Climate tech’s $26 billion half is really the data-center trade

The biggest climate tech deal of the first half of 2026 was a $4.5 billion Series C for DayOne, a Singapore data-center developer that until recently was the international arm of Chinese operator GDS Holdings. The second biggest was $2 billion for Nscale, a London AI-cloud builder backed by Nvidia, at a $14.6 billion valuation. Neither company sells decarbonization. Both sell racks.

Those two deals sit at the top of Currence’s H1 2026 funding report, published Monday, which puts climate tech venture investment at $26.1 billion for the half. That is up 55% year over year and the strongest first half since 2022. It is also, on inspection, substantially one trade: low-carbon data centers took 34% of all funding. DayOne and Nscale alone account for a quarter of the entire tally. The report’s Built Environment vertical, which now houses the data-center deals, grew more than 800% and overtook Energy as the largest sector — a sentence that would have been unintelligible to a climate investor two years ago.

A boom that fits in ten term sheets

Currence frames the moment expansively. Discussing DayOne’s round, the report says:

The AI compute race is driving a VC boom even larger than the heady days of 2021-2022.

Test that against the report’s own deal count and the framing wobbles. Deals fell 25% year over year, to a five-year low. The top ten deals captured 42% of all funding. Series C rounds went from 16% of funding a year ago to 40% now. The 2021 boom was wide — money spraying across seed decks for everything from alt-protein to carbon accounting. This is narrow and tall. Fewer companies are raising, and the ones that do raise enormously, because they have a customer whose demand curve is drawn by Nvidia’s shipping schedule.

Whether the half even counts as growth depends on whose spreadsheet you open. Latitude Media, working from Net Zero Insights data, counts $41.3 billion across private climate financing including debt — down from $43.6 billion in H1 2025, with equity deal count at a record low and average round size jumping from $18 million to $27 million. Same market, opposite headline. Both are honest. “Climate tech is booming” and “AI compute ate climate finance” are the same dataset wearing different heds; the choice of which to write is mostly a choice about whether a gigawatt of grid capacity procured for GPUs counts as climate progress or as a customer acquisition.

What the trade left behind

Look away from the data-center column and H1 gets bleak fast. Carbon-markets equity fell 61%, its weakest half since 2020 — though Currence notes a structural reason alongside the cyclical one: “capital is also taking the form of offtake instead of equity,” with buyers contracting for future removal tons rather than investors taking ownership stakes. That softens the collapse without erasing it. Low-carbon fuels dropped 56%, which the report pins on sunsetting US subsidies and pending European policy reviews. Growth-stage funding overall slid 18% to $3.6 billion.

The exceptions prove the pattern. Series A funding rose 55% to $4.5 billion, powered by nuclear startups — companies whose pitch decks open with data-center demand. The biggest deal outside energy and transport was ICEYE’s $521 million for radar satellites, an adaptation play that watches disasters rather than preventing them. Everything growing has either a hyperscaler or an insurance company at the end of it.

The public markets tell the same story with better liquidity. Fervo Energy’s May IPO raised $1.89 billion and popped 35% on debut — the largest clean-energy IPO on record, per Latitude. X-Energy, the Amazon-backed small modular reactor developer, raised about $1 billion in April in what TechCrunch called, accurately, a data-center-driven IPO. The bull case writes itself: AI demand is finally the offtaker of scale that clean firm power spent two decades waiting for, and geothermal and nuclear companies that could never reach commercial escape velocity on climate sentiment alone are now getting there on compute hunger.

The bear case is that the money follows the workload, not the carbon. If the hyperscaler is the customer, “low-carbon” is a procurement preference, and preferences bend when the queue gets long.

Same electrons, different ledger

Monday made that concrete. The same day Currence published, Williams announced a $5.34 billion investment from a Blackstone-led consortium — with Apollo and KKR participating, as Bloomberg reported — for a 49% stake in its Power Innovation portfolio: five behind-the-meter, gas-fired generation projects serving data centers, 2.6 gigawatts announced and more than 6 gigawatts in backlog.

Be precise about what this is, because it is not venture capital and it appears in nobody’s climate tally. This is infrastructure and asset-based credit money buying a noncontrolling equity position in a listed pipeline company’s gas plants — Blackstone Credit & Insurance, not the growth funds writing checks to Fervo. Conflating the two would flatter the climate numbers and muddy the point. The point is sharper kept separate: the same customer that pulled $8.9 billion of “climate tech” venture funding this half is simultaneously pulling multibillion-dollar credit deals for new gas combustion, for the same electrons, on the same timelines. “Williams is a leader in meeting the country’s rapidly growing power demands,” Blackstone’s Robert Horn said in the release. Carbon does not come up in the sentence.

Our read: the H1 report is best understood not as a climate tech recovery but as the moment climate tech’s definition became the contested asset. The category now includes anyone selling low-carbon electrons or efficient racks to AI, which pulls the totals up while the deal count — the measure of how many distinct bets the market is making on decarbonization — hits a five-year low. The AI-demand story is genuinely funding clean firm power; Fervo’s balance sheet is real. It is also genuinely funding gas, through doors the climate dataset doesn’t track.

Williams named its five behind-the-meter projects Socrates, Apollo, Aquila, Neo — and Socrates the Younger. Williams holds rights to buy its partners out between years seven and fourteen, which is to say the plan runs to 2040. They burn natural gas.