Companies have outside-facing functions and inside-facing functions (Fig. 1). Sales and marketing are examples of outside-facing functions. They are the company’s sensors. They bring information into the company to help with product development and decision-making. Engineering and design are examples of inside-facing functions. They allow the company to act on outside information by creating products and services. Program or project management are cross-functional groups that coordinate this information flow. According to a study by KPMG, executives cite coordination as the number one challenge for their companies.

Coordinating human workers and AI agents is even more complex. Researchers like Andrew Ng describe how AI agents will perform many workflows in the future. According to the venture capital firm NFX, “US businesses spend upwards of $5 trillion on knowledge workforces. By comparison, companies spend about $230 billion on B2B SaaS.” A system that harmonizes the outputs of humans and AI agents is the key to unlocking the value at the intersection of knowledge work and B2B SaaS.

Companies have shortened the loops between sensing, deciding, and acting through agile software development. In agile, products are developed iteratively through cycles of building and testing. Agile can help companies arrive at product-market fit (PMF) much faster, especially when implemented alongside a strong product vision. Companies that do not expose themselves to user feedback until product delivery often fail. Steve Blank, the legendary Stanford entrepreneurship professor, says that “there are no facts inside the building.” He implores students to interact with the outside world (customers, suppliers, partners, etc.) as much as possible to feed their product and company development.

Bringing world data into the company is also critical for strategic decision-making by executives. These decisions may involve mergers and acquisitions (M&A), supply chains, partnerships, manufacturing, etc. Strategic decisions are generally made by executives at the top of the organization, and tactical decisions are made at the bottom by individual contributors (ICs) or managers.

Subscribe now

Momentum removes the need for coordinating

This inside/outside dichotomy existed where sensing, acting, and deciding were separate activities. What if they became simultaneous and synonymous? In such a world, companies could sense as they were acting and act as they were sensing. These companies would react and change directions instantaneously in response to stimuli.

Nature provides helpful analogies for our product (Momentum). A cheetah chasing a gazelle does not waste time thinking during the chase. It also does not waste energy by acting without purpose. Doing so would diminish its limited resources. The cheetah chases after its goal, and it changes directions instantaneously in response to new information. It senses, decides, and acts simultaneously. There is no separation between inside-facing and outside-facing units of the cheetah. The cheetah’s sensing units (eyes), its acting units (legs), and its deciding unit (brain) are all connected as a single system through its central nervous system (CNS).

For the first time, Momentum’s AI gives this capability to companies. I wrote earlier about how we built an AI prefrontal cortex (PFC) to capture the company’s world model. We build this model using information sensed inside and outside of the company. Some of these sensors are Generative AI systems, which allow us to capture the contexts of text, images, videos, and other pieces of information. The world model automatically updates goals in response to the streams of information it receives. Sensing, acting, and deciding turn into a unified process. It also integrates into all of the company's tools, allowing the company’s workers to react to updated goals immediately. The cheetah becomes one of the “auto communicators” about which I wrote previously. In this article, I describe “what” we are building. In a future article, I will describe “how” we build it.

Our company has learned from nature and modeled many of our systems on natural phenomena. However, our imagination and the power of AI have allowed us to go further than the limitations of a single physical body. The cheetah can only resolve its needs serially. It can hunt, drink water, rest, or find a mate one after the other. There are no such physical limitations in the digital world. The digital cheetah (company) can simultaneously receive help from multiple AI agents (Fig. 2). The company pursues its goals while its agents find new opportunities, resources, partnerships, threats, etc. Information from the world is invaluable for making decisions. Momentum constantly aligns the goals of the agents with those of the company. This paradigm is the world of harmonized auto-communicators that I described earlier.

Momentum is building the operating system of work

According to the Wall Street Journal, workers spend 57% of their time coordinating with each other, instead of discovering or acting on information (Fig. 3). They find this coordination work frustrating, and they feel that they are ineffective at it. We spoke with 1000 professionals last year (600 executives and 400 ICs) to understand the barriers to productivity. The frustration with coordination was clear at every level of the organization. Companies like Microsoft have developed tools like Copilot that are intended to improve the productivity of individuals. However, the real barriers to productivity are organizational, not individual.

What are workers doing when coordinating? They are aligning information flows (sensing), decision-making, and execution (acting) around company-wide and individual goals. Solving all of these issues simultaneously is the only way to solve them properly. It is also the only way to build the immediate feedback loops of the earlier cheetah analogy. We are building the operating system of work to remove the need for coordination, and simultaneously solve the problems of sensing, deciding, and acting within companies. We refer to it as an operating system because it handles all of the coordination necessary between workers, tools, and AI agents. Using Momentum, workers can focus all their attention on creativity, which they strongly prefer over coordination.

How engineers experience our operating system of work

Most of the obstacles that engineers face are about coordinating information flows. Engineers need code documentation, vendor APIs, open-source project updates, and details of new technologies from outside of the company. At the same time, they need recent goals, engineering dependencies, updates from peers, and details of the code base from inside the company. Finally, they need to use available tools to develop code, document it, communicate it to peers, and receive feedback. The description above shows some of the sensing, deciding, and acting necessary for engineers to complete their work.

Momentum relates each engineer’s individual goals to the company’s organizational goals. The system provides the right external and internal information to the user at the right time to make decisions without sensory overload. Momentum integrates with all of the tools required by the worker. These pieces remove the need for coordination, and they reduce friction to the point that sensing, deciding, and acting become continuous processes. Taken together, Momentum acts as the operating system of work for this engineer.

Other functions (design, sales, marketing, etc.) require their own information flows and custom views of the Momentum operating system. Momentum connects all of these information flows to create a single unit that simultaneously senses, decides, and acts. This unit has AI agents outside that help with sensing information from the world, and AI agents inside that help with acting on this information. Harmonizing the goals of these agents is key to Momentum’s approach. Humans are in the loop at each step to make decisions and evaluate the work of agents.

Limerick summary

"In the dance of work, coordination's plight,
Agile shortens loops, but still, a fight.
Momentum, the OS, brings harmony's light,
Sensing, acting, deciding unite.
From engineers to execs, productivity takes flight."

Subscribe now

From Tehran to Tennessee

"I,
empty of love and anger,
empty of self and other,
floating delirious,
between being and non-being."
~Khali (Empty) by Ebi

A combination I like

A scene I like

Jack: “Is that what a man looks like?”

Tyler: “Self-improvement is masturbation. Now, self-destruction?!”

Oakland and its surrounding towns (Berkeley, Alameda, San Leandro, etc.) are located in Alameda County. Alameda County stretches along the east bank (East Bay) of the San Francisco Bay from Berkeley in the North to Fremont in the South. Some of the most prominent deep tech startups in the energy transition are located in Oakland and its surroundings (Fig. 1). I have not seen a similar density in other energy transition hubs (i.e. Boston, Houston, etc.) This high density will increase the momentum of the region as the center of the deep tech energy transition. Oakland has gained this role because of its central location, a history of heavy industry, proximity to educational and capital resources, and relative affordability. Disclaimer: I love Oakland and live nearby, which may affect my analysis. Nevertheless, I have provided the evidence necessary for readers to make their judgments.

Subscribe now

Location, location, location

The proximity to UC Berkeley and the Lawrence Berkeley National Lab (LBNL) has been critical to the role of Oakland as an energy transition hub. These centers have produced a steady stream of scientists, engineers, and business professionals to start and grow such startups. In particular, the Activate program at LBNL has incubated Cuberg, Inlyte, Twelve, Sepion, Fervo, Brimstone, Limelight, and Form among other successful startups. Founders and early employees who moved to the area for Activate’s two-year incubation period have tended to put down roots and grow their companies in the region.

Stanford University and Silicon Valley VCs are just a 45-minute drive away. Stanford has also been a source of engineers and entrepreneurs. Companies like Cuberg, Inlyte, Sila, Fervo, Molten, Twelve, and others were founded by Stanford alumni. The VC firms surrounding the university (DCVC, Capricorn, Khosla, A16Z, Sutter Hill, etc.) have been major sources of funding for Oakland area startups. The proximity of Silicon Valley VCs has been critical to the growth of the area’s startup sector. Some climate-focused VCs, like Powerhouse VC, are even headquartered in Oakland. I consider Powerhouse’s Watt It Takes podcast tied with Shayle Kann’s Catalyst podcast as the top two podcasts about energy transition startups. Watt It Takes episodes are often recorded in Powerhouse’s Oakland offices.

The East Bay (i.e. Alameda County) is more affordable than Silicon Valley (i.e. Santa Clara County). The price of housing is one of the most important contributors to this affordability gap. The median housing prices in Oakland ($710k), Alameda ($1.1M), and Berkeley ($1.2M) in Alameda County are lower than those in Santa Clara ($1.6M), Mountain View ($1.6M), and Palo Alto ($2.8M) in Santa Clara County. Silicon Valley hosts the fastest-growing and highest-margin startups, which are mostly in software. Heavy industries generally have lower margins and slower growth than enterprise and consumer software companies. The relative affordability of the Oakland area helps deep tech startups attract and retain their workforce.

All six routes of the Bay Area Rapid Transit (BART) train system are centered in Oakland (Fig. 2). The BART connects the Bay Area’s three major cities, two of its major airports, and many of its outer suburbs. Oakland can easily attract workers residing throughout the Bay Area because of its location at the center of the BART system. Oakland is also home to engaging museums, good food, fun bars, diverse cultures, interesting hikes, and stimulating activities. It is the kind of city that attracts the energetic and adventurous types who enjoy working in startups.

The Matthew effect of accumulated advantage

Social scientists refer to the accumulation of advantages to a person or group as “the Matthew effect.” The principle is named after the parable of the talents in the book of Matthew: “For to everyone who has will more be given, and he will have abundance; but from him who has not, even what he has will be taken away.” The Matthew effect also applies to regions. New York City attracts new financial institutions, Los Angeles attracts new movie studios, and the Bay Area attracts new tech companies. Oakland has a history of heavy industry, which is partly responsible for the current wave of startups. I expect that the city’s advantage in the deep tech energy transition will grow as these startups mature.

West Oakland is home to the Port of Oakland, which is one of the busiest ports on the West Coast. It is also at the intersection of major East-West and North-South rail lines. Its position on the San Francisco Bay and access to transportation made the region the center of the Bay Area’s heavy industries in the 19^th and 20^th centuries (Fig. 3).  The automotive industry bloomed in the region starting with a Chevrolet plant in 1916. A map of the area from 1926 shows companies like Commercial Acetylene Supply, Detroit Steel Products, Westinghouse High Voltage Insulator, Penzoil, and Linde Air Products among others. West Oakland has been economically damaged from the offshoring of the country’s heavy industries in recent decades. However, it still has many space, manufacturing, transportation, and licensing resources for hosting such industries.

The automotive industry continues its presence in Alameda County with the Tesla plant in Fremont, CA. Bay Area battery companies like Sila Nano and Our Next Energy (ONE) are automotive industry suppliers. Chevron, one of the world’s largest oil & gas (O&G) companies, is headquartered in the Bay Area (San Ramon, CA). This O&G giant operates a major refinery in Richmond, CA, which is just north of Alameda County. Richmond is also the headquarters of SunPower, a clean energy pioneer, with expertise in semiconductor manufacturing for photovoltaics. I have met many engineering leaders in Oakland’s deep tech startups who used to work for one of these two companies. The chemistry and high-temperature manufacturing expertise of Chevron and SunPower, respectively, has helped grow these deep tech startups. 

Within the Bay Area, Alameda County has some of the clearest rules on the operations of industrial facilities (e.g. rules on the handling of industrial solvents). Just as Delaware’s clarity on business law has made it the preferred state for incorporating US companies, Alameda County’s clarity on heavy industry has made it the preferred destination for the Bay Area’s deep tech startups.

Discussion

Despite the hype, we are still in the early stages of the energy transition. Changes in the energy industry are measured in decades, not years. Tesla introduced its first mass-market vehicle (Model 3) in 2016. The oldest startups listed in Fig. 1 (Sila Nano and Twelve) also began operating in the East Bay within the past decade. Nevertheless, according to the “Matthew effect,” the high density of deep tech energy transition companies in Oakland and surrounding towns will beget even higher density in the future. I expect that Oakland, CA will be seen as the capital of the energy transition in the years to come.

The fundamentals of the area’s proximity to education, capital, transportation, and manufacturing resources will remain the same. In addition, the region will continue to build expertise in relevant fields. For example, electrochemistry has become the king-science of the energy transition. It is the core science of battery companies like Cuberg, sustainable aviation fuel (SAF) companies like Twelve, and metal processing companies like Magrathea. The LinkedIn profiles of engineers at many of these companies show that they have worked in other startups in the area. The cross-pollination of talent between these startups will mean that the most experienced electrochemists in the world will reside in the Oakland area.

Oakland’s position as an energy transition hub has been helped by its proximity to Silicon Valley VCs. Other parts of the Bay Area have seen similar specialization of industries. For example, there is a concentration of biotech companies in South San Francisco and a concentration of semiconductor companies in Santa Clara. Many of these companies also draw their funding from the VC firms located in Palo Alto and Menlo Park. As the Oakland energy transition hub matures, I would expect that more energy- and climate-focused VCs, like Powerhouse, will appear in Oakland.

Subscribe now

From Tehran to Oakland

"No damage
So Oaklandish
Deep rooted
In the town
Where we compound soul
And street music"

A combination I like

A short movie I like

The Life Age is starting. Generative AI tools are the enabling technology of this age. They allow formerly mute processes like AI agents or organizations to come alive. These units can “auto communicate” to connect automatically. The Life Age can only scale if these units are harmonized. Two features are required for harmonization:

1. The units must act based on their goals.

2. The units must share a medium of communication.

We are building both of these features at Momentum. We have invented the field of Emergent AI for this purpose. Emergent AI harmonizes the actions of AI agents to allow a higher-level behavior to emerge.

Subscribe now

Background: The Information Age required curation

Some inventions are so powerful that they define human eras. The period before the invention of writing is called “prehistory.” The period following the invention of the steam engine is called the Industrial Age. Computers and the Internet began the Information Age. The ubiquity of these technologies is responsible for this name. The total number of people using the Internet increased from 2.6M in 1990 to 5.4B in 2024.

Most adults are connected to the Internet and most data is digitally transmitted. Navigating the large number of connections and data has required curation. Search engines like Google are gateways to the Internet, social networks like Facebook help manage relationships, and e-commerce platforms like Amazon curate our product discovery. The Information Age has created a feedback loop between connections and curation. More curation is needed as connections increase, and the ability to curate allows for more connections.

Problem: The Life Age requires harmonization

Generative AI models are “super communicators.” They are the enabling invention of the Life Age. They have not delivered on their initial promise so far. High-flying startups are being acqui-hired, researchers are questioning their economic impact, and big tech is tempering expectations. This shortcoming is caused by the use of Gen AI for curation (search) in an Information Age paradigm. Their true value is to enable “auto communication” between formerly lifeless and mute processes (AI agents and organizations). Humans feel more powerful when these units come to life to serve them.

Achieving the higher-level goals of AI agents and organizations requires harmonization. Biological processes offer helpful examples of harmonized behavior. The video below shows a white blood cell pursuing a bacterium. Neither cell has eyes or a brain. However, they each share three features:

1. Goals: Survival for the bacterium and pursuit for the white blood cell.

2. Action: Movement through the liquid.

3. Communication: Awareness of each other through chemicals in the liquid.

We harmonize auto communicators (e.g. AI agents) by allowing them to pursue and communicate their goals. The Life Age creates a feedback loop between auto communicators and harmonization. More harmonization is needed as AI agents increase, and the ability to harmonize allows for more AI agents.

Momentum harmonizes similarly to the Prefrontal Cortex

We are addressing the need for harmonization in the Life Age. Coordinating the outputs of humans was already a challenge in the Information Age. Workers spend 57% of their time on coordination. Executives cite “super alignment” as their most desired superpower (we interviewed 1000 professionals). It is more difficult to harmonize (align) the outputs of groups of humans and AI agents in the Life Age. Our solution (Momentum) harmonizes around goals. Humans perform a critical role in our solution: they dream, set goals, and lead. Our solution harmonizes the goals of groups of humans and AI agents to accomplish their higher-level group objectives.

Using the brain as an analogy (Fig. 2), our solution functions like the prefrontal cortex (PFC). Broca’s area is for understanding speech. ChatGPT maps well to this area, as it relates the contexts of language inputs and outputs. The Visual Cortex is for understanding images. Dall-E maps well to this area, as it relates the contexts of language and images. Animals have similar areas in their brains for auditory and visual inputs. What differentiates us as humans is our larger PFCs, the centers of planning and action. We have larger PFCs for two reasons:

1. We perform more complex planning and acting than animals.

2. We are social animals. Planning and acting are more complex in social settings.

Our PFC uses the other parts of the brain as sensors and tools to model the “self” and the world. These abstractions help with planning and allow the PFC to predict the results of its actions. Our solution sits on top of Gen AI models. In addition to building a world model, Momentum provides the Gen AI models with the most accurate “context window” based on the current goals. An accurate context window is critical to obtaining the most relevant results from Gen AI models. Building Momentum has required the invention of a new field distinct from Gen AI. We have come to call this field Emergent AI. At Momentum, we are first applying this field to bringing organizations to life and making them antifragile.

Subscribe now

From Tehran to the Subcontinent

A combination I like

A scene I like

Father: “You boys are about all I’ve done in life. Otherwise I’ve drawn zilch. You are all I have. You’re all I want to have. Sweet boy.”

I want to understand the strategic choices that successful startups make to win markets. My articles discussing moats or pricing strategies are parts of this exploration. A friend told me that “7 Powers: The Foundations of Business Strategy” is a favorite of Peter Thiel and the Founders Fund universe (Fig. 1). I have read the book and been impressed with the insights and the framework. The book first defines Power as “the set of conditions creating the potential for persistent differential returns.” This definition is similar to the concept of a “moat.” The book then defines strategy as “a route to continuing Power in significant markets.” The book explains that the two most important attributes of Power are the magnitude and duration of the differential returns. In this article, I present these seven types of moats with examples of startups in the energy transition.

Moat 1 – Scale economies: Northvolt

A “scale economies” moat is “a business in which per unit cost declines as production volume increases.” Certain fixed costs like engineering design or marketing remain constant regardless of the number of units produced. The per-unit cost of these activities decreases at higher volumes. The book argues that this relationship can be an effective moat for different businesses. New entrants would lose money if they produced at lower volumes, which would dissuade them from entering the market. Scale economies can act as a moat in other contexts as well. For example, the distribution network density of UPS and the purchasing economies of Walmart act as a barrier to new entrants into their respective markets.

Subscribe now

Northvolt is an automotive battery manufacturing startup that is building such a moat. The company was founded in 2015 and it has raised $9B. Northvolt does not have a differentiated technology, business model, or distribution network. However, it entered the market when there were few European automotive battery startups. It picked an already-established technology and scaled rapidly to reach a valuation of tens of billions and 4400 employees within just eight years. TechCrunch reports that Northvolt has “Secured more than $55 billion in orders from customers like BMW, Fluence, Scania, Volvo and Volkswagen.” The founders of Northvolt were previously supply chain executives at Tesla. They both had experience building large organizations. The scale of Northvolt means that the European market will likely not see another entrant unless the startup aims for a different type of moat.

Moat 2 – Network economies: Dandelion and Rana Energy

A “network economies” moat is “a business in which the value realized by a customer increases as the installed base increases.” This type of moat has dominated consumer software for the past two decades, especially in social networks and two-sided marketplaces (Uber, Facebook, Venmo, etc.). In this scenario, individual users benefit from the large number of other users on the platform, which disincentivizes them from moving to a new and less popular one.

I had trouble thinking of many energy companies that have a network economies moat. Such moats are most associated with consumer media or service businesses where there is little transactional friction. The energy sector is highly regulated, which increases transactional friction. In addition, services often depend on large infrastructure like the electrical grid or major companies like electricity suppliers and not just individual consumers or service providers. The first business model of Dandelion Energy depended on network economies. The company was a two-sided marketplace that connected homeowners with independent contractors capable of installing geothermal heat pumps. The company has since changed that model and now employs the workers directly.

Rana Energy plans to build two-sided marketplaces for electricity production, storage, and consumption on residential microgrids in Nigeria. Developed countries have historically built their infrastructure around producing electricity centrally and distributing it to individual homes (Fig. 2). The distributed storage and solar production now available can conflict with this centralized model. Many developing countries leapfrogged landline telephones and directly adopted cellular phones. Rana wants to repeat this leapfrogging for decentralized electricity. It plans to build a two-sided marketplace for electricity where individual households can buy and sell electricity from each other on a microgrid. Such a two-sided marketplace is possible in a developing country where less infrastructure and regulation reduce transactional friction.

Moat 3 – Counter-positioning: Sunrun

A “counter-positioning” moat forms when “a newcomer adopts a new, superior business model which the incumbent does not mimic due to anticipated damage to their existing business.” Counter-positioning is not the same as a disruptive technology. A disruptive technology "does not necessarily lead to… [a moat], but it can sometimes create the circumstances in which… [a moat] may be established.” Counter-positioning is about a new and better business model that the incumbents ignore until it is too late.

Sunrun (founded in 2007 and market cap of $2.1B) has revolutionized the residential solar market by helping customers buy electricity through power purchase agreements (PPAs). In the past, homeowners who wanted solar energy had to pay for the system upfront (~$50k in 2008). Instead, Sunrun owns the panels for ~20 years and sells electricity to the homeowners at a fixed rate over that period. Offering financing, and especially PPAs, for purchasing home solar and battery is now the dominant business model in this sector. SunPower, Sunnova, and Vivint Solar all offer multiple options. Sunrun’s approach has been adopted by other home energy solutions. For example, Dandelion partners with banks to provide financing for its geothermal heat pumps. This business model innovation has contributed to a 100X increase in the annual residential solar capacity installation in the US since 2007.

Subscribe now

Moat 4 – Switching costs: Terabase

A “switching costs” moat is “the value loss expected by a customer that would be incurred from switching to an alternate supplier for additional purchases.” The author presents “SAP’s [market cap of $167B] paradoxical combination of high retention and low satisfaction” in the enterprise resource planning software (ERP) space. Fundamental company operations like receiving, tracking, and fulfilling orders rely on ERP software. It is risky and expensive to switch ERP systems after adopting the software, integrating it into the organization, training employees on it, and building relationships with the vendor. Most companies choose to stay with SAP and even purchase additional tools from them to avoid this risk. These high switching costs act as a moat for SAP.

Terabase (founded in 2019 and raised $77M) is building a suite of software tools for the development, construction, and operation of utility-scale solar power plants (Fig. 3). Their tools look useful, but I do not see radical technical advancements that would provide a technology moat. Altogether, the software suite reduces the cost of building and operating these facilities. I like their creative product strategy. Some companies sell solutions for a single area and then convince the user to buy their adjacent product offerings. For example, if a customer buys an order tracking solution, they may also purchase an adjacent solution to predict restocking needs. Terabase products are adjacent in time. If a customer uses Terabase for the development of a new plant, they will be incentivized to purchase additional tools for the subsequent construction and management phases. Terabase’s moat is based on this desire to avoid switching costs.

Moat 5 – Branding: Span

A “branding” moat is “the durable attribution of higher value to an objectively identical offering that arises from historical information about the seller.” A branding moat builds over time, as the company delivers a high-quality product and builds trust. Consumers buy branded products to associate with a particular identity and reduce the uncertainty in their purchasing decisions. After building a branding moat the company can charge a premium for its product.

Span (founded in 2018 and raised $230M) is building a branding moat around its smart electrical panels. I have written before about how Span used a significant portion of its $3.5M seed round on industrial design. Arch Rao, the founder and CEO of Span, said that,

It's very unusual for a seed-stage company, with a few single-digit millions in the bank, to go spend money with an external industrial design company. But from the very beginning, it was super important to us to build a… hardware and software product that was stunning, and I think we’ve achieved that.

Span’s panel looks symmetric, sleek, and compact compared to the product from Square D (Fig. 4), a subsidiary of Schneider Electric (founded in 1836 and a market cap of $88B). It offers convenient features like lights on the sides of the panel to increase visibility. Finally, there is a large company logo at the top of the panel, just as Apple products prominently display their logo. All of these features point to a branding moat around a luxury product. Interestingly, both companies charge $3,500 for their panels. I think there are two reasons for this decision. First, according to the author the ability to charge customers a branding premium forms over time. Second, these electrical panels are intended to last for decades and have high switching costs. Span can generate revenues from selling services and software after the hardware installation. This future revenue incentivizes the company to lower the barrier to initial adoption.

Moat 6 – Cornered resource: KoBold Metals

A “cornered resource” moat is “preferential access at attractive terms to a coveted asset that can independently enhance value.” A cornered resource may be physical, like exclusive access to raw materials. The resource may be technological, like exclusive rights to intellectual property (IP). The resource may be regulatory, like a license from the Nuclear Regulatory Commission (NRC), which has not been granted in half a century. The resource may simply be information. Many artificial intelligence (AI) companies build “data moats” that ensure them the best-trained models compared to competitors. Companies can even corner human resources by hiring the best in a particular field. The compensation of individuals with in-demand skills increases dramatically as companies compete to hire them. A typical engineer at OpenAI (maker of ChatGPT) earns $1M annually.

KoBold Metals (founded in 2018, raised $400M, with a valuation of $1.15B) is building multiple of these moats. The company trains AI models on geological and Earth observation data to discover and develop metallic mineral deposits. These metals are critical to the energy transition. The company’s AI models are proprietary, which is one cornered resource. It has received investment from the oil company Equinor, partly to gain access to their geological data, which is another cornered resource. Combining these resources and the investment from Andreessen Horowitz I expected that they would have a SaaS business model. However, they take a stake in their mineral finds, jointly develop the mine with partners, and generate revenue from selling metal. Mineral rights are another cornered resource. Experts in this field have told me that companies relying only on discovery can be squeezed easily, but those with mineral rights have much more power.

Moat 7 – Process power: Tesla

A “process power” moat is “embedded company organization and activity sets which enable lower costs and/or superior product, and which can be matched only by an extended commitment.” The author emphasizes that this moat is not merely operational excellence, which can be mimicked. This power captures the unwritten ways that organizations operate and those within them relate to each other. These relationships may allow companies to gain a sustained advantage over time. I think this moat is akin to “company culture.” Even companies with nominally the same organizational structure often operate differently based on their cultures (Fig. 5).

The book uses the example of the NUMMI plant in Fremont, CA. NUMMI was jointly operated by GM and Toyota starting in the 1980s so that they could learn from each other. Toyota was able to gain from this experience, but GM executives had difficulty replicating the Toyota Production System (TPS) in other plants. Tesla took over that Fremont plant in 2010. Tesla is large enough and old enough to have built a process power moat. Employees often point to the intensity, personal responsibility, and pace of change at the company. A former Tesla engineer told me that “Tesla [59k employees] is a big company that moves like a startup.” These characteristics are part of the reason that Tesla has a process power moat (among other moats). Tesla has a price-to-earnings (P/E) ratio of 67, which is an order of magnitude larger than that of GM (3.9), Ford (4.6), and Toyota (11).

Discussion

These moats operate on a different plane than the dual framework of technology and business moats. For example, Northvolt’s scale economies is largely technological, whereas Walmart’s scale economies is a business moat based on purchasing power. Companies can also build moats in multiple categories. For example, Tesla has a counter-positioning moat (direct sale to consumers), a branding moat (luxury product), a cornered resource moat (preferred access to battery materials), as well as a process power moat.

Investors, employees, and strategists need to be aware of these categories. A former executive who previously ran a large organization may be more successful in building a scale economies moat. Instead, a founder pursuing the cornered resource of exclusive IP may benefit from a deep technology background. Traction must be measured somewhat differently in each category. Companies pursuing scale economies may quickly raise capital and grow their teams. However, this rapid expansion may be the sign of an undifferentiated product that scales unencumbered by technology development. A differentiated technology may provide a better long-term moat than a highly scaled one.

These categories are important for thinking through market dynamics. It is likely too late to build another European battery company aiming for a scale economies moat. The newcomer would find it difficult to compete with Northvolt. However, a newcomer with exclusive access to IP, raw materials, or even one with a novel business model may still be viable. According to Richard Rumelt, “strategy is the application of strength against weakness.” It is important to know where the incumbent is strong for the newcomer to avoid that territory.

Question

What is your favorite moat and which company is a good example of this moat? Tell me in the comments!

Subscribe now

A song I like

"Everything will disappear but
The wind will carry us"

Hot pilates beat

Featuring Idris Elba (aka Stringer Bell)

A scene I like

Hannibal Lechter: “He covets. That is his nature. And how do we begin to covet, Clarice? Do we seek out things to covet?… No. We begin by coveting what we see every day.”

Artwork credits

House and power plant

The Department of Energy (DOE) has produced a series of “Liftoff Reports” to “create a common fact base and a tool for ongoing dialogue with the private sector on the pathways to commercial liftoff” of clean energy technologies. The Liftoff Report about “Low-Carbon Cement” points out that “Cement represents ~7–8% of global CO₂ emissions and ~1–2% of U.S. CO₂ emissions.” This report was the main topic of conversation between Jigar Shah of the DOE and Brimstone CEO Cody Finke for the Demonstrate Deploy Decarbonize (#Deploy23) event in September (Fig. 1). Brimstone was founded in 2019, and it has so far raised $60M over two rounds from Breakthrough Energy Ventures, DCVC, Fifth Wall, and other investors. I love Brimstone’s approach to cement decarbonization because of its i) elegance, ii) timeliness, iii) low business risk, and iv) the founders’ mindset.

A new recipe for Ordinary Portland Cement

Cement is the key binder in concrete, which is one of the most common man-made materials in the world. Different types of cement exist; however, Ordinary Portland Cement (OPC) is by far the most common form. It constitutes 95% of the cement produced in the United States. Calcium-based minerals are key ingredients in OPC. Historically, the main feedstocks for cement production have been limestone or gypsum, which are both rich in calcium (Fig. 2). As I have written before abundance is a key requirement for the adoption of industrial-scale materials. Both of these feedstocks are abundant globally. The process that uses gypsum also produces sulfuric acid. In the past, this co-product was a useful secondary source of revenue for cement producers. However, there are now other ways of producing sulfuric acid. The need to safely discard this material has caused cement producers to turn away from gypsum as a feedstock.

Subscribe now

Limestone has become the primary source of calcium for making OPC because it does not produce the same caustic co-product as gypsum. However, the calcium in limestone is attached to carbonate, which is released as CO₂ when heated to 1,500 °C. This release accounts for more than half of the CO₂ emissions of cement. Another 34% of the emissions are due to the fuel that is burned to heat the cement kilns in this step. Brimstone has invented a new recipe for OPC, which uses calcium silicate rock as the feedstock. Over the Summer the company received the ASTM C150 certification for its OPC recipe. This feedstock not only avoids producing harmful co-products but also leads to valuable Supplementary Cementitious Materials (SCM) and Magnesium compounds. SCM is one of the components of concrete (in addition to cement), which would otherwise have to be purchased separately. Magnesium absorbs CO₂ from the atmosphere, which makes this process carbon-negative for a variety of heating processes. Brimstone has created an elegant solution that avoids the worst parts of the current process while improving the economic outcome because of the valuable co-products.

Why now? A secular shift in SCM supply

SCM is used as a partial replacement for cement in mixing concrete. Different materials can be used as an SCM as long as they are rich in silicon and aluminum. Fly ash, which is a residue of burning coal, has been one of the most common SCMs in concrete production. Coal consumption in the US nearly tripled over half a century, since the 1950s. This increase was driven by the use of coal for electricity production (Fig. 3). Starting in the mid-2000s US coal consumption began a precipitous decline, where it now stands at less than half of its peak. US natural gas production has doubled during the same period because of fracking. Natural gas has replaced coal as the dominant source of energy used for electricity production in the US.

The phasing out of coal power plants has significantly reduced the production of fly ash in the US. The US now imports fly ash from China to be used in concrete as an SCM. Fly ash represents a multi-billion dollar industry. The fact that Brimstone co-produces an SCM along with their cement makes their product more economical. As Cody Finke told the CleanTechies podcast, “[If] supplementary cementitious materials were abundant, then our process would probably be too expensive. But because they're not abundant, because of this transition away from coal, our process, we believe, would be lower cost at scale.” Brimstone’s approach is right on time, because of this secular shift in the supply of SCM.

Subscribe now

Brimstone has low business risk

Brimstone is aiming for an already-existing cement market, which is valued at $340B globally. Brimstone is a quadrant 4 company, and it faces a low business risk (Fig. 4). As I have written before, I like this risk profile because there is no question that if they are successful at scale the market will adopt their product. Some of the other cement decarbonization companies plan to charge a “Green Premium” for their products. I have written before that quadrant 4 companies cannot charge a Green Premium at scale, because they cannot charge for a “sustainability” feature. Companies in this quadrant can only differentiate their products on price or specifications, and not on features. Brimstone understands this fact and it is acting accordingly. Cody told the CleanTechies podcast,

I don't know too many folks that want to buy more expensive concrete, but I do know a lot of folks that want to buy clean concrete… If you can make this the same material for a lower price point without the CO₂ emissions, or even the same price point… that's a huge win. But any more expensive… it's not a win for the world cement industry.

Brimstone’s founders have the right mindset

I love listening to Cody talk about Brimstone’s journey and approach. He is aware of the pitfalls of deep tech climate entrepreneurship. These entrepreneurs are often idealistic about their technologies and carbon reduction missions. I greatly admire them personally, and I hope that they succeed in their work. However, when I mainly hear technology specs or carbon reduction goals from some founders, I wonder if they are sufficiently focused on the economic drivers of product adoption. What we verbalize is a sign of what we think about. Why would a customer buy this product at scale? Will it ever be cheaper or better than the incumbent? Is it risky to assume that the customer’s decision will be driven by climate considerations instead of economic ones? Instead, in the same interview Cody talks about minimizing economic risks. He says that Brimstone has

An unrelenting attitude of if it doesn't work in today's economic assumptions, then the process is not good enough… we're not multiplying risks that we can't control together… like we realized, ‘Oh, this process is great, except electricity has to be 24 hours a day and clean and 0.1 cents per kWh. Oh, this is great, but we need a $50 per ton carbon tax. Or, oh, this is great, but we just need to ship rock from Hawaii all over the world’… a lot of folks will either maybe fool themselves a little bit with assumptions which is really easy to do because you love what you're doing… or they will assume things that are really risky.

As Richard Feynman put it, “The first principle is that you must not fool yourself and you are the easiest person to fool.” Reducing emissions is important for the health and safety of humanity globally. Building a venture-backed startup can be a powerful tool for addressing this issue. However, the common rules of business will still apply to companies in this sector. A beautiful technology or an important cause will not automatically lead to customer adoption if the economics do not make sense. As Cody put it, “We're all trying to build businesses here and you're not going to change the world or solve a climate problem or whatever if your business doesn't go anywhere.” That is the right mindset for building a deep tech climate startup, and it is one of the reasons that I love Brimstone.

A song I like

Oh, father of the four winds, fill my sails 
Across the sea of years 
With no provision but an open face 
Along the straits of fear 

A scene I like

Don Draper: “Get out of here and move forward. This never happened. It will shock you how much it never happened.”

Look at the light on Don’s face at the end: half lit and half dark. It is reminiscent of the lighting in The Godfather.

Earlier this month the Department of Energy (DoE) published a report titled “Pathways to Commercial Liftoff: Virtual Power Plants (VPPs).” VPPs are “Aggregations of distributed energy resources (DERs) such as rooftop solar with behind-the-meter (BTM) batteries, electric vehicles (EVs) and chargers… and flexible commercial and industrial (C&I) loads that can… provide utility-scale and utility-grade grid services like a traditional power plant.” The report notes that “With electricity demand growing… and fossil assets retiring, deploying 80-160 GW of [VPPs]… by 2030 could support rapid electrification while redirecting grid spending from peaker plants to participants and reducing overall grid costs.” We can use cellular networks as an analogy to better analyze the product strategies of companies involved in the residential VPP market.

Subscribe now

Cell networks as a residential VPP analogy

Cellular networks are ubiquitous (Fig. 1). Network operators AT&T, Verizon, and T-Mobile have a combined 475M US subscribers. A majority of the cell phones on these networks are Apple iPhones. Furthermore, each iPhone may connect to peripherals like Bose headphones or host dozens of applications made by companies like Google, Netflix, and Meta. The cellular model shows that successful products can exist at each level (network, device, and application) of the system. However, each level and type of product (software versus hardware) offers different strengths and weaknesses.

We can use cellular networks as an analogy to better understand the large number of companies active in the residential VPP space (Fig. 2). These companies operate at the network, household, and application levels. This analogy is particularly useful now that electricity moves bidirectionally between the grid and homes, and it can even be stored at home (in batteries or EVs), much like data can be stored in phones.

Weavegrid

Weavegrid helps utilities balance the load on their systems by managing the charging patterns of EVs on the grid. In this analogy, the utility plus Weavegrid is the equivalent of AT&T. Apoorv Bhargava is the CEO and co-founder of Weavegrid (full disclosure: I know Apoorv personally). As he explained on the MCJ podcast “You as an individual actor can do all the right things that you believe are the right things. And yet, at the same time, it may not actually tackle the system's problem.” The average US household uses 30 kWh of electricity in a single day. A Chevy Blazer EV has a battery capacity of 80 kWh, and the new Chevy Silverado EV has a battery capacity of 200 kWh! Adding EVs to the electric grid can increase demand by up to an order of magnitude, which is why Weavegrid has focused on mitigating this issue. In the cellular analogy, this work is the equivalent of managing video streaming at the network level, which accounts for 2/3 of web traffic.

SPAN

SPAN designs and builds smart electrical panels, which are increasingly required when homeowners add solar, battery storage, or EVs to their homes. The SPAN panel is the single point through which all electricity within the house flows, and where the exchange between the house and the grid takes place. SPAN has some features for adding resiliency to the household in case of a power outage, but it does not turn the house into a VPP. As I have argued before adding such a feature can not only act as a source of revenue but also act as a protective measure against encroachment from Tesla. I believe that SPAN can play nearly as big a role in the residential electricity network as the iPhone plays in cellular networks.

Lunar Energy

When I first heard of Lunar Energy, I was most fascinated by its fundraising model and product. The company raised $300M in two tranches from strategic investors Sunrun (solar) and SK Group (batteries) before exiting stealth mode with a product and hundreds of employees. Their physical product combines solar generation with battery storage. This combination is offered by other players like Generac, SunPower, and Qcells as well. However, Lunar also offers Gridshare, which is an “energy management platform, designed to optimize millions of DER.” Lunar simultaneously offers the hardware to turn millions of homes into VPPs, as well as the grid-management software to maximize the value of the VPPs for the grid and the homeowners. The fact that Lunar’s software product integrates with third-party storage, solar, EV charging, and HVAC makes it even more powerful.

General Motors (GM)

Just as Weavegrid has understood the importance of EV charging and approached it from the grid, many car companies have approached this problem from the EV. General Motors (GM) is one of those companies. An interview with Ty Jagerson, leader of V2X at GM, on the Catalyst podcast went viral when Jigar Shah of the DoE tweeted about it (Fig. 3). The episode explains that there is “managed charging, or V1G, and vehicle-to-grid, or V2G. There’s also a third technology called vehicle-to-home [V2H] that allows an EV battery to power a building, just like a home battery. Collectively these technologies are called V2X.” V2H can be used to add energy resiliency to the home, while V1G or V2G can reduce the strain on the grid or allow the home to act as a VPP. Once again, the large capacity of EV batteries is the key to their importance in this context, and why GM and other car companies are increasingly active in this area.

OhmConnect

OhmConnect was founded in 2013, and it precedes all of the other efforts mentioned. The company reduces stress on the utility by paying homeowners to lower their electricity usage during times of peak demand. They also sell smart plugs that can turn off power to homeowners’ appliances remotely to facilitate this process. OhmConnect has been an early proof of the success of VPPs in managing demand. However, the number of EVs and residential solar electricity production have increased by an order of magnitude since the company was founded (Fig. 4). The company’s tools address neither these components nor battery storage, which account for up to 80% of a modern household’s needs (Fig. 2). In the future, as more homes add solar, battery, and EV charging OhmConnect must either address these elements, or it will not be nearly as effective as competitors.

Subscribe now

Software scales faster, but hardware gives strength

Facebook engineers commit new code continuously, which is released to Facebook’s 3B users within a few hours. Apple has sold 2.3B iPhones since their introduction in 2007, and each product cycle lasts two years. Venture funding poured into mobile software applications in the years following the launch of the iPhone. Many thought that it was easier to build high-margin software businesses than to develop cutting-edge hardware. However, access to users through a hardware platform has gained importance in recent years.

Apple’s changes to the iPhone’s privacy settings lowered Facebook’s advertising revenues by $10B last year. The company potentially made these changes to boost its own advertising revenues. Even companies primarily known for their software offerings have released hardware products that reduce their dependence on outside platforms (e.g. Microsoft Surface or Google Pixel). Between 2019 and 2022 Meta invested $36B into its VR headsets partly to create a new hardware route to its users. However, platform access is not an issue for hardware applications like headphones. Apple would never make it more difficult to connect Bose headphones to your iPhone just to promote its own AirPods.

In a residential electrification setting the impact of hardware can be even greater. Homeowners expect to keep their smart electrical panels or solar panels for decades. They also hold on to their cars on average for seven years. A stand-alone software offering may scale faster. However, a software product that is attached to a piece of hardware can lock in users for years as the switching costs will be significant. That is why I find Lunar Energy’s combined hardware and software product strategy so exciting. It is also why I think OhmConnect’s software-first strategy leaves it without a moat against SPAN.

A SPAN panel already can monitor and control every circuit in the house. Homeowners can get much more leverage out of a VPP software built by SPAN, than one built by OhmConnect that only affects the energy usage of appliances. If there was a VPP operating system for a house, the smart panel would be the logical host for it. The panel has an overview of all of the functions in the household. Would you rather fully charge your EV, change the thermostat, run your dishwasher, or sell electricity to the grid? The smart panel is the central place to store all of these preferences and tradeoffs.

What will the future look like?

The cellular network analogy breaks down at some point. For example, a single iPhone can host hundreds of applications, some of which represent trillions of dollars in market cap. Instead, there are a handful of applications in each home, and the market opportunity is also smaller. The truth is that we are in the proliferation phase of the home VPP and resiliency markets. There are dozens of companies with various hypotheses and product strategies working to solve roughly the same problem. In each market, after the proliferation phase comes the consolidation phase. The outlines of the user’s preferred product will gradually emerge. The companies with more resources will capture the market and acquire smaller competitors. The less successful companies will exit.

The successful companies will have a unique strength that can be sustained (a moat). SPAN’s unique strength is that it is at the center of the action, and this strength can be sustained because of high switching costs. Lunar Energy’s unique strength is that it combines the high switching costs of hardware with the fast scaling of software. Car companies like GM sit on top of the biggest storage component at the home, which is the EV battery. However, most people expect their car to work with their house, not the other way around. The VPP center of the home must remain in the house, so I doubt if GM will be able to capture much of the value here.

As I have mentioned before, OhmConnect is currently in a weak position compared to its competitors. It must strengthen its position in the coming years to remain competitive in this market. As for Weavegrid, if I were to summarize my understanding of Apoorv’s pitch, it would be: “It’s the transmission, stupid!” Many things can happen at individual homes to manage demand. However, the utility must still coordinate supply and demand at the network level. Network-level coordination is where companies like Weavegrid will be needed, and this function cannot be taken over by any products solely focused on the home.

A song I like

Did you get enough love, my little dove?
Why do you cry?
And I'm sorry I left, but it was for the best
Though it never felt right
My little Versailles

A scene I like

Senator Pat Geary: “I despise your masquerade, the dishonest way you pose yourself. Yourself and your whole fucking family.”
Michael Corleone: “Senator, we're both part of the same hypocrisy, but never think it applies to my family.”

Some business ideas only make sense at scale. Businesses that rely on delivery networks are one example. Amazon has rarely paid a dividend to shareholders since it went public in 1997, because it has continuously reinvested its profits into expanding its distribution network. Two-sided marketplaces are another example. Uber spent more than $23B (!!!) in 2018 and 2019 alone to subsidize its riders. It finally reached profitability this year, 14 years after its founding. The TV show Silicon Valley even dedicated an episode to this concept (YouTube link).

Tesla famously took a different approach to the high marginal cost of low volumes. The company initially aimed at the luxury market, which has higher profit margins. It built a strong brand, where buyers could feel that their purchase of a $100,000 Roadster gave them something more than just a vehicle. Subsidizing products with the help of investor money or aiming for the luxury market are two ways in which startups have tried to address the issue of higher upfront costs. As far as I can tell, charging customers a Green Premium is a new method, that is increasingly championed by many energy and climate thinkers.

Subscribe now

Customers pay for scaling with the Green Premium

According to the Breakthrough Energy page dedicated to the subject, “The Green Premium is the additional cost of choosing a clean technology over one that emits more greenhouse gases.” The website explains that,

The average retail price of ground beef is $3.79 per pound, while a plant-based burger is $5.76 per pound. The Green Premium for a zero-carbon burger is the difference in cost between the two, or $1.97. But the regular burger doesn’t reflect the true cost of emissions generated in its production.

Any salesperson can tell you that people buy a new product over an incumbent if it’s i) better, ii) cheaper, or iii) lowers risk, and lowering risk has the same effect as lowering cost. Products that charge a Green Premium are by definition not cheaper. They also generally do not directly reduce risk for the buyer. Climate change is a significant risk, but that risk is borne by everyone, as emissions are an economic externality. The buyer cannot draw a direct line between their purchasing decision and their reduction of risk, the way that they can for buying a home security system. The only remaining solution is to convince the buyer that this greener product is better.

What constitutes a “better” product? Other than price, which was discussed earlier, products can be better on features or specifications. For example, you may decide to purchase a new cellphone because it has an ultra-wide angle lens, which was a missing feature on your previous phone. Alternatively, you may purchase a new phone because it has a faster processor than your previous one, which was poorer in this specification. This kind of product differentiation is key to building new markets or capturing underserved market segments.

“Lowering emissions” is a product feature

I have written before that startups with high business risk (quadrants 2 and 3) are creating new markets or product categories, whereas startups with low business risk (quadrant 4) are taking over existing markets (Fig. 1). Business risk is lower for quadrant 4 companies because there is never a question of what features to build. The product features in the market are already known, and the new entrant merely has to beat the incumbents on specifications or price. For example, the quadrant 4 company Sila Nano builds Titan Silicon™, which mimics graphite in battery anodes in every way, with the difference that it increases energy density by 20%. Sila’s product differentiation is on specifications, and it may even charge a higher price than the incumbents. Electric Hydrogen (EH2) is a quadrant 4 company helping to produce hydrogen from electrocatalysis instead of fossil fuels. Hydrogen is a commodity and an element on the periodic table, so there is no way to beat the incumbents on specifications. As I have written before, EH2’s main innovations are around ways to meet or beat the incumbents on price.

Lowering emissions is a product feature, and startups in quadrants 2 and 3 may be able to charge a Green Premium for this feature. Startups in quadrant 4 cannot differentiate their products based on features. Therefore, they cannot charge customers a Green Premium at scale. I have emphasized “customers” because these companies may be able to charge other entities such as the government a Green Premium in the form of incentives. I have emphasized “scale” because it may be possible to charge more for pilots or corporate initiatives, but not the level of output that impacts the day-to-day operations of most customers.

Branding commodities is ineffective

A few of the most famous climate tech categories (steel, cement, electrofuels, hydrogen, etc.) fall in quadrant 4, as large markets for these products already exist. Some startups active in these categories, which are mostly commodities, have created trademarked brands for their products to differentiate them as less carbon-intensive. They may be planning to use this branding as a way to charge customers a Green Premium. Firms that produce commodities are generally price-takers, which means that they “must accept prevailing prices in a market.”

Other commodity producers have tried branding their products as well, without much success. Chevron with Techron® and Shell V-Power® are gasoline brands based on the name of their proprietary engine-cleaning detergent fuel additives (Fig. 2). The content of gasoline, including the addition of detergents, is tightly regulated and controlled nationally. That fact may be why consumers are most likely to choose gas stations based on price (>70% for some surveys), and they have a much weaker association with gasoline brands (<35% for most surveys). This exercise reminds me of an episode of Mad Men, where Don Draper encourages the executives of the tobacco company Lucky Strikes to brand their cigarettes with “It’s Toasted.” One of the executives protests, “But everybody else’s tobacco is toasted.” Don clarifies, “No. Everybody else’s tobacco is poisonous. Lucky Strikes is toasted.”

Electric Hydrogen shows the path forward

The seasoned founder and CEO of a battery technology startup once told me, “The software business is all tactical. This business is all strategy. We started this company more than a decade ago. Right now I am not thinking about next year. I am thinking about five years from now.” In Good Strategy/Bad Strategy, Richard Rumelt notes that “The most basic idea of strategy is the application of strength against weakness. Or, if you prefer, strength applied to the most promising opportunity.”

Cleantech commodity startups have multiple strengths. As quadrant 4 companies, they are aiming for enormous existing markets like steel ($550B), cement ($420B), and hydrogen ($150B). The secular shift in the cost of electricity is another strength for these companies. Renewables like solar photovoltaics have reduced in price by an order of magnitude over the past decade, and they are now less expensive than fossil fuel alternatives (Fig. 3). These companies should be able to leverage such strengths and use investor funding to scale their operations, just as Uber did.

One of the reasons that I like Electric Hydrogen is that they have been applying these strengths to building an economical product. Raffi Garabedian, their co-founder and CEO, told the MCJ podcast,

Looking back at hydrogen over the years… we could never make ourselves convinced that hydrogen could be produced economically relative to the alternatives. And the alternatives are fossil resources. So, now fast forward to the end of 2020… and the situation had changed… Renewable electricity generation, both solar and wind, are now… cheaper than any other form of generation.

That's a key input to the production of renewable hydrogen. The missing link is … [a system that can perform] hydrogen conversion in a really cost-effective manner… Dave [our CTO] had a hypothesis… around how electrolysis can be driven in the same kind of way that we drove very successfully over the decades solar technology… By improving conversion efficiency you drive down the cost.

Electric Hydrogen is using the secular shift in electricity prices and the founders’ expertise in building efficient conversion systems to create a new and economically relevant source of hydrogen. Please note that capturing this market with such an approach only made sense at the end of 2020, when the price calculations finally made sense. It is bad to be too late, but also it is bad to be too early. In the same MCJ interview, Raffi makes it clear that his company makes economic sense regardless of any policy provisions.

It's all about economics. For this to be successful, it's got to be cheap. And what does cheap mean? It's got to be on par with the fossil resource. Now, we can sit around and wax on philosophically about, ‘Gee, wouldn't it be great if there were a carbon tax? Gee, wouldn't it be great if we could monetize… the societal costs of CO₂ emissions?’ … We've been waiting for a long time. We finally have the technology, we believe at Electric Hydrogen, and the resource to achieve cost parity in the next few years with the fossil resource, without actually having to have all those other great policy things that could accelerate adoption even further.

Electric Hydrogen shows the path for building a cleantech commodity startup. The company does not expect its customers to fund its scaling efforts by paying a Green Premium. Instead, it is using less expensive inputs and a more efficient process to build a commodity that can compete with the fossil fuel alternative on price. The fundamental economics of the company make sense. That is why even in the absence of any policy initiatives it can use investor financing to scale.

Conclusion

The Green Premium is a way to charge customers for scaling costs, which have historically been borne by investors. Companies with high business risk may be able to charge a Green Premium for the product feature of “lowering emissions.” Companies with low business risk can only differentiate themselves on price or specifications, and cannot charge a premium for such a feature. Furthermore, branding commodities in this quadrant is an ineffective product differentiator. Companies with low business risk have the advantage of pursuing known and sizable markets. They should be able to scale using investor funding as long as they can show that secular trends and their technologies allow them to compete on price. Companies with low business risk may not have access to the Green Premium, but if they are set up properly they should not need it either.

A song I like

A scene I like

Cool Hand Luke: “Sometimes nothin’ can be a real cool hand.”

1. How can manufacturers produce more EVs while scaling mining more slowly?

2. What are the technology and business risks of Magrathea?

3. Why has the founder of an LFP cathode startup invested in Magrathea?

Background

According to the Energy Information Administration (EIA), the number of light-duty electric vehicles (EVs) globally will increase from 9 million in 2020 to nearly 700 million in 2050 (Fig. 1). Building hundreds of millions of new EVs requires tens of millions of tons of battery materials (lithium, cobalt, nickel, etc.). New mining ventures often require more than a decade to produce minerals at scale. Observers expect a shortage of these minerals in the coming years. Battery manufacturers like CATL and even automakers like Tesla have used vertical integration to ensure a steady supply of these minerals. At the same time, startups like Mitra Chem (founded in 2021 and raised $87M) have focused on other battery chemistries, such as iron-based cathodes instead of cobalt-based ones, partly to circumvent this mineral shortage.

These efforts are helpful and necessary, but they have all focused on optimizing the “electric” part of EVs. What about optimizing the “vehicle” part? American cars are heavy and getting heavier. Magrathea (founded in 2022 and raised $12M) addresses this challenge by producing magnesium, which can be used to create lighter structural aluminum and steel alloys. The company calls this approach “dematerialization.” Alex Grant, the founder and CEO of Magrathea, noted that dematerialization “is technology innovation focused on ‘doing more with less’ by, for instance, achieving higher economic output per mass of natural resources consumed.” He added that using this metal can “eliminate up to 75% of dead weight from some metal parts used in vehicles.” I like Magrathea because I like i) this philosophy, ii) their low business risk, and iii) their fun attitude!

In a complex system, try removal before addition

Nassim Taleb (Black Swan, Antifragile, Skin in the Game, etc.) is one of my favorite non-fiction writers and thinkers. His work often deals with the complexity of decision-making within large and disorderly systems. In a chapter called “Naïve Intervention” in Antifragile, he discusses how doctors can avoid Iatrogenesis, which are the negative side effects of medical interventions. In a lecture at Stanford as part of the book tour, he said,

Sometimes curing people is much easier to do by subtraction… All these rules are positive. They say, ‘Do this, do this, do this.’ Charlatans have typically nothing but positive rules. People who are scientists or philosophical have negative rules: ‘Don’t do this. Don’t do that.’ It’s much more rigorous actually, and they don’t have side effects… For example, remove cigarettes from society… It’s more effective than anything we are spending on drugs, and cumulatively more than anything done since penicillin… [or] if you remove corn sugar. … The first thing they do is try to give you a drug [for diabetes], instead of removing [sugar]… But pharma doesn’t make any money removing things from your system… This is the problem.

Interventions are often necessary to solve problems. However, interventions that consist of adding new elements to a complex system can have significant unknown side effects. A more robust solution would be to remove problematic elements from the system. This “negative” intervention would be much less likely to cause harmful side effects.

The global supply chains for structural metals and battery materials are also highly complex. Manufacturers’ response to the global demand has been to mine and process ever more iron and aluminum for structural metals, and cobalt, nickel, and lithium for EV batteries. This voracious appetite has led to environmental damage and human rights abuses in different countries. Magnesium-based alloys require less steel and aluminum for the same strength, which makes them lighter. A lighter vehicle can also operate on a smaller battery for the same driving range. In addition, Magrathea extracts magnesium from ocean water, which is much less invasive than mining hard rock. Solving mobility challenges by reducing the amount of structural and battery metals used in EVs is like treating a diabetic by removing sugar from their diet (Fig. 2). I think this philosophy is a truly innovative approach, and it is one of the main reasons I like Magrathea.

Magrathea has a low business risk

I have previously written about the technology and business risks faced by startups in the energy transition. Much like the other materials companies that I analyzed in that piece, Magrathea is a Quadrant 4 company facing high technology and low business risks (Fig. 3). I like this risk profile because of my experiences as an entrepreneur and venture capitalist. I generally dislike business risk, because it means that the company can execute perfectly over multiple years to build a great solution, but then realize that there is no market for that solution. Technology risk can be unpredictable and time-consuming, but startups pursuing very large markets can still have successful exits even if their timelines inevitably slip by multiple years.

According to Business Wire, the global “magnesium market was valued at $4.37B in 2021, and the market size is estimated to reach $6.36B by 2028.” This market size is respectable and roughly equivalent to the global markets for lithium ($5B) and cobalt ($8B). However, Magrathea is making a much bigger implicit bet. It is betting that the shortage of raw battery materials will make magnesium-based structural metals a much more common component of mobility in the future and that this market will be significantly bigger down the road. I think that is a reasonable bet.

Vivas Kumar, the CEO and co-founder of Mitra Chem, which is building lithium iron phosphate (LFP) batteries, has personally invested in Magrathea. One potential reason for this decision is that LFP cathodes offer a lower energy density and a shorter driving range compared to cobalt-based (NMC and NCA) cathodes. This fact has historically kept LFP cathodes out of the most high-end and high-performance applications. A significantly lighter vehicle means automakers can use LFP batteries across their fleet without sacrificing driving range, expanding Mitra Chem’s market.

Alex Grant previously co-founded the direct lithium extraction company Lilac Solutions, where he gained experience with brine mining. In a white paper published in Light Metal Age, he noted that producing magnesium from brine, “involves four main steps: hydrometallurgy, dehydration, electrolysis, and metal foundry.” Magrathea’s main innovations are likely in the dehydration and electrolysis steps, as the other steps are not differentiable. The information online is limited, so I would be curious to find out what is their actual technology innovation. Does this innovation lend itself to a continuous improvement that builds a technology moat? Instead, do they plan to build a moat with highly efficient operations and economies of scale?

From what I gather, it seems like the company used their $2M angel round to build a proof of concept. It then raised a $10M seed round led by Capricorn. Capricorn is a top-tier deep tech VC that likes to see working prototypes before investing. Alex Grant’s previous experience, the existence of a working prototype, and the involvement of Capricorn are all positive signs regarding Magrathea’s technology.

Magrathea looks like fun!

Magrathea’s website looks completely different from other deep tech startups. Other websites tell the visitor “We are nerds building something difficult” or “We are altruists saving the planet.” I think Magrathea’s website tells the visitor, “Pssst! Let me tell you a strange secret. Isn’t this cool? Weee!!!” The yellow, pink, and purple hues of the website remind me of the thematic colors of the video game Grand Theft Auto: Vice City (Fig. 4). The name Magrathea itself is a reference to The Hitchhiker’s Guide to the Galaxy, a book I mostly heard about as an undergraduate engineering student. At least from the outside, the company has an air of youthful enthusiasm. I feel that many people who think about the climate only have a mindset of concern. They do not always acknowledge that this work can be fun! I am glad that Magrathea is injecting a more enthusiastic energy into this effort.

A song I like

Now here you go again, you say you want your freedom
Well, who am I to keep you down?
It's only right that you should play the way you feel it
But listen carefully to the sound of your loneliness

A scene I like

Gunnery Sergeant Hartman: “Private Joker is silly and he is ignorant, but he's got guts. And guts is enough.”

1. What is the biggest threat to SPAN’s business model?

2. Why is SPAN’s market a natural extension of Tesla’s product strategy?

3. How should SPAN react to this threat from Tesla?

Tesla is the biggest threat to SPAN’s business model

SPAN builds smart electrical panels that help homeowners monitor the production, storage, and consumption of electricity at their homes. They have raised $230M from top-tier climate VCs, like Congruent and Capricorn, and they are by far the most visible brand in this category.

In a previous post, I argued that SPAN does not have a “technology moat,” as other companies can quickly copy its technology and build a lower-cost alternative. I observed that it instead has a powerful “business moat,” through its luxury brand, great design, and ease of operation. I noted that Apple has pursued a similar business moat across its different product lines, and it has succeeded in that effort.

Tesla is another example of a company with a luxury brand and great design. SPAN is highly susceptible to its market being taken over by Tesla. SPAN’s business moat is ineffective against this potential competitor (Fig. 1). In addition, SPAN has now proven the existence of a significant market for a smart electrical panel, which is adjacent to several of Tesla’s core product lines. Tesla would find this complementary market a natural area to gain additional revenue and a way to further strengthen the position of its core business.

SPAN complements Tesla’s hardware and software products

In his interview with the MCJ podcast, Arch Rao, the founder and CEO of SPAN, observed that,

Where we are seeing the most amount of traction and demand today is retrofitting into existing homes like yours and mine that have been around for a few decades or are getting solar and battery and EV charging or some combination thereof.

The purchase of a SPAN electrical panel is triggered when homeowners integrate solar, battery, or EV charging into their homes, which are all products made by Tesla (Fig. 2). SPAN is building a complementary product to Tesla’s core product offerings, and customers make this purchasing decision simultaneously. Building a smart electrical panel would be a natural extension of Tesla’s product portfolio.

Such a move for Tesla is not just offensive, but also defensive. In the same podcast, when describing the origin story of SPAN, Arch recalled his time working at Tesla,

And as you think about the rate at which we were beginning to deploy solar and EV charging in batteries, it became more and more common that the electrical panel either had to be upgraded or replaced… You kind of had to be the person either responsible for or very close to the deployment of thousands of these systems to then see that the electrical panel was a bottleneck.

Every successful company must control as much of its path to user adoption as possible. Tesla deployed a national network of Superchargers when it realized that the availability of charging stations would play a major role in the adoption of EVs. If the electrical panel is a major bottleneck to the adoption of Tesla’s other products at home, then the company must do everything in its power to remove this bottleneck without relying on a third party.

The real market is much bigger than the hardware. The real market is about the data and the granular control of every circuit in the house. In the same MCJ interview, Arch noted that,

Within our panel, we've integrated the ability for us to monitor the grid and determine when it's safe to disconnect and when it's safe for the join. We have the ability to measure every circuit… We have the ability to control every circuit and also control some subset of major appliances in your home, and it's got… a computer built into it that is purpose-built to optimize the power flow in your home.

There is an enormous market for this data and the ability to control circuit-level electricity flow, both on the consumer side and the utility side. On the consumer side, products like the Google Nest Learning Thermostat monitor the performance of the heating furnace and offer to send a handyman if the furnace is malfunctioning. This service is a source of revenue. On the utility side, companies like OhmConnect monitor the electricity usage of homeowners and pay them to reduce their usage during peak hours to reduce demand on the utility. For a utility, this demand management has the same effect as adding more generation capacity.

The electric panel produces highly granular data about electricity generation, storage, and usage in the home, and it has the means to control those features. Tesla already offers software products for this purpose in the commercial sector. Tesla’s “Autobidder provides independent power producers, utilities and capital partners the ability to autonomously monetize battery assets.” In addition, Tesla’s “Powerhub provides a single interface for managing many combinations of energy assets including solar, storage and select non-Tesla assets.” Tesla can provide the same services on the residential side as an extension of its commercial strategy.

Even though SPAN is a very well-funded startup, it can never win a head-to-head battle against Tesla. Tesla is not only one of the largest companies in the world ($730B market cap) but also it is a highly dynamic and innovative company. Old behemoths like Intel or Cisco that used to be great can be disrupted, but not Tesla. Tesla’s hardware expertise means that it can produce an excellent electrical panel, and use its balance sheet to undercut SPAN on price until the young startup runs out of cash. Knowing this, SPAN should either build a bigger moat or try to get acquired by Tesla.

SPAN should build a bigger moat or get acquired

SPAN does not have the resources of Tesla, but it has the benefit of having deployed units that are generating data. SPAN can potentially build a powerful moat with this data. Google, Meta, and Amazon often capture up to three-quarters of all digital advertising dollars. Their granular user data allows them to deploy the most targeted ads to their users. Competitors that lack this data have fallen behind for years. SPAN should be creative about the services that it provides using this data. Monitoring the electricity usage of appliances can not only lead to predictive maintenance services or virtual power plants but also potentially advertising!!! 🙂 For example, if the electricity usage of a washing machine becomes irregular, the homeowner may receive advertising for a new one. These kinds of high-margin services increase the profitability of the product, and they give consumers and partners additional reasons to work with SPAN instead of Tesla.

The other potential solution is to get acquired by Tesla. Gaining a great new product, brand, and team can also benefit that company. In the past, Tesla has acquired companies that either strengthen its technological edge (e.g. Maxwell Technologies) or complement its product portfolio (e.g. SolarCity). For a venture-backed startup, a good rule of thumb for a successful exit is one valued at 10X the total funding raised. At its current market cap, Tesla can easily acquire SPAN for $2.3B. In 2016, when Tesla was valued at ~$30B, it acquired SolarCity for $2.6B. For SPAN, the door to an acquisition will not remain open forever. Most companies cannot handle a multi-billion-dollar acquisition, and Tesla’s appetite will wane if the acquisition price increases significantly.

Conclusion

SPAN’s main moat, around the function and design of its luxury product, is ineffective against Tesla, which has a similar product and branding strategy. SPAN’s smart electrical panel naturally complements Tesla’s hardware and software product offerings. In addition, helping to increase the adoption of such a product will remove one of the main bottlenecks in the adoption of Tesla’s solar, battery, and EV product lines. SPAN’s only options are to build a bigger moat or get acquired. SPAN has a head start in generating data from its panels. It can use this data to build a data moat and add more high-margin services. Alternatively, SPAN can be a great acquisition target for Tesla, where it can fetch a price that would constitute a successful exit.

A song I like

“So I called up the Captain,
‘Please bring me my wine’
He said, ‘We haven't had that spirit here
Since 1969’”

A scene I like

Tony Soprano: “Well you know what they say, ‘Be careful what you wish for, you might just get it.’”

1. How do venture capitalists (VCs) analyze the risks associated with technology startup investments?

2. How do startups prevent competitors from taking over their markets?

3. What do startups with low business and high technology risks have in common in the energy transition?

Background

Entrepreneurs or VCs often consider two forms of risk: technology risk and business risk. Technology risk is roughly, “Can I build it?”, and business risk is roughly, “If I build it, will people buy it?” Technology startups in the energy transition can also be placed on these axes (Fig. 1).

To my knowledge, it is not possible to build a technology startup that faces neither of those risks. Otherwise, there are many examples of successful companies that have faced any combination of those risks. The investment theses of VCs sometimes bias them towards certain types of risk. For example, I have heard some investors say that they invest in companies with low technology risk, and others claim that they prefer low business risk. Those preferences sometimes relate to the expertise, network, or other resources that the VCs have for derisking the investment. Entrepreneurs should be aware of those theses and preferences, and focus on fundraising from firms that align with their businesses.

Employees may also have preferences for joining companies that face particular types of risk. They may have this preference for two reasons. First, the type of risk faced by a company relates to the types of work most valued by the company. For example, a researcher would likely find the work available in quadrants 3 and 4 companies to be most rewarding. On the other hand, a product manager would likely find the work in quadrant 2 companies to be the most engaging. Second, employees are some of the biggest shareholders (and by extension investors) of venture-backed startups. Just as VCs may have a preference for the type of investment risk that they take on, so should the potential employees.

Quadrant 1: Low Business and Technology Risks

As far as I know, it is not possible to build a venture-scale startup in this quadrant. It often takes 7-10 years to build a successful software startup, and it can easily take twice as long to build a “deep tech” company. Entrepreneurs and investors take enormous risks to build and fund such companies. However, if they are successful they can build a monopoly in a multibillion-dollar market. This new business forms a monopoly by building a “moat” around its market, to keep away competitors.

This moat is based on factors relating to the startup’s business, technology, or both. For example, for the past two decades, Google has had a monopoly in the highly lucrative online search business, where it holds a 90% market share. Google’s technology moat is based on the algorithms underlying its search engine, which provide some of the most relevant results for each query. Google’s business moat is its brand name recognition. We even use the phrase “Google” as a verb to mean online search. It is very difficult for a new startup to build effective algorithms and a brand name to gain market share from Google. The greatest threat to date has come from Microsoft Bing, which has a radically new technology (ChatGPT) and a strong brand name of its own. Peter Thiel’s Zero to One provides the best explanation of this principle that I have read.

It is not possible to build a venture-scale startup in the first quadrant because it is not possible to build a moat around the new company. The type of risk that a startup takes and the type of moat that it builds are nearly synonymous. If a startup does not take the necessary risks to build a moat, then any incumbent or newcomer can easily encroach on its territory by cutting prices. As the proverb goes, “Nothing ventured, nothing gained.”

Quadrant 2: High Business and Low Technology Risks

The technologies in this quadrant are not “easy.” However, they are difficult in relatively predictable ways. That predictability makes the endeavor low-risk. The technology work in this quadrant is primarily in the form of engineering, and building features that have been already built before in a different context. However, the business risk in this quadrant can be unpredictable. There is generally no evidence that customers would be willing to adopt such a product in large numbers, and more importantly, that they would be willing to pay for it. The right combination of these factors is called product/market fit (or even product/market/sales fit).

Twitter

I have included Twitter as an outlier among companies involved in the energy transition. I did so to point out how network effects are the dominant form of a business moat that the most famous Silicon Valley startups have built over the past 15 to 20 years. These companies include social media (Twitter, Facebook, TikTok, Snapchat, etc.) as well as two-sided marketplaces (Uber, Airbnb, DoorDash, Etsy, etc.). Being a member of one of these services is valuable because everyone else we know is on these services. That fact makes competing services that have fewer members less valuable. Instagram launched Threads in July 2023 to compete directly with Twitter. They started working on this project in January, and launched the app six months later, which is further evidence that Twitter did not have a technology moat. At the same time, Instagram leveraged its own 2.5 billion users to overcome Twitter’s network effects. Threads gained 100 million users within five days.

Terabase Energy

Terabase offers an integrated suite of software tools to developers and operators of solar power plants to reduce their costs. Building these tools is not a high-risk proposition. However, knowing which tools to build to gain rapid industry adoption, and then getting integrated into the plant’s Supervisory Control and Data Acquisition (SCADA) from the beginning is an effective moat against competitors. Their customers will find it difficult to change all of the systems where Terabase’s product has been integrated, and train their workforces on a different platform. Using multiple system integrations and significant user adoption as a business moat is similar to the strategy of Microsoft Office for maintaining market share.

Dandelion Energy

Dandelion helps homeowners replace their heating and cooling units with geothermal heat pumps. Their primary technology innovation is in better calculating the length of the ground loop required for their units (Fig. 2). Much like Amazon, their moat is mainly operational. They own the equipment used for drilling holes for the ground loop, and they employ the workers who install these systems. Finally, they work with banks to provide financing to their customers, and their size and track record allows their customers to benefit from very favorable rates. Any new entrant into this market would find it difficult to compete with Dandelion’s economies of scale or provide similar financing terms. Originally, Dandelion was built as a two-sided marketplace to connect homeowners with independent contractors who could install geothermal units for them. This model had some success; however, they realized that they could improve user adoption if they employed the workers and provided the full installation service. This kind of iterative product discovery is common for companies in quadrant 2, which can relatively easily change their product offerings to better meet customer needs.

SPAN

SPAN provides smart electrical panels that help with the monitoring and management of electricity production, storage, and usage at home. The technologies that it has developed are useful and easy to understand, but they do not represent a technology moat. SPAN has built integrations with battery systems and relationships with distributors, neither of which provide a significant business moat. I believe that SPAN’s main business moat is its seamless operation, design, and above all brand. SPAN has many of those features in common with Apple (Fig. 3). Apple has built a reputation as a luxury brand with beautiful and useful products that function seamlessly. The iPhone is not competing with Samsung on megapixels or processor speed, but on providing a delightful user experience. While doing so, it is capturing half of smartphone sales revenues globally, and a whopping 85% of profits! Those profit margins are only possible because of its position as a luxury brand. A few years ago, it even hired the head of the luxury clothing brand Burberry to help revamp its physical and online stores.

SPAN is following a similar strategy in building a branding moat. The interview that Arch Rao, the founder and CEO of SPAN, gave the Watt It Takes podcast provides further evidence. The company built an early version of its product within a year using its $3.5M seed financing round. Companies with high technology risk generally require multiple years, and one to two orders of magnitude more funding to reach that stage. At the same time, SPAN spent a sizable portion of that funding on hiring an industrial design firm. As Arch explained in that episode:

It's very unusual for a seed-stage company, with a few single digit millions in the bank, to go spend money with an external industrial design company. But from the very beginning it was super important to us to build a… hardware and software product that was stunning, and I think we’ve achieved that.

SPAN’s look, feel, and brand are a significant part of its business moat. As Apple has shown, such a moat can be quite effective.

Quadrant 3: High Business and Technology Risks

Startups in this quadrant are building technologies that require a significant amount of scientific invention or are very difficult to scale. Furthermore, these companies are generally creating new products or markets that did not exist before. Therefore, there is no evidence that customers are willing to buy such a product, how much they are willing to pay for it, and what features the product must have to make it valuable. These same business risks exist in quadrant 2 as well. However, in quadrant 3 entrepreneurs commit to building a technology for years before knowing whether there will be any customers. At the same time, successful quadrant 3 startups can create enormous new markets with significant technological and business barriers to entry for competitors.

Tesla

Tesla is a famous quadrant 3 company. There was no electric vehicle industry before Tesla was founded in 2003. There were no direct sales figures that the founders could use to determine the market size or pricing for each vehicle. The adjacent internal combustion engine (ICE) vehicles industry provided clues in terms of price, expected range, features, or potential market size. However, all of these components were merely educated guesses before the company built the vehicles and consumers started buying them. There was also an enormous technological and scaling challenge involved in chasing this uncertain market. Five years after its founding, the company launched the Roadster in 2008 and sold 100 units that year. A decade after its founding, Tesla launched the Model S in 2012 and sold 3,000 vehicles in that year (Fig. 4). In 2023, two decades after its founding, Tesla is expected to deliver up to 2 million vehicles, and it is valued at nearly one trillion dollars! Tesla’s founders took on significant technology and business risks, and their success created a lucrative new market, where their company is one of the leading figures.

Form Energy

Form provides long-duration (100 hours) grid-scale storage of electricity. In what analysts have called a “duck curve,” the load on California’s power generation drops markedly in daytime hours while solar energy is plentiful (Fig. 5). This load increases suddenly in the evening as the sun sets and demand increases. As the grid adds more intermittent renewables like solar and wind, long-duration storage will most likely be necessary to smooth the swings in supply and demand, and increase resiliency. Form is placed in this quadrant because of its relatively high technology and business risks. An electrochemical storage unit of this magnitude has never been built before. Form needs to prove that it can achieve this scale reliably and repeatably. On the business side, there is currently no significant market for long-duration grid-scale storage technologies. The adjacent market for “peaking plants” or customer conversations may help with product specifications or pricing. However, the only true evidence of market demand is when customers pay for the product. While these risks are significant, a successful launch and scaling effort will create an entirely new industry where Form will have technological and business barriers to entry against future competitors.

Rondo Energy and Antora Energy

Like Form, both Rondo and Antora are involved in storing intermittent electrical energy. While Form stores electrochemical energy, Antora and Rondo store thermal energy and help industrial companies use this energy as process heat. Antora stores the heat in graphite blocks, and Rondo does so in bricks. Similar to Tesla and Form, Rondo and Antora also face higher technology and business risks, but with an important difference. Internal combustion vehicles and peaking plants could provide a helpful template for Tesla and Form’s products and markets, respectively. Process heat has different requirements in various industries like food and beverage, cement, or steel making. Furthermore, in most cases, the heat is part of the plant, and not an external product or service. In addition, grid operators do not need to make significant changes to their systems to source or sink electrical energy from a Form battery. Whereas, factories need to make capital investments in their operations in order to source their process heat from a thermal battery. The lack of direct market references and the greater difficulty in integrating thermal batteries into ongoing operations increase the business and technology risks of Rondo and Antora compared to Form. However, as with other venture-backed companies, building or accessing large markets is only possible through taking risks.

Quadrant 4: Low Business and High Technology Risks

Much like the companies in quadrant 3, quadrant 4 companies also face an uncertain R&D and technology scaling path to profitability. However, companies in this quadrant know that if they are successful, there is an existing multi-billion (or multi-trillion!) dollar market awaiting them. It is interesting to note that all of the companies from Fig. 1 in this quadrant are materials companies. They are all aiming at existing markets, where the buyer, product specifications, price, and competition are well-known. The challenge for these companies is to produce this known product with higher performance (Sila), lower cost (Electric Hydrogen), or in greater abundance (KoBold and Lilac).

Sila Nanotechnologies

Sila is replacing the graphite used in lithium-ion battery anodes with silicon to increase the energy density of these devices by 20%. Sila surrounds the silicon particles with a nanostructured material to absorb the 300% (!!!) swelling of lithium-saturated silicon. Sila’s commercial product, Titan Silicon, is a one-to-one drop-in replacement for graphite. According to Business Wire, “the global graphite market was valued at $24B in 2022 and is expected to grow to $38B by 2028.” Founded in 2011, Sila has needed more than a decade to engineer a product at scale that is substantially better than the incumbent, which was a significant technology risk. However, the company has faced low business risk, as it was aiming for a large market with established product specifications and buyers.

Electric Hydrogen

Electric Hydrogen (EH2) is building a new generation of electrolyzers that can produce hydrogen from water electrolysis at prices comparable to Methane Steam Reforming (MSR). There is already an enormous market for hydrogen as a feedstock for the petrochemical and fertilizer industries, and this market can grow much larger if hydrogen takes on a bigger role in transportation. According to Business Wire, “the global hydrogen generation market value surged from $148B in the year 2021 and is expected to reach $279B by the year 2030.” While EH2 technically sells hydrogen-producing machines instead of hydrogen, they often repeat their metric of interest as $/kg of H₂. The product and the market for hydrogen are understood well. The main question for EH2 is whether it can produce hydrogen at a scale and cost that allows them to capture a significant share of this market.

Lilac Solutions and KoBold Metals

Lilac is mining for lithium in brine, and KoBold is mining for lithium, cobalt, nickel, and copper in hard rock. These metals are used as raw materials for electrification and for building lithium-ion batteries. The markets for lithium ($5B), cobalt ($8B), nickel ($36B), and copper ($291B) are already quite large. An additional $15 trillion of these metals are required by 2050 to meet world demand (Fig. 6). Once again, there is little risk that these companies will take on years of R&D effort to produce a product without any buyers. However, they each face a significant challenge in successfully scaling their technological and operational efforts to capture a major portion of this enormous market.

1. What other industries will be affected by the energy transition?

2. What are the common sources of innovation in the materials transition?

3. Why do the economics of the startups in this space make sense now?

Background

I have been working to better understand the technological and business landscape associated with the energy transition. In the energy transition, we are not only changing the forms of energy that we use (renewable electricity instead of fossil fuels), but also we are changing the materials associated with those forms of energy. Overcoming the intermittency of renewable electricity and finding carbon-neutral resources are major causes of this materials transition.

Electric vehicle (EV) batteries are an interesting example of this materials transition. In a world where hundreds of millions of cars run on gasoline, we must produce and process a significant amount of petroleum. Entire industries are dedicated to this effort. However, in a future world where most cars are powered electrically, we must build new industries to produce and process the metals used in EV batteries.

While studying this subject, I have noticed that those active in this materials transition often refer to the three variables of abundance, concentration, and cost to describe their work (Fig. 1). These variables are interrelated. Engineers always have to make tradeoffs between interrelated variables to accomplish their goals. New companies have arisen within the past few years to make novel and innovative tradeoffs between these variables. In this post, I want to discuss how Lilac Solutions, KoBold Metals, and Electric Hydrogen each address the tradeoffs between these variables for their respective businesses.

Lilac Solutions

Lithium (the third element on the periodic table) forms the lightest and smallest metal ion, which makes it ideal as the charge carrier in EV batteries. For this reason, lithium-ion batteries have the highest energy density amongst common battery formulations. A typical electric sedan or light truck may have 50 kg (110 lbs) of lithium in the battery pack. The current lithium supply is inadequate for producing tens of millions of new EVs in the coming decades.

Lithium is either extracted from hard rock or from underground brine solutions. The traditional method of extracting lithium from brine involves pumping lithium-rich brine solution into large overground pools (Fig. 2). The brine solution is then further concentrated through evaporation (a process that can last multiple years!). Lithium is actually an abundant metal. There is even lithium in ocean water. However, at a concentration of 0.17 mg/L, 300 million liters (120 Olympic swimming pools) of ocean water would be necessary to produce a single EV. Processing that much ocean water is too costly to be economical.

The highest concentrations of lithium brine are found in underground reservoirs beneath arid high-elevation dried lakebeds. This geological formation is most common in the border region between Bolivia, Chile, and Argentina. Lithium mining is economically relevant in this region because the brine solution has become highly concentrated through natural means. At a concentration of up to 1,400 mg/L, 36,000 liters (1/70^th of an Olympic swimming pool) of brine solution from this region produces enough lithium for a single EV.

We already see the tradeoff between the variables of abundance, concentration, and cost in this example. Ocean water is much more abundant than the geological features that give rise to lithium-rich brine solutions beneath dried lakebeds. However, sourcing lithium from ocean water is very costly because the concentration of lithium in that solution is so low.

Lilac Solutions addresses this tradeoff in a new way. Dave Snydacker, the founder and CEO of Lilac Solutions, told My Climate Journey (MCJ),

For people who have studied oil and gas, or mineral resources, there's this classic pyramid where at the top you have a small number of sites with a very high-grade resource, and as you go down the base of the pyramid, you get to lower concentrations of lithium, but a much larger resource base with more sites that can be developed. And so, outside of the existing production, which was three or four brine projects, there were dozens of sites with large amounts of lithium, but at a lower concentration, where conventional technology would not be applicable. And it's those sites that seemed like the frontier for lithium.

Lilac aims for a more abundant resource, which is brine solutions with a slightly lower concentration of lithium. They overcome the lower concentration with a process that concentrates the lithium faster than traditional methods. They use ceramic beads that absorb lithium from the brine. They subsequently wash these beads in hydrochloric acid to produce lithium chloride. They process this salt into lithium carbonate or lithium hydroxide, which are the raw materials needed for batteries. While the additional processing adds cost, it gives Lilac access to much greater reserves of lithium.

KoBold Metals

KoBold Metals is another company active in the materials transition associated with electrification. They aim to increase the supply of copper, nickel, cobalt, and (hard rock) lithium. The world needs $15 trillion (!!!) of these materials by 2050. These metals are important in conducting electricity and building energy-dense EV batteries.

In his interview with MCJ on the Stanford campus, Kurt House, the co-founder and CEO of KoBold Metals, noted that,

I would take a step back actually and think about what a[n] ore deposit actually is. So you could dig down underneath Stanford's campus right now and you'll find cobalt guaranteed… You'll find cobalt probably around 20 to 50 parts per million… [There’s] something like 10 to the 18^th kilograms of cobalt near Earth’s crust…

The problem is it's in this low, very low concentration background concentration, and so to try to extract it from that background concentration would be way too expensive… An economic ore deposit is when you find a chunk of earth that doesn't have 30 or 40 parts per million cobalt, but rather it has 20,000 parts per million cobalt. That's what you want.

Kurt House is referring to the same resource pyramid that Dave Snydacker mentioned. However, KoBold pursues a different strategy compared to Lilac. Whereas Lilac moves down the concentration pyramid to increase abundance, KoBold increases abundance by finding concentrated metal deposits deeper in the Earth’s crust (Fig. 3). They do so partly by modeling the geological features and soil content that led to previous discoveries, and by making more sophisticated decisions about where to drill test wells. This effort, which is enabled by advances in artificial intelligence, allows KoBold to reduce the cost of discovering rich ore deeper in the Earth’s crust. A similar tradeoff between abundance, concentration, and cost holds here as well.

Electric Hydrogen

Electric Hydrogen (EH2) is also involved in the energy and materials transition, albeit not in electrification. EH2 will start by addressing an existing market, which is in supplying industrial-scale hydrogen for the production of fertilizer and petrochemicals. However, it expects that in the future there will be a much larger market for hydrogen as a way to store chemical energy, and it intends to address that market as well.

Industrial-scale hydrogen is currently primarily produced through methane steam reforming (MSR). In this reaction, methane (CH₄) endothermically reacts with steam (H₂O) in the presence of a catalyst to produce hydrogen (H₂) and carbon dioxide (CO₂). The same variables of abundance, concentration, and cost are relevant here. Methane is abundant (in the form of natural gas) and it is a concentrated source of hydrogen (25% by mass). In addition, MSR is a relatively inexpensive process.

EH2 uses water (H₂O) as its source of hydrogen atoms, which is abundant. Water is also a relatively concentrated source of hydrogen (11% by mass). EH2 produces hydrogen using the water electrolysis reaction, where water is split into hydrogen (H₂) and oxygen (O₂) with the application of electrical energy to an electrocatalytic stack.

The main drawback of MSR is the production of CO₂, which is a global warming agent. The wider adoption of MSR over electrocatalytic water splitting relates to the cost required to access that hydrogen. Historically, most of the electricity in the US came from power plants that burned coal or natural gas. The hydrogen produced from electrocatalytic water splitting using this electricity was far too expensive compared to the hydrogen produced from MSR. However, EH2 is benefitting from one major trend and two innovations:

1. Solar and wind power are not only carbon-free, but also they are now less expensive than power produced with fossil fuels.

2. EH2 has increased the efficiency of its electrocatalytic stacks with device physics innovations.

3. EH2 is building much larger electrolysis plants (100 MW instead of 25 MW) to reduce costs using economies of scale.

Combining these three elements, EH2 is aiming to lower the cost of its electrolytic hydrogen to be on par with hydrogen produced using MSR. In his interview with MCJ, Raffi Garabedian, the co-founder and CEO of EH2 said that, “we're running as fast and as hard as we can towards the capability to produce at fossil-parity. That means under $1.50 a kilogram of hydrogen produced. We think that that's possible in the next few years.” In this case, the main difference between new and traditional methods is in the cost and not in the abundance or the concentration. However, we can still analyze this innovation using the same three variables, where producing hydrogen at “fossil-parity” cost is the main challenge for EH2.

Conclusion

The world is poised to experience a materials transition in parallel with the energy transition of the coming decades. In this materials transition, there is a tradeoff between the abundance and concentration of materials and the cost to produce or process them. Lilac Solutions, Kobold Metals, and Electric Hydrogen have each addressed these tradeoffs in a unique way (Fig. 4).

Lilac’s technology allows them to extract lithium from less concentrated but more abundant brine resources. KoBold’s technology allows them to find concentrated mineral ores deeper in the Earth’s crust, where they are more abundant. For both companies, production costs from these new resources will likely be higher than traditional resources. However, these metals are in such high demand that the new operations will still be economical. We see the same principle in the extraction of other resources. For example, on-shore petroleum from the Middle East costs $31/barrel to produce, whereas hydraulic fracturing petroleum from North America costs $44/barrel to produce. In August 2023, when each barrel of petroleum sells for $82, both operations are profitable.

EH2 is extracting hydrogen from water instead of methane. There is no significant difference in the abundance or concentration of methane or water. Historically, producing hydrogen through water-splitting was much costlier than MSR. However, EH2 is taking advantage of new industry and technology trends to reduce this cost difference to parity. Cost parity is what allows them to substitute electrocatalytic hydrogen for MSR hydrogen.