Anurag Kamal is the co-founder and CEO of ElectricFish, a company rebuilding the electrical grid from the ground up through a decentralized network of community-scale batteries. Each ElectricFish unit is a 400 kWh system that sits between a home battery and a utility-scale one — hardware and control software in a single box — deployed at gas stations to enable extremely fast EV charging while quietly turning that site into a two-way grid resource. The flagship product was unveiled at CES 2026, and units are already live across six U.S. cities.
In this conversation, part of our “AI Meets the Physical World” series, Anurag explains why the energy challenge is a delivery problem rather than a generation one, why ElectricFish is emphatically not an EV charging company, and why the real prize isn’t the electrons you sell but the moments you show up for the grid.
Topics include:
Why “we’re not an EV charging company” and why calling them one is like calling Amazon a bookstore
The generation-vs-delivery reframe, and why we curtail terawatt-hours of clean energy
Cutting peak load at the source vs. the ~$1 trillion in utility upgrades built for four worst hours a year
Everything-in-one-box vs. co-located batteries: how permitting and footprint decide scalability
Gas stations becoming “energy stations” as homes start generating their own power
The three revenue layers — EV charging, energy arbitrage, and grid services as “a hedge fund of energy”
The capital stack today (VC + project finance + grants) and how asset ownership shifts after Series B
China hardware, U.S. manufacturing, and why service and software are the durable edge
Distributed compute and powering the “missing middle” once the centralized data-center build-out cools
The data-center build-out gets blamed for straining the grid, but Anurag’s framing is sharper: the build-out didn’t create the bottleneck, it exposed one that was already there. We can make plenty of power. We just can’t get it where it’s needed, when it’s needed. Each year California and Texas curtail terawatt-hours of clean generation — energy that gets thrown away because there’s nowhere to put it.
“We don’t have an energy generation issue. It’s more of a delivery issue. Curtailment means you’re just killing the energy, putting it to the ground, because you can’t store it.”
Renewables are cheap and abundant; the hard part is intermittency. And the old answer — co-locate a giant battery next to a giant solar or wind farm and ship the power across long transmission lines — is, to Anurag, last century’s logic applied to this century’s problem. Generation has decentralized. Storage hasn’t followed it yet.
“The old way was centralized because you couldn’t have a single solar panel generating energy. Now renewable energy is abundant, it’s everywhere, and everybody can own it. So a new distributed network of storage needs to emerge.”
The punchline is an economic one. The cost of energy is migrating away from the fuel you burn and toward the infrastructure that stores and moves it.
“The cost of energy is not going to be mostly in generation anymore. It’s going to be in storage and delivery.”
ElectricFish’s beachhead is extremely fast EV charging at gas stations, where landowners are eager to diversify a business under long-term pressure. But Anurag is adamant about what the company actually is — and the analogy he reaches for is deliberately grandiose.
“Even if EV charging is how we’re getting sites today, calling ourselves an EV charging company would be like calling Amazon a bookstore. In the early days, people did call them that.”
The real product is a grid resource: a fleet of distributed batteries that can absorb, store, and re-inject power on demand. EV charging is simply the wedge that gets the hardware into the ground and the unit economics moving. Every architectural decision — where to deploy, what to build, how to go to market — is made against that longer arc, not the charging business that funds it today.
Anurag knows most of his competitors and their founders personally. Some bring mobile batteries to a site; some co-locate a large stationary battery. He respects the transitory use cases — racing, roadshows, places you can’t build permanent infrastructure — but doesn’t believe they scale to mass penetration. His objection to the “big co-located battery” model is brutally physical.
“Have you seen a neighborhood gas station? There’s no space. They’re using every inch they have. If you want to put a big battery, you need a power cabinet, a dedicated parking spot, you rip up their lot and run thick conduits of copper. I don’t know how that scales.”
ElectricFish’s answer is integration. Battery, fast charger, and optimization all live in a single enclosure, so the only thing the site has to bring in is a thin trickle of grid power.
“Everything is packed inside the box. The only power you need to bring in is 100 amps through a tiny 1.5 to 2 inch conduit.”
That simplicity is the scalability thesis. Gas stations are everywhere, they already store community energy in the form of fuel, and as homes start generating their own power, the site’s role flips.
“The gas station is going to change into an energy station — storing energy for the community the way it stores fuel today.”
To understand why distributed storage matters now, Anurag points at how utilities are forced to think. The grid is sized for the worst-case hour — the few hours a year when every AC, oven, and light switches on at once — and then over-built to survive them.
“You’re always planning for those four worst hours, and then you beef up the whole infrastructure just for them. Imagine how much wastage that is.”
Utilities have papered over this for years by deferring upgrades, pushing the cost down the road and onto ratepayers. But the bill is coming due, and AI load is accelerating it. The scale of the spend is staggering — and most of it is the least glamorous part of the system.
“Around a trillion dollars is what utilities expect to spend in the next six or seven years in the US alone, and 60% of it is just distribution upgrades.”
ElectricFish’s pitch is to make a chunk of that spend unnecessary by cutting peak load at the source, before it ever reaches the distribution or transmission grid. EV charging is the first peak it shaves — five times a gas station’s load, hundreds of times a home’s — but the same logic extends to holiday surges and, eventually, to local compute load. At network scale, the model gets ambitious: Anurag estimates a few thousand units in a state like California could blunt the kind of large, sudden swings that data centers and weather impose on the grid.
A unit like this has layered revenue. The first is the obvious one: charging EVs. The second is arbitrage — buy power when it’s cheap, sell it when it’s dear, and live on the spread, the same instinct a gas station’s floating fuel price already encodes. But the layer Anurag built the company around is grid services: being in the right place, at the right time, with stored power the grid is willing to pay a premium for.
“Grid services is much bigger than the arbitrage. It’s understanding how much the grid will pay for a certain amount of storage and power at a certain time. It’s almost like running a hedge fund of energy.”
It’s not steady income — it’s optionality on volatility, the value of being available when the system is stressed. He cites Texas, where prices have spiked a hundred- or thousand-fold in a crisis, but is careful not to romanticize the chaos.
“It won’t be a lot of money every day. It’s about predicting those crisis moments and being available during them.”
Anurag is a battery researcher by training, and the constraint he keeps returning to is that a battery is a living, electrochemical-thermal system, not a static asset. The complexity isn’t only in building one — it’s in keeping it cool, safe, certified, and economically serviceable across a long life. Get the maintenance economics wrong and the whole model collapses.
“This is batteries plus power electronics, and batteries need to be kept happy. If you’re spending too much keeping the system up, that’s eating your margins, and you won’t be successful scaling this.”
Which is why he’s dismissive of the idea that selling cheap hardware is a business at all.
“Just selling this hardware doesn’t seem like a great business. That’s what the Chinese are doing — another race to zero. But this is a 10-to-15-year asset. If it can’t run with high uptime at reasonable maintenance cost, it doesn’t work.”
His view on Chinese batteries is refreshingly unideological. He’s grateful for the vendors he buys from; the quality meets his spec. Domestication matters, but cheapest-bill-of-materials thinking misses the point when the asset has to perform in the field for over a decade. The defensibility, he argues, isn’t in owning the cells — it’s in service, localization, and the software that decides when a fleet of assets charges, discharges, and bids into the grid.
“We don’t have to be a hardware manufacturer always. We can expand more to software and let it work with other hardware too — as long as we understand the markets and can dispatch the assets.”
A customer chasing the lowest sticker price might save upfront, but loses on the long-run ROI if the system isn’t optimized, serviced, and tuned to their schedules. That gap — quality plus local service plus intelligent dispatch — is where Anurag thinks the durable margin lives.
Because ElectricFish often retains ownership of the units, the natural question is whether this is venture or project finance. Anurag’s answer is that it’s both, and that the mix shifts as the technology de-risks. Today, gas station owners want the solution but can’t underwrite the tech risk — and neither can their banks — so project-finance partners and grants carry the early deployments, with owners taking a small revenue split.
“Right now it’s venture capital plus project financing and grants — a lot of grants. I’m very grateful for the innovation economy of California and other states.”
As the product proves out, the capital structure is designed to flip: the asset becomes bankable, owners borrow against it themselves, and ElectricFish steps back from financing while keeping operational control.
“In three or four years, owners will get their banks or cooperatives to finance it because the product is proven. We’ll make a smaller margin, but we won’t be project-financing it ourselves.”
The most forward-looking thread is where distributed energy meets distributed compute. Today’s AI build-out is centralized — enormous data centers with enormous power needs, built far from the cities they serve. Anurag’s instinct, as a distributed-grid person, is to question the premise.
“Why are we doing it in a centralized way? After this phase of the centralized build-out is done, we’ll have to get to distributed compute — so we’re not dependent on moving terabytes to the data center continuously.”
If inference moves closer to where it’s consumed, it needs power closer to where it’s consumed too — and that is exactly what a community-scale battery network provides.
“We are distributed energy. We can power distributed compute in the communities, and we’ll fill the missing middle of compute.”
He’s testing prototypes in the field and running the numbers, but he’s disciplined about not over-promising on a thesis that’s still maturing — a fitting note from a founder who treats hard physical constraints as the thing to respect, not wish away.
“I’d be very, very happy to talk about that maybe two years from now.”
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.