Small clean energy startups are landing deals that would have seemed impossible just a few years ago: short-term contracts with major tech companies at $100 per megawatt-hour, or roughly double typical market rates.
Hyperscaler willingness to pay such a premium doesn’t necessarily extend to the whole power sector. But for the companies offering what the tech companies want — generation of clean, firm power that they can get online fast — it’s a lucrative moment. That said, for those lucky startups, the question of whether current demand signals can last remains open.
Last week, the Chinese AI company DeepSeek announced what appears to be breakthrough efficiency gains in its large language models. And as a result, both AI and energy companies are grappling with potential changes to their most fundamental assumptions about how much energy the technology requires. But the development also raises questions for the wave of energy startups banking on the tech giants’ seemingly bottomless appetite for power.
Gabriel Kra, the founder of the investment firm Prelude Ventures, has seen a steady uptick in small companies seeking venture capital backing based on small deals with AI developers.
“I am seeing more and more companies coming in and saying ‘Hey, we spoke to Amazon, we spoke to Meta, we spoke to Google, they’re offering us this trial, they are offering us this pilot, they’re offering us these prices,’” said Gabriel Kra, founder of Prelude Ventures. “I do not doubt that those companies are getting those actual price indications from the hyperscalers.” (Editor’s note: Prelude Ventures is an investor in Latitude Media.) But as the DeepSeek news suggested, those prices aren’t guaranteed to endure as the booming industry evolves.
These startups — pitching everything from geothermal and nuclear to biomass burning, natural gas, and energy storage — are bringing agreements that range from MOUs to take-or-pay contracts, often with the potential to scale up in coming years. The majority are structured as virtual PPAs rather than behind-the-meter arrangements.
“If you have a credible delivery of this power in 2027, 2028, those customers are willing to pay what, by every other measure, is an exorbitant fee for that power,” Kra said. “That’s a story, I think, that hasn’t run its course yet…We’re not eager investors at this moment in companies that need to sell their electricity at 100 bucks an hour. We’re not sure we believe that’s a long-term trend.”
Investor outlook
Prelude is relatively comfortable with the types of agreements startups are bringing with hyperscalers, Kra said — even though they’re not necessarily long-term PPAs.
What he’ll be watching for, other than whether these startups can deliver on their systems, is whether they can get their costs low enough to be able to sell to a wider market, outside customers like hyperscalers who are willing to pay a huge premium.
Another key question for Kra: “whether this AI driven energy market dynamic starts to genuinely spill over to other commercial customers and residential customers as to whether they start paying higher electricity prices.” If that happens, there are likely to be “really interesting consequences for how electricity gets regulated in this country,” Kra said. “It’s the one place that public sentiment could really change things…I think I’ll be paying a lot of attention to how that develops.”
For clean energy investors, the current trend of high energy prices has a little bit of ironic déjà-vu, he added.
Several years ago, PPA prices were as low as $20 per MWh. And thanks to major advancements in solar and wind efficiency, lower interest rates, and less grid congestion, Kra said the industry was speaking in terms of “energy abundance.” And at that time, investors were fielding pitches for business plans predicated on accessing $20-per-MWh electricity.
“People would come in and talk to us about all the wonderful things that they’re going to do and build with this cheap, abundant electricity and this cheap, abundant energy,” he explained. “And we’d push back on these companies and say ‘Wow, I don’t know where you’re getting this $20 power…what happens to your long run model at $50?’ And that knocked a lot of business models right out of the playing field.”
It was a prescient response on Kra’s team’s part; prices haven’t been that low since early 2020.
There’s of course a difference between the 2020 companies, who were consumers of power, and 2024’s crop of startups, who are producers. Today’s market is driven by the hyperscalers, who have a need for speed to power — and are willing to pay $100 per MWh to get it.
But again, Kra said, these prices may not last: “Is this any more believable as a long-term business plan than 20 bucks per metawatt hour as a customer? Can you rely on 100 bucks a megawatt-hour if you’re a supplier?”
