You’re out of free articles.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
Sign In or Create an Account.
By continuing, you agree to the Terms of Service and acknowledge our Privacy Policy
Welcome to Heatmap
Thank you for registering with Heatmap. Climate change is one of the greatest challenges of our lives, a force reshaping our economy, our politics, and our culture. We hope to be your trusted, friendly, and insightful guide to that transformation. Please enjoy your free articles. You can check your profile here .
subscribe to get Unlimited access
Offer for a Heatmap News Unlimited Access subscription; please note that your subscription will renew automatically unless you cancel prior to renewal. Cancellation takes effect at the end of your current billing period. We will let you know in advance of any price changes. Taxes may apply. Offer terms are subject to change.
Subscribe to get unlimited Access
Hey, you are out of free articles but you are only a few clicks away from full access. Subscribe below and take advantage of our introductory offer.
subscribe to get Unlimited access
Offer for a Heatmap News Unlimited Access subscription; please note that your subscription will renew automatically unless you cancel prior to renewal. Cancellation takes effect at the end of your current billing period. We will let you know in advance of any price changes. Taxes may apply. Offer terms are subject to change.
Create Your Account
Please Enter Your Password
Forgot your password?
Please enter the email address you use for your account so we can send you a link to reset your password:
Artificial intelligence wants the energy and has the money, and climate tech companies need buyers.

Their founders wanted to make transmission lines, powertrains, and electrical switches more efficient. Or maybe they wanted to unlock the potential of geothermal energy or low-carbon cement. Wherever they began, a bevy of deep tech climate startups, clean energy producers, and sustainable materials companies have found their way to the same destination: Building and powering data centers in the most energy efficient way possible.
“They might not have started out as data center companies, but they’ve been pulled — because of this huge market movement towards data centers — into being that,” Lee Larson, an investor at the venture firm Piva Capital, told me.
With power demand from artificial intelligence on track to grow as much as 30x from 2024 to 2035, and the Trump administration seeking to fast track data center buildout, there’s a wealth of opportunity — and literal cash — for startups that can help hyperscalers meet their clean energy targets while cramming as many high-powered computing chips into a data center as physically possible.
“I think the proportion of pitches that we see that reflects some kind of data center messaging has gone from maybe one out of 20 to one out of five,” Matthew Nordan, co-founder and general partner at Azolla Ventures told me. “It’s a lot.”
Perhaps the most obvious data center pitch is for companies offering clean, firm power or energy storage. In Azolla Ventures’ portfolio, that includes the geothermal exploration and development company Zanskar and the underground pumped hydro storage company Quidnet. While neither has announced any data center tie ups to date, both are having conversations with all the usual suspects — a group that includes Google, Microsoft, Amazon, and Meta. “Virtually any reasonably mature, ready to deploy clean, firm power technology company is talking to the same people,” Nordan told me.
Some big deals have already made headlines, especially in the nuclear and geothermal sectors. There’s Microsoft’s plan to reopen the Three Mile Island nuclear plant and Google’s deal with the small modular reactor startup Kairos, plus Fervo’s partnership with Google and Sage Geosystems’ partnership with Meta on the geothermal side. But fusion companies also see data centers as a viable option. Google already has an offtake agreement with Commonwealth Fusion Systems, while Microsoft has a deal with Helion Energy.
But it’s not just the big name cleantech companies that are turning into data center service providers. The AI boom also presents a major opportunity for deep tech startups working on electrical infrastructure. While companies in this sector might not scream “climate tech,” behind the curtain they’re driving significant gains in energy efficiency that data center operators are eager to tap into.
In Azolla’s portfolio, these include Scalvy, founded to build modular powertrain electronics for electric vehicles. The company’s small, distributed units connect directly to EV battery cells, converting DC power from the batteries into AC power for the motor. “The hyperscalers started coming to the company saying, can you do what you’ve done in reverse?” Nordan told me. “Can you take the AC coming in off the grid and then convert that to DC, and then interface with the load and energy storage systems?”
That proved easy, and now Scalvy’s small, building-block style approach allows data centers to control power flow on the server rack itself, as opposed to taking up valuable space with a separate power rack. While the details haven’t yet been announced, Nordan said the startup “has recently done their first agreement for data center power, and it’s with one of the large names that you would expect.”
Piva Capital has also invested in a number of under-the-radar companies in this arena — Veir, for instance, initially proposed to build “high-temperature superconducting transmission lines” that could carry electricity with near-zero resistance, and thus very low energy loss. But after seeing some early interest from data centers, the startup learned that hyperscalers were not only struggling to build transmission lines to their substations, but were also experiencing severe bottlenecks in their low-voltage distribution networks, responsible for getting power into and around data centers.
“We realized we can apply essentially the same superconducting technology that we’re targeting for transmission and distribution applications and build a low-voltage set of products for data centers, specifically, that can allow you to shrink the size and weight of conductors and bus bars [which distribute power within data centers] by 10 times,” Veir’s CEO Tim Heidel told me. With this newly refined focus, the company raised an oversubscribed $75 million Series B round in January, which included participation from Microsoft’s Climate Innovation Fund.
Piva is also an investor in Menlo Micro, a spinout from General Electric that uses a proprietary metal alloy to make high-performance electrical switches that are smaller, faster, and more energy efficient than the industry standard. The startup has already commercialized its tech for use in high-speed radio frequency devices, as well as for testing the performance of semiconductors.
Ultimately, the company is aiming to integrate its switches into a wide range of high-performance electrical equipment, data center power systems very much included. In this context, the startup’s switches could be embedded directly into semiconductor packages or circuit boards rather than installed on racks, leading to more compact and energy efficient data center power management. The switches’ small size and low resistance would also generate less heat than what’s used today, further increasing overall energy efficiency.
Menlo Micro’s CEO Russ Garcia told me that five years down the line, he expects a third of the company’s revenue to come from power applications such as data centers, growing to two-thirds in 10 years’ time.
Even sustainable materials companies are getting pulled in, Nordan told me. The primary example there is Sublime Systems, which inked a purchase agreement with Microsoft for up to 622,500 metric tons of low-carbon cement. The deal gives Microsoft the right to use the cement if and when it's useful, but more importantly, it entitles the tech giant to the cement's environmental attributes — that is, the carbon savings associated with producing it. The idea is that the tech giant can catalyze market demand without the emissions impact of shipping the cement to its data center sites.
Amazon has also invested in a number of companies in this sector, including Brimstone and CarbonCure, which are working to decarbonize cement and concrete, as well as Electra, which is working on green steel. The hyperscaler is also trialing products from Paebbl, which produces a carbon-negative mineral powder that can partially replace cement, on the construction of an Amazon Web Services data center in Europe.
While the current administration may not be exerting pressure on hyperscalers to reduce their emissions, Nordan told me that the tech giants are thinking about the long term. “If the tide turns and there will be real or effective costs to emissions in these data centers, they want to do everything they can to bankroll emissions reductions now. And that manifests itself in low-carbon cement, in green steel, in all sorts of technologies.”
At least some of the aforementioned investments — especially those that increase efficiency while decreasing the size of data center components — won’t necessarily lead to emissions reductions, however. Much as when the Chinese AI firm DeepSeek released its cheaper and more efficient AI model, the idea of Jevon’s Paradox looms large here. This is the theory that making products more efficient and cost-effective will lead to an overall increase in consumption that more than offsets the efficiency gains.
Heidel, for one, told me that Veir’s potential customers don’t see energy efficiency in itself as the startup’s main draw. “It’s actually the space savings, the real estate savings, the ability to lay out data centers and configure them in new ways,” he told me. Mainly what Heidel is focusing on with his customers-to-be is, “how much smaller can you make the building, or how many additional AI pods or servers could you fit into the same footprint, or how much higher of a server density could you achieve using our solution?”
Of course, one day Veir may fulfill its original dream of creating superior transmission infrastructure, just as Scalvy could circle back to its initial focus on EV drivetrains and Menlo Micro could wriggle its way into a whole host of electronic devices.
As Heidel told me, he sees this data center buildout as just the first push in what will be an ongoing effort to meet the world’s growing electricity demand. “If we can figure out how to serve all of this demand at the speed at which data centers are growing, and do so cost effectively, and do so in a low-carbon way, then we can take those learnings and apply them to all of the other industries that are coming in the future that'll also be facing enormous electricity demand,” he explained.
But for the time being, as Larson of Piva Capital told me, investors are simply trying to get their portfolio companies “to skate where the puck is going.” And that’s more than okay for Heidel. As he put it, there’s “so much enthusiasm for data centers today that we are having trouble just keeping up with all the interest in that market.”
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
We didn’t know days like this could happen. Then we learned how bad they really are.
When I woke up this morning in Chicago, the Air Quality Index was in the 300s, and I could barely see the top of the skyscraper across the street. The weather app on my phone featured a little image of a man wearing a World War I-style full-face gas mask. That’s fun, I thought. I didn’t know it could do that.
I went downstairs. Old photographs of the city were hanging in the hotel lobby — girls playing in bathing suits next to the lake — and I realized that the haze shrouding the old Lakeshore Drive condos was in fact haze, smoke, particulate matter, and not a lens artifact. It really used to be that smoky all the time, back before the Clean Air Act. Then I glanced up and saw that the haze out the window was far worse than the century-old pollution in the picture.
It’s significant, I think, that a mass smoke-out like this has now happened to the eastern U.S. for a second time. Second times matter. When exhaust from Canadian wildfires blanketed the Northeast and parts of the Midwest in June 2023, exposing more Americans to wildfire smoke than on any previous day in history, one could almost write it off as a freak occurrence. It was upsetting, sure, and reminiscent of California’s climate-addled amber skies. But didn’t wildfire smoke also descend on New England once in the 1780s? Even on a warmer planet, couldn’t this remain a once-in-a-century blip?
Twice in just over three years, though — that‘s more than a hiccup. That’s almost a trend. To get smoked out once may be regarded as a misfortune; for it to recur again, without any plan to respond, starts to look like carelessness. The federal government is doing roughly diddly squat about adaptation — President Trump can build a fan on the border and make Canada pay for it — but state and local governments across the eastern U.S. will now need to reckon with a new form of extreme weather. You grew up with snow days, but now we’ll have smoke days — and schools and sports leagues and concert venues will need rules about how to deal with them. When should games be canceled, tickets refunded? Is smoke more like a heat wave or a hurricane? Hotels and office buildings will need to review their ventilation policies and possibly upgrade their equipment; municipal emergency response plans will be revised and printed in triplicate.
All this will happen because the smoke has invaded a second time — and arguably a third, if you count last year’s minor episode — and that means it could come back again. For that reason, this event strikes me as a much bigger deal than what happened in 2023. The smoke is now a fact of life; institutions will need a policy about it. The tortious creep of litigation risk will enforce that outcome, even if no federal official enforces it.
So it goes. But to be clear, this new inconvenience is not what worries me most about today’s events. No, what frightens me instead is that today’s airborne toxic event is not something that was supposed to happen. Until a few years ago, we had not thought too hard about whether a major smoke exposure event like this could happen on the East Coast at all. It had not seemed possible.
For years, economists and climate scientists have simulated how global warming might affect the U.S. and global economies. They poured years of careful work into this modeling, and they simulated — with ever-increasing levels of statistical persnicketiness — what extreme heat and sea-level rise might do to agricultural yield, labor productivity, energy demand, heat mortality, and real estate values, among other potential sources of damage. This work was useful; it improved our practical understanding of coastal flooding, to name one example. It also helped calibrate U.S. regulatory policy, even if it never achieved the crowning heights of helping to set a national carbon tax.
Yet these careful models almost never accounted for mass smoke exposure days. Indeed, the kind of thing that happened this week — when heavy haze blows down from Canada and exposes more than 100 million people to hazardous air — was not countenanced by the simulations at all. Only in recent years did economists begin to study events like these, and only because mass exposure events like 2023’s happened first.
We’ve long known that the tiny shreds of particulate matter in wildfire smoke dance across the body’s barriers and penetrate its deep places, etching their way into lung, heart, and brain tissue. Inflammation follows. What makes days like today unique is the scale: Tens of millions of Americans inhaling wildfire smoke at the same time. As we’ve started studying this phenomenon, it’s become clear that the mortality effects of days like today, the deaths elevated above what you’d otherwise expect, can persist for years. That becomes extraordinarily expensive for society.
How costly? “When monetized,” a group of Stanford and Princeton economists wrote in Nature last year, in the first major study on the topic, “the climate-driven smoke deaths result in economic damages that exceed existing estimates of climate-driven damages from all other causes combined in the U.S.A.”
You read that right: The cost of climate-worsened wildfire smoke alone is larger than what earlier studies said every other estimated cost of climate change would be, combined.
To summarize, wildfire smoke did not appear in our economic simulations of climate change. As recently as a few years ago, we did not really know that days like today — or June 7, 2023; or September 15, 2020; or September 9, 2020 — could occur. Then they happened. And happened again. And then we studied them and discovered that, in fact, they may be more expensive for the U.S. economy than we once thought climate change itself would be.
That worries me. Now we know these smoke-out days can happen; now they are fast becoming a rare but predictable feature of summer life. But until recently they were unimaginable. What other ignominies, what other tail risks and airborne surprises, are lurking in the uncontrolled experiment we’re running on the biosphere? What else — unforecast, unmodeled, unstudied, unthought of — lies ahead? After 10 years of covering the climate system, I am not someone who lies sleepless fretting about atmospheric CO2. But I do wonder what else we don’t know enough about to ask.
“Microsoft, you can’t hide, we can see your dirty side!”
Protestors interrupted one of the final sessions of PNW Climate Week — a conference that brings together climate leaders across Washington, Oregon, and British Columbia — objecting to Microsoft’s rising carbon emissions from data centers and partnerships with oil and gas companies. The company’s Chief Sustainability Officer Melanie Nakagawa was having a one on one conversation with GeekWire climate reporter Lisa Stiffler at Seattle’s City Hall when protestors carrying signs reading “Microsoft’s AI pollutes” and other slogans began shouting from the audience.
I was there, having just moderated the prior panel on how to finance Washington’s clean energy ambitions. Early on there were some rumblings in the crowd from up front. “Climate leaders don’t build gas pipelines in Moses Lake,” was the first objection I heard clearly. It came shortly after Nakagawa kicked off the conversation by highlighting Microsoft’s partnership with sustainable aviation fuel startup Twelve, which recently opened its first commercial-scale SAF plant in Moses Lake, Washington. The tech giant has supported the project through a strategic investment from its Climate Innovation Fund, as well as an offtake agreement for the fuel that will help offset its emissions from employee travel.
Whether Microsoft is building a gas pipeline in this particular community I haven’t been able to determine, though it seems irrelevant to Twelve’s SAF facility, which doesn’t rely on natural gas. But it is true that Microsoft is one of the largest power consumers in Grant County, Washington, home to Moses Lake, where a natural gas pipeline operator is looking to expand its network to accommodate data center load growth.
Another audience interruption was more pointed. “How does signing a 20-year deal with Chevron help you reach your clean energy goals?,” one protestor asked, referring to Microsoft's recently announced power purchase agreement with Chevron for nearly 2.7 gigawatts of natural gas-fired power to supply a West Texas data center. The project represents one of the largest gas-powered artificial intelligence developments in the U.S., and Stiffler acknowledged that she had been planning to ask about it, herself.
Nakagawa answered the question. at least in part, saying “that project with Chevron is initially using natural gas and it’s a natural gas contract,” before emphasizing that the company has built “over 4.5 gigawatts of clean energy already today,” and remains committed to balancing speed-to-power with its clean energy goals. She added that, “with this deal in particular, we’re looking at a range of tools in our toolbox to ensure that we can continue to grow our power, but also do so in a way that is responsible and sustainable.” She stopped short, however, of making any commitments to transitioning the project to renewable energy over time.
The session became more chaotic from there. Another protestor stood up, shouting that “Microsoft is enabling genocide in Palestine.” Other activists joined in, while still other audience members shouted back. As Nakagawa recovered and resumed answering a question from Stiffler about Microsoft’s recent decision to pause its carbon removal purchases after years of dominating the nascent industry, protestors throughout the crowd began a chant of “Microsoft, you can’t hide, we can see your dirty side.” Security eventually shepherded many of them out.
Stiffler continued speaking with Nakawaga about the company’s clean energy efforts, touching on many of the protestors’ concerns as she asked about community opposition to data centers, the role of large corporations in the clean energy transition, and whether Microsoft can realistically achieve its goal of becoming carbon negative by 2030.
Nakawaga emphasized that the company must, “first and foremost, listen to where the communities are and what they are calling for.” Regarding the concerns she hears most often, she explained that “first has been transparency. Second has been around resource uses and what are we doing about those resource uses. We’re hearing about jobs and employment and investments in education, investments in housing.”
If this session was any indication, those concerns won’t go away anytime soon.
Heat kills more Americans than any other extreme weather event in the United States. But wildfire smoke — while not strictly “weather” — appears to kill even more. Current excess death estimates put American heat mortality at about 10,000 people per year, or possibly as high as 12,000. Recent studies on wildfire PM 2.5 exposure suggest a mortality of double that: 24,000 all-cause deaths every year.
Needless to say, wildfire smoke is definitely not something you want to inhale if you can avoid it. (And really, you should try to.) But for the 115 million Americans in the Great Lakes and Northeast regions of the country who’ve been exposed to hazardous air from the fires in Ontario and Minnesota this week, there’s a chance that the damage is already done. According to a wildfire smoke mortality estimation tool from Cornell University’s School of Public Health and the Northeast Regional Climate Center, the total mortality for this smoke event could already be as high as 424 people so far, including nearly 100 in Michigan and more than 50 in both New York and Wisconsin.
Alistair Hayden, an assistant professor of practice in Cornell’s Department of Public and Ecosystem Health, stressed to me that the tool is a “first draft,” and that his team is still working on getting it peer-reviewed. “We intend it as a hypothesis that people can test in the coming weeks or months to confirm our numbers,” Hayden told me. “I’m really hoping to be proven wrong.”
But Hayden also emphasized that while the West Coast might historically be where many smoke-related deaths have occurred, “this is the third out of four years [in the Northeast] that we’re having the smoke, so it seems like something we should be planning for,” he said. “It reminds me of that saying: ‘Fool me once, shame on you. Fool me twice, shame on me.’”
Admittedly, the smoke this week is a bit of a freak occurrence. A cooler-than-average sea surface pattern across the North Pacific, known as a negative phase of the Pacific Decadal Oscillation, helped produce weak low-pressure areas in the northwestern part of the United States, which in turn allowed for heat domes to develop across the Southwest and Plains. After one did just that earlier this month, the hot, high-pressure dome then shifted north, where it developed “dryness across Canada, followed by the lightning-producing thunderstorms,” Chad Merrill, a senior meteorologist at AccuWeather, told me. Then, boom: widespread fires.
“It is very unusual to have a combination of an El Niño and a negative phase of the Pacific Decadal Oscillation,” Merrill went on. “That’s one of the unusual factors this year, which contributed to the heat dome being farther north in that particular position.” The heat dome and jet stream then worked together to direct the thick smoke down into some of the most populous regions of Canada and the U.S.
That’s what makes this particular smoke event so bad. Were the smoke blowing over remote regions of Canada, as it would under more usual conditions, “then the big cities and the Great Lakes wouldn’t experience the smoke; it would have gone north toward the Hudson Bay and then Greenland,” Merrill said. In fact, the Canadian fire season is tracking below average overall; it’s the meteorological conditions that made this week’s smoke events, as one outlet put it, “the perfect storm.”
Wildfire smoke in the region is not historically anomalous, however. A 1903 article in The New York Times describes a “yellow day” similar to smoky events in 1894, 1881, and earlier. But large-scale burns in Canada’s dense, remote boreal, which produce more smoke, are increasing. Though it’s difficult to attribute any one wildfire directly to climate change because of the complex nature of such events, we do know that fire weather is becoming more common with the warming of the atmosphere from greenhouse gas emissions. As modeled by Zeke Hausfather in the Friday edition of his newsletter The Climate Brink, “hotter, drier seasons burn the most” in Canada — and “recent years cluster there” as the country has outpaced the global average in warming.
But as Hausfather also writes, “While overall area burned is the climate-linked trend, who breathes the smoke on a given week in July is mostly driven by the weather.” This is similar to the way that, though it may be a quiet year in the Atlantic, it only takes one hurricane making landfall in the right (or wrong) spot for the season to be remembered as catastrophic.
On the other hand, as foolish as it might be for the Central Plains and East Coast to still believe smoke is the exclusive domain of Westerners, it is also a mistake to assume smoke only comes from without. As I reported earlier this year, the Eastern half of the country has seen a 10-fold jump in the frequency of large burns over the last 40 years. Nowhere is safe from the smoke.
Planning and preparation, then, should be paramount. But as Grist learned last month, there are no established Air Quality Index numbers that would trigger the postponement, relocation, or cancellation of, say, a FIFA World Cup game, including the final, which is set to be played in New Jersey on Sunday. White House officials are reportedly meeting with FIFA’s president on Friday to discuss contingencies, given the unhealthy air quality in the region.
Which brings us back to Hayden’s modeling. He offered a note of optimism in that research by Stanford’s Sam Heft-Neal and his colleagues indicates that emergency room visits do not rise in tandem with increasing wildfire smoke. “As smoke gets bad, the health impacts get bigger. But then as smoke gets worse and worse, the amount of health impacts actually goes down, measured for emergency room visits,” Hayden said. “The idea is that people modify their behavior in higher smoke” — say, by staying indoors, wearing masks, or canceling outdoor events.
It’s time to treat smoke as an East Coast phenomenon, in other words. Doing so will save lives. “Will [smoke events] become more frequent in the future? Most likely we will see a recurrence,” Merrill, the meteorologist, told me. “How often they happen is yet to be determined.”