Antora Raised 550 Million To Sell Heat, Not Power

The flagship customer is not a data centre

At Big Stone City in South Dakota, a POET Bioprocessing ethanol facility is being fed by more than 200 thermal battery modules totalling 5 gigawatt-hours, under a long-term heat offtake agreement. That plant, not a server hall, is the reference installation behind the 550 million dollar Series C that Antora closed on 30 July, co-led by G2 Venture Partners and Eclipse, with Ribbit Capital, Salesforce Ventures, Activate Capital, John Doerr, the Westly Group, StepStone Group and Liberty Mutual Strategic Ventures joining alongside existing backers including Breakthrough Energy Ventures and Lowercarbon Capital.

The headlines put a data centre in the first line and the invoice does not. Antora's own chief executive, Andrew Ponec, framed the round around a broader claim, that energy is the bottleneck to industrial growth. Canary Media reports the company has signed agreements with some of the largest data centre operators and that future systems could run five to ten times the size of the South Dakota installation. Those are forward commitments. The revenue that proved the machine came from making steam.

The company has now raised roughly 770 million dollars since its founding in 2017, following a 150 million dollar Series B in February 2024, and separate reporting puts total corporate and project financing near one billion dollars. Its manufacturing sits in San Jose, California, expanded to a three-building campus. Read the round as the scaling of a working industrial product rather than the funding of a data centre thesis.

Efficiency is the wrong test to apply

The instinct with any storage technology is to ask about round-trip efficiency, and on that measure a thermal battery looks poor. Antora's system uses electricity to resistively heat blocks of solid carbon to as much as 2,400 degrees Celsius, then converts the emitted light back to electricity using thermophotovoltaic cells. Electricity in, heat, then electricity out again is a lossy chain compared with a lithium-ion cell, and no amount of engineering makes it otherwise.

That comparison only matters if the customer wanted electricity. If the customer wanted heat, the conversion step never happens and the loss never occurs. This is why the ethanol plant is the honest test case: it buys the output of the first half of the process. Applying a battery benchmark to a device sold as a furnace is a category error, and it is the error most coverage of this round has made.

The number that deserves the attention is 2,400 degrees. Heat pumps and electric boilers have made real progress at low and medium temperatures, and industrial process heat above roughly a thousand degrees has stayed stubbornly fossil because nothing electric reached it economically. Cement kilns, glass furnaces, steel reheat and much of heavy chemicals sit in that band. A device that reaches it addresses a segment that electrification has been unable to touch, which is a different and larger claim than storing power for a server hall.

Twelve months against a queue measured in years

The operational fact worth carrying out of this announcement is a schedule. Antora says the Big Stone system advanced from initial construction to delivering energy in under twelve months, with commissioning beginning in May 2026. Set that against what an industrial site or a compute campus faces when it asks for more grid capacity, where connection dates in Europe and the United States alike are quoted in years rather than seasons.

Speed, not cost, is the property being sold here. A thermal store does not create energy; it moves consumption from hours when power is cheap and plentiful into hours when it is not. For a facility whose constraint is a peak connection it cannot get, that shift changes what it has to ask the grid for. The interconnection request is priced against the peak, and a store lets the site present something closer to its average.

This is the same logic that has driven compute operators to build their own generation rather than wait, a pattern this journal has traced through fuel-cell campuses and private turbine fleets. Thermal storage arrives at that problem from the opposite direction: instead of adding a private power plant to escape the queue, it reduces the size of the connection the queue has to grant. Both are answers to a wait, and only one of them puts a combustion permit on the operator's desk.

What has to be true for this to reach Europe

None of this is deployed in Europe, and the company's own framing is explicitly American. That is the first thing a European operator should hold onto before treating this round as a procurement signal. The manufacturing is in California, the flagship is in South Dakota, and the stated purpose of the capital is a second domestic manufacturing hub and a strengthened domestic supply chain.

The economics rest on a spread, and the spread is local. A thermal store earns the difference between the price of power in cheap hours and the cost of the fuel it displaces in expensive ones. South Dakota supplies that spread through abundant low-cost wind. A German, Spanish or Nordic industrial site has to check its own version of that number, and the honest answer differs sharply by market: Iberian and Nordic price curves look very different from a German one, and an Italian one different again.

Two European factors push the other way and are worth weighing properly. Industrial electricity is more expensive here than in the United States, which hurts the charging side of the arithmetic. But the carbon price under the EU Emissions Trading System is a cost on the fossil heat being displaced, and it does not exist in the American case at all. Whether those two cancel is a site-level calculation, not a continental one, and anyone selling you a continental answer has not done it.

What to do with this before it is buyable

For most European operators the correct action this quarter is measurement rather than procurement. Pull the hourly power price history for your own connection point over the last two years and calculate how many hours sat below the level at which storing heat would beat burning gas. That single number determines whether any of this technology class is relevant to your site, and it is knowable today without a vendor conversation.

Then find out what temperature your process actually needs. Many industrial sites carry a mixed heat load, and the portion below roughly 200 degrees is already addressable by heat pumps that can be bought in Europe now. Splitting the load by temperature tells you which part is a live decision and which part is waiting on a technology that has not arrived here yet. Firms tend to treat process heat as one line item, and that is why the addressable share is usually underestimated.

Finally, keep the claim in proportion. One 5 gigawatt-hour installation at one ethanol plant, however impressive its build time, is a proof of capability and not a track record. The useful posture is to know your own spread and your own temperature curve, so that when this class of equipment does appear on a European price list you are evaluating it in weeks rather than starting the arithmetic from zero.

Read next: One Commissioner Stalled a 2 Gigawatt Campus  |  The Turbines Come Out In 2027, Not This Year