For 30 to 40 years, the U.S. grid load was effectively flat. Efficiency gains absorbed every new appliance, every new screen, every new server. That era is over.
It’s three forces bombarding the grid all at once:
AI-led load growth that combines with industrial electrification and heat pumps, EVs, etc.;
a fractured generation mix that ranges from a few dispatchable plants to a fractured mix of solar, wind, geothermal and peakers with a wide variety of dispatch curves; and
an unpredictable climate with longer storms, deeper droughts and multi-day cold snaps.
The grid doesn’t break in any single hour. It breaks across days. And the asset class built for hours, lithium-ion, was never going to be the answer.
Arvin Ganesan is the CEO of Fourth Power. He has sat in three different seats of the energy transition: senior roles at the US Environmental Protection Agency on the regulation side, Head of Global Energy and Environmental Policy at Apple on the corporate policy side, and now as an operator building one of the more physically unusual long-duration energy storage (LDES) systems in the world - a thermal battery that stores electricity as light at 2,400°C in blocks of carbon.
In this conversation, part of our “AI meets the Physical World” series, Arvin walks through why the grid is structurally different now, why lithium-ion and LDES are not competitors but complements, and why supply chain (not technology) increasingly decides which storage technology actually scales.
His perspective is grounded in both operator reality at Fourth Power and a regulatory-and-policy view of how the global grid is being rewired in real time.
The three structural forces breaking a grid designed for the 1990s
Why the “short vs long duration” debate is the wrong frame
The thermal battery that runs hotter than an incandescent filament
A theory of abundance: building energy storage from petroleum coke waste
Why supply chain has become a proxy for scalability
Hyperscalers vs utilities: two customer segments, two clocks
The first-of-a-kind capital gap that kills hardware startups
Why the next decade is a golden time for energy technologies that deliver
Load demand in the US was almost perfectly flat for three to four decades. Small rises, mostly offset by efficiency. That equilibrium is gone.
Arvin frames the new pressure on the grid as a three-part story:
“Load growth is happening. This is not only data centers. This is heat pumps, electric vehicles, electrification of industry.”
That is one. Two is the diversification of supply. The grid has moved from large, predictable, centrally dispatched plants to a far more distributed mix, and “distributed” here doesn’t only mean rooftop solar. It means geothermal, nuclear, wind, solar, and gas all on the same wires, each with its own dispatch curve. Three is the physical reality of climate change pressing on infrastructure that was never built for it: longer storms, deeper droughts, multi-day cold snaps.
The winter storm example is the one that lands hardest. When gas gets diverted between electricity generation and home heating during a multi-day freeze, the system runs out of optionality. That is the gap LDES is built to fill -peak reliability for one to three days, not hours.
Arvin’s most clarifying point is that the entire short-vs-long duration debate is a labeling problem the industry imposed on itself.
“If I was able to go back in time 20 years, I would get rid of all of these labels. Like much of mankind, this is all a spectrum.”
The better question is not how many hours a battery stores. It is what value the asset provides to the grid.
Lithium-ion is a buy-low, sell-high asset. It takes cheap electricity and dispatches it at expensive hours, usually over two to six hours. Frequency response, fast ancillary services, capturing renewable overabundance -lithium-ion does all of that better than almost anything else.
Fourth Power sits in a different slot. Longer hold times. Capacity over days, not hours. Other technologies (like Form Energy’s iron-air chemistry) sit further out still, at the hundred-plus-hour end. None of these compete head-to-head. They slot into different grid needs.
Investors see it differently. The better question is not “ which LDES technology will win?” but “what technology fits into what grid service and what do cost, safety, and scalability for that grid service look like?”
Fourth Power’s system is unusual enough that it’s worth describing physically.
Electricity comes in from the grid. It is converted to high-temperature heat by heating tin and radiating light as the heat transfer mechanism. That energy is stored in highly insulated blocks of carbon, which heat up to 2,400°C - the temperature of the tungsten filament in an old incandescent bulb. The blocks hold that temperature for days or weeks.
When the grid calls for power, a thermophotovoltaic panel is positioned eight inches from the heat source. It captures the most efficient part of the light wavelength and converts it back to electricity, within seconds.
The physics the company is exploiting is in the name. As temperature rises, light emission rises proportional to the fourth power of temperature. At 2,400°C, the light intensity is enormous and a specially tuned solar cell captures it efficiently.
The most strategically important sentence in the entire conversation is this one:
“We’re designed to operate from a theory of abundance. 98% of our system comes from carbon.”
That carbon is petroleum coke - a waste product of petroleum refining. Every country that has a petroleum refinery produces it. Fourth Power adds a binder, bakes it, and that becomes the bulk of the system. The insulation is raw petroleum coke.
Compare this to lithium-ion: 65% of the world’s lithium is processed in China.
This is not a marketing distinction. It is a scalability one.
“Supply chain is another word to talk about scalability.”
Arvin’s argument is that investors and customers are not buying one-off products. They are buying the credible path from current scale to large volumes. A product that performs as expected but has a bottlenecked or geopolitically exposed supply chain prices in that risk. A product with a domestically manufacturable supply chain in every major market prices in optionality. That is why Foreign Entities of Concern (FEOC) rules now show up in investor diligence on hardware companies, and why a 98%-carbon system has a different scaling profile than a lithium-dependent one.
Fourth Power’s pilot project is in Massachusetts, commissioning this year. Commercial demonstration follows, with commercial products available in 2028, sold in 25-megawatt modules.
When asked whether LDES go-to-market shifts from conservative utilities to capital-rich hyperscalers, Arvin’s answer is both - but the clocks are different.
Hyperscalers care about time to power. LDES is a way to bypass long interconnection queues, bring capacity to market faster, and disconnect from the grid during peak hours. The urgency is acute, the dollars are sized for it, and the desire for clean capacity is real.
Utilities are the longer game. A vertically integrated utility has an obligation to serve everyone, an obligation to keep rates low, an obligation to maintain reliability, and is regulated by the Public Utilities Commission. That makes engagement slower. It also makes the market enormous.
“If you look just at peaking plants in the United States, there’s 250 gigawatts of peakers. The average age is over 30 years old.”
That fleet is at the end of its natural life. Replacement plus growth makes utilities a structural long-term buyer for any credible storage technology.
The strategic point: LDES companies that build a business only around hyperscaler urgency are mispricing the second leg of their own market.
When asked where most hardware startups fail, Arvin is direct. The hardest leap is not Series A to Series B. It is first-of-a-kind.
“How do you avoid venture dollars for your first commercial product before the finance sector is going to issue debt for those projects?”
Venture capital remains interested in energy technology that addresses load growth. But there is no clean, one-size-fits-all financing path for the first commercial project. That gap (between fully derisked technology and bankable infrastructure ) is where most hardware companies stall.
His broader observation on what separates survivors from casualties is almost stubbornly simple:
“The best way to get to market is to have your product work to the specs that you say it’s going to work. Success tends to go towards companies that deliver on what they say they’re going to deliver.”
That sounds obvious. It is harder when a company has raised at sky-high expectations and is incentivized to scale to manufacturing before technology risk is sufficiently retired. Fourth Power’s discipline -much of the basic technical work came out of MIT labs before the company was started- is the deliberate counter-pattern.
The closing frame is the one worth holding on to.
“The average energy consumption of a human being now is very, very different than it was 20, 30 years ago. We’re still at the early days of AI. The market for technologies that can deliver reliable, clean, and perhaps most importantly, affordable energy becomes massive. This is the beginning of a golden time.”
Two layers in that sentence are worth separating. The first is that LDES is not an environmental story - it is a reliability and affordability story, and the addressable customer set is far broader than the climate-tech framing has historically allowed.
The second is the AI overlay. AI is pulling forward clean energy investment in geographies like Texas, where solar has ramped extraordinarily fast. It is also pulling forward long-duration gas peaker builds and locking in 20- to 30-year exposure to a commodity whose pipeline infrastructure isn’t keeping pace. LNG export prices set the floor on domestic gas prices. The hedge against that exposure ( for utilities, hyperscalers, and entire economies) is storage that runs on something else.
At 2,400°C, carbon blocks are one answer to that question. There are others. But the framing -abundance rather than scarcity, value-by-grid-service instead of hours-of-duration, FOAK as the true capital chock point - is correct regarding where the next decade of energy storage is heading.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.