But it wasn’t the only issue at play. Critical minerals were another hot topic, while conversations around adaptation and resilience are finally becoming a bigger part of the picture. I also moderated a surprisingly technical panel on distributed energy resources and virtual power plants, though that inevitably managed to touch on data centers and strategies for managing AI-driven load growth, too.
At Heatmap House, our day of conversations and roundtables with leading climate thinkers, one investor mentioned he had recently backed a lab-grown meat startup – a true contrarian investment if I’ve ever seen one. And my colleague Robinson Meyer hosted a fascinating pair of back-to-back conversations on a controversial geoengineering approach known as solar radiation management, which proposes using aerosolized chemicals to reflect sunlight away from Earth. He first spoke with the CEO of Stardust Solutions, a private company actively building this tech, followed by an advocate for research into solar engineering but certainly not near-term commercial deployment.
It’s impossible to capture the exact essence of a conference with hundreds of individual events — at some level, it’s always going to be what you make of it. But as I bopped around the city shaking hands, I picked up a range of interesting perspectives, along with three pieces of news that I thought were worth unpacking here — one related to funding for critical minerals, and two focused on bringing data centers online as quickly and cleanly as possible.
Atana Elements Spins Out of Lilac Solutions, Raises $27.5 Million
At a Climate Week event, Atana Elements CEO Thomas Wilson disclosed that the critical minerals exploration startup has quietly closed its seed round, which totals $27.5 million, according to an SEC filing. The round includes participation from Earthshot Ventures, as well as Lowercarbon Capital, and Hitachi Ventures. Last year Atana officially — but stealthily — spun out of Lilac Solutions, a startup developing a cleaner method of extracting lithium from saltwater brines.
But while Lilac is focused on commercializing its novel lithium extraction technology, Atana is tackling the broader upstream mineral discovery process. Its scope includes lithium, but extends to other so-called “flowing” critical minerals dissolved in brines, such as helium, hydrogen, and copper. In the years before the spinout, Atana compiled reams of historical geological datasets — think “Soviet-era oil and gas reports,” Wilson said. It used these to train predictive artificial intelligence models designed to identify previously overlooked mineral deposits.
“You can think of Atana as somewhat analogous to Kobald, but for flowing minerals such as lithium brines rather than hard rock resources,” said Matt Logan of Earthshot Ventures at the event, hosted by the nonprofit climate tech investor Elemental Impact. Kobald similarly uses AI for minerals discovery, and following a $537 million Series C round last year, is reportedly valued at nearly $3 billion.
Atana formed as a team within Lilac back in 2019, benefiting from the more mature startup’s relatively long and well-funded runway — Lilac has raised about $315 million to date. “We have found some of the biggest deposits in the world, and we’ve drilled 19 exploration wells across three continents,” Wilson said. “Around 2% to 3%of the world’s new minerals have been found by this particular team.” That’s a huge number for a startup that’s yet to even formally launch.
To date, Atana has identified a high-grade lithium brine resource in an Argentinean salt flat and secured 1.5 million acres across Germany and Poland, where it’s conducting exploration for lithium brine deposits. While lithium is likely to remain a core market, Wilson said he’s looking forward to broadening Atana’s ambition, asking “now that we’ve been released from the Lilac lithium play, what can we do in copper, helium, hydrogen, and where can we do that in other parts of the world?”
Silicon Valley Power, Emerald AI Launch First Pilot to Accelerate Flexible Data Center Interconnections
Data center-driven load growth, speed-to-power, and grid flexibility dominated the conversation at SF Climate Week, and the much-hyped data center management platform Emerald AI came prepared with a fitting announcement: It’s partnering with Silicon Valley Power, Santa Clara’s municipally owned utility, not only to demonstrate the benefits of flexible data centers for the grid, but to actually attempt to implement a program that expedites grid interconnection for data centers with flexible loads.
The latter objective differentiates this from Emerald AI’s earlier utility pilots, which were primarily technical demonstrations of its software — proving it can slow, pause, or reroute AI workloads during periods of peak demand without disrupting critical operations, which research shows could unlock nearly 100 gigawatts of grid capacity. This new pilot appears to go a step further by explicitly linking that flexibility to interconnection outcomes. As Emerald AI’s business development lead Daniel Padilla confirmed at a panel, data centers operating flexibly in Silicon Valley Power’s territory “will get material acceleration in time-to-power.”
Santa Clara, which sits about 45 miles south of San Francisco, is a major West Coast data center hub, with roughly 58 facilities packed into 19 square miles, according to Chris Karwick, Silicon Valley Power’s assistant director of utility operations, who spoke later at the same event. Karwick confirmed that the pilot with Emerald includes a “flexible load interconnection program,” and noted that while utilities broadly recognize the need for solutions to rising data center load growth, few are eager to be first movers. “We’re the electric utility for a city. We’re not known for being innovative — we’re usually followers. So this is big for us,” he explained.
Since emerging from stealth last summer, Emerald AI has already raised $67.5 million, and is now working with Nvidia to develop a 96-megawatt flexible data center facility in Virginia called Aurora, which Padilla said is expected to come online in October.
Meta Locks in 100 Gigawatt-Hours of Ultra-Long Duration Storage from Noon Energy
As Heatmap’s end-of-year survey revealed, experts widely consider Meta to be among of the worst hyperscalers when it comes to its climate impact and sustainability efforts. But the company nevertheless maintains a net-zero by 2030 target, even as it continues to bring plenty of new natural gas capacity online to power its AI expansion. Now, however, the company is throwing its weight behind a markedly greener — and less proven — technology, the ultra-long duration energy storage startup Noon Energy.
Meta announced this week that it has reserved 100 gigawatt-hours of storage capacity from Noon, which completed a successful demonstration of its 100-plus-hour carbon-oxygen battery earlier this year. Noon’s system charges by breaking down CO2 and discharges by recombining it using a technology known as a reversible solid-oxide fuel cell, and is certainly one of the earliest-stage data center power technologies that Meta has supported.
“There’s an urgency now that I don’t think existed before,” Carolyn Campbell, head of clean technology innovation at Meta said at a Climate Week panel, referring to the need to deploy emerging energy tech to meet the surge in data-center driven electricity demand. She added that Meta is evaluating how its procurement strategy can help commercialize early-stage climate tech — an area it so far hasn’t backed as extensively as its peers Google and Microsoft.
“When we sign a partnership agreement with a new company, does that help them with their next financing round because their investors see a different level of interest in the technology than they would have otherwise?” Campbell speculated. “Can we provide some upfront development capital to support a pilot that was maybe conceptual — going from concept to reality? So I think that’s one of the things that I’m really excited about with the Noon partnership.”
As I reported earlier this year, Noon CEO Chris Graves expects initial commercial deployments to begin as soon as next year, with early systems installed onsite to allow data centers or other large loads to draw power directly from Noon’s batteries rather than interconnecting to the grid itself. The startup’s collaboration with Meta will kick off with a 2.5-gigawatt-hour project, scheduled for completion by 2028.
A cutaway of the MARVEL reactor, showing its size relative to a person. Idaho National Laboratory
