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:
After years of planning, the Tropical Forests Forever Facility has so far failed to take root.

In selecting a location for this year’s United Nations climate conference, host country Brazil chose symbolism over sense. Belém, the site of this year’s summit, is perched on the edge of the Amazon rainforest. The setting is meant to foreground the importance of nature in fighting climate change — despite the city’s desperately inadequate infrastructure for housing the tens of thousands of attendees the conference draws.
That mismatch of intention and resources has also played out in the meeting rooms of the gathering, known as COP30. The centerpiece of President Luiz Inácio Lula da Silva’s agenda was meant to be the Tropical Forests Forever Facility, an international finance scheme to raise at least $2 billion per year to fund forest conservation and restoration. After an inauspicious launch in which presumed supporters of the facility failed to put up any actual financing, however, it’s unclear whether the TFFF will have a chance to prove it can work.
Deforestation rates have hardly budged globally since 2021, despite more than 100 countries signing a pledge that year to halt and reverse deforestation and land degradation within the decade. The world lost more than 8 million hectares of forest to deforestation last year, causing the release of more than 4 billion metric tons of carbon dioxide into the atmosphere — nearly as much as the entire U.S. energy sector.
First proposed by the Brazilian government in Dubai at COP28, the TFFF was devised to deliver a more consistent source of funding to countries in the global south for forest conservation that would not depend on foreign aid budgets or be vulnerable to the ups and downs of the carbon market.
The plan involves setting up a fund with money borrowed from wealthier countries and private investors at low interest rates and invested in publicly traded bonds from emerging markets and developing economies that command higher interest rates. After paying back investors, the revenue generated by the spread — roughly a 3% return, if all goes to plan — would be paid out in annual lump sums to developing countries that have managed to keep deforestation at bay. Participating countries would have the right to spend the proceeds as they choose, so long as the money goes to support forests. At least 20% of the funds would also have to be set aside for indigenous peoples.
Brazil lined up substantial support for the idea ahead of this year’s launch. Six potential investor countries — France, Germany, Norway, the United Arab Emirates, the United Kingdom, and the United States — as well as five potential beneficiaries — Colombia, the Democratic Republic of Congo, Ghana, Indonesia, and Malaysia — joined a steering committee to help shape the development of the fund. The Brazilian government ultimately proposed a fundraising target of $25 billion from the sponsor countries, with the idea to attract about $100 billion from private investors, for a total of $125 billion to get the fund off the ground.
Once the fund started generating revenue, private investors would be paid out first, sponsor countries second, and forested countries last, with the $25 billion serving as insurance to the private investors should the emerging market bond issuers default on their payments. The fund itself would be managed by the World Bank, while a separate entity would govern payments made to forested countries.
While many in the international environmental community were enthusiastic about the plan — especially as a shift away from controversial carbon markets — some raised alarms.
Max Alexander Matthey, a German economics PhD student studying international finance, first saw a presentation on TFFF at COP29 and was baffled by its simplicity. “If it was that easy to make this 3% on borrowed money, why wouldn’t everyone else be doing it?” he recalled thinking at the time. After digging into the Brazilian government’s financial analysis and doing some of his own, Matthey came to believe that the fund’s proponents had underestimated the risk inherent to the investment strategy, as well as the cost of managing the $125 billion fund, he told me.
The whole reason these emerging market bonds command a higher interest rate, Matthey explained, is because they are riskier. If and when countries default on their debts, whether due to global financial shocks like pandemics or wars, or simple mismanagement, the “free money” available for forests will dry up. “These 3% are not up for grabs,” he told me. “They compensate for actual risk and defaults that will happen over time.”
The TFFF was designed to create an incentive for countries with tropical forests to invest in policies and programs to protect forests — to hire rangers to prevent illegal deforestation, to pay farmers not to raze their forests, to implement fire prevention strategies. “They have to heavily invest,” Matthey told me. “If we as the Global North say, Well, thanks for investing large shares of your budget into rainforest protection, but you won’t get any money from our side because financial markets turned the wrong way, that’s just not how you build trust.”
Matthey outlined his analysis in a Substack post in September with University of Calgary economist Aidan Hollis. They found that the JP Morgan EMBI index, which tracks emerging market sovereign bonds, has seen regular downturns of between 18 and 32 percentage points over the past two decades. In the case of the TFFF, a single 20-point loss would wipe out the $25 billion in sponsor debt “and halt rainforest flows, possibly before they even begin,” they wrote.
The energy research firm BloombergNEF seems to agree. In a report published last week outlining the state of international biodiversity finance ahead of COP30, BNEF forecast there would be “little progress” on the TFFF. “The 3% spread is not a money faucet, but a risk premium; studies on the TFFF appear not to have properly conducted risk analyses,” the report said, warning that in effect, the scheme would eat up development finance just to absorb private investor losses. (After this article’s publication, a representative from the World Wildlife Fund reached out to Heatmap to say that this analysis was based on earlier financial modeling, and that a new model had been released in October.)
Just prior to that report’s release, confidence in the TFFF appeared to dip. Brazil’s finance minister lowered his fundraising ambition for the facility to $10 billion by 2026. A few days later, on the eve of the launch, Bloomberg News reported that the United Kingdom would not be contributing to the fund after the country’s treasury department warned it could not afford the investment, despite its significant involvement in the fund’s design.
Following the launch, Indonesia and Portugal each committed $1 billion, while Norway pledged $3 billion, although only if the fund successfully secures at least $10 billion. France also promised €500 million, or just over half a billion dollars, while Germany said it would contribute “significantly,” although it hasn’t said how much yet. All in all, countries committed just $5.5 billion above Brazil’s own initial $1 billion commitment — with at least $3 billion of that contingent on further fundraising.
Andrew Deutz, the managing director for global policy and partnerships at the World Wildlife Fund, which has also been heavily involved in developing the TFFF, assured me this was not the disappointment it appeared to be.
"I look at what just happened last week as validation that the model can work and that countries have confidence in it,” Deutz said. He pointed to the fact that 53 countries, including 19 potential investors, have endorsed the scheme. “A bunch of sponsor countries who haven’t been that engaged said, We like this idea, and I think that creates the opportunity and the momentum that we can get one or two more rounds of capitalization at least.” Deutz was bullish that Germany would come to the table with a pledge between $1 billion and $3 billion, and that the UK would “get guilted in” shortly. He expects to see additional pledges at the World Bank’s Spring Meetings next April, and a few more at the UN General Assembly next September.
As for criticisms of the fund’s investment strategy, he brushed them off, arguing that the risk was "quantifiable and manageable.” He has faith in the TFFF’s modeling showing that the fund’s managers will be able to earn high enough returns to pay back investors and still generate enough funds to pay tropical forest countries.
Charles Barber, the director of natural resources governance and policy at the World Resources Institute was more cautious on both fronts. “We’re glad it’s got as far as it has, but there’s a whole lot of questions that will need to be answered to really get it up,” he told me. Barber saw arguments both for and against the risky investment strategy, but he was skeptical that a starting point of $10 billion would be enough to attract sufficient private investment or give tropical forest countries enough of an incentive to participate.
Matthey has called the idea of a scaled-down TFFF a “worst-case scenario for everyone involved,” due to the high fixed costs of managing the fund, monitoring deforestation, administering the proceeds, etc. The potential payouts to forested countries would be so diminished as to amount to a “rounding error” rather than a true incentive, he wrote.
Deutz told me the TFFF’s architects always expected there to be a three- to four-year ramp-up period. If the fund gets one or two more rounds of capitalization, “we’ll see if it works — and then, assuming it works, you can keep adding to it,” he said. “This is something new and different, so it might take us a little while to prove it out and for people to get comfortable.”
Editor’s note: This story has been updated to reflect new information provided after publication.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
The spinoff of Lawrence Livermore National Lab has a new 10-point plan to get onto the grid by the 2030s.
One of fusion energy’s newest startups, Inertia Enterprises, is betting that the fastest route to commercial fusion runs through one of the field’s oldest ideas. The company, which raised a $450 million Series A earlier this year, plans to build a power plant based on the laser-driven fusion system pioneered at Lawrence Livermore National Laboratory’s — the only tech yet to have produced more energy from a fusion reaction than it took to initiate it. Now, Inertia has shared its commercialization roadmap exclusively with Heatmap, detailing the 10 near-term capabilities it must demonstrate before this landmark experiment can become a grid-scale power plant by the mid-2030s.
The roadmap offers a route from the national lab’s impressive but commercially impractical fusion demonstrations to an economical power plant capable of producing electricity for the grid. At its core are a set of milestones — mostly aimed at developing cheap, mass-manufacturable components — that Inertia says it must clear before those individual systems can be integrated into a working plant. This road is not necessarily linear, however, as various teams will likely be working on many of these goals simultaneously.
At least the physics of Inertia’s approach are already proven, the startup’s CEO Jeff Lawson told me, pointing to the fusion experiments at Lawrence Livermore’s National Ignition Facility as a proof-of-concept. The lab’s demonstration of net energy gain caps more than six decades and $30 billion (in 2026 dollars) of U.S. fusion research. The remaining challenges, he argued, are all engineering-related, requiring “elbow grease, hard work, and smart people” rather than breakthroughs in fusion science.
"It seems to us like a startup or a commercial company of any variety should be focused on commercializing a proven scientific result, as opposed to actually trying to demonstrate the basic science to begin with," Lawson told me. Basic science, he argues, is better left to national labs and universities, where researchers can pursue "unbounded problems" that don’t align with the expectations and timelines of venture-backed startups.
Indeed, no fusion startup has yet achieved scientific breakeven, the milestone Lawrence Livermore first hit in 2022, and has since repeated numerous times. But leading players such as Commonwealth Fusion Systems and Helion Energy maintain that it’s only a matter of time before they validate the physics behind their own reactor designs, which they claim will be highly cost-competitive.
Lawson, on the other hand, readily acknowledged that Lawrence Livermore’s tech is uneconomical in its current form. His bet is simply that the more predictable path to a commercial reactor is to drive down the cost of the lab’s validated fusion approach, known as inertial confinement. This system relies on high-powered lasers firing at a millimeter-scale pellet of fusion fuel, compressing it to extreme temperatures and pressures until the atoms fuse. Today, the National Ignition Facility makes each individual fusion target by hand, a workable solution given that it only uses about a dozen per year.
That production model, however, isn’t remotely plausible for a grid-scale power plant. Because each fusion reaction lasts just a fraction of a billionth of a second, a commercial facility must fire its lasers at a fresh target about 10 times per second to generate continuous electricity — requiring the production of hundreds of millions of targets each year.
Scaling production to roughly a million pellets per day and making them inexpensive enough for commercial operation without compromising the strength or precision required for fusion ignition is central to Inertia’s roadmap. That includes goals five, seven, eight and nine — industrializing the manufacturing of the carbon shells that hold the fusion fuel, making the thin films that hold those carbon shells both durable and cheap, scaling up and automating fusion target assembly, and speeding up how fast targets are filled with the requisite deuterium-tritium fuel.
The other central focus of the roadmap is the laser system, which will ultimately consist of 1,000 individual units operating in concert to compress and heat the fusion fuel. Key priorities include reducing the system’s cost (goal two), dramatically increasing its firing cadence (goal three), and bolstering its durability to withstand high-intensity operations (goal four). Goal six also complements these efforts, calling for the development of a control system capable of tracking moving fusion targets to precisely align each laser shot.
Goals one and 10 bookend the journey with some broader milestones. The first focuses on increasing the fusion target’s energy gain — the ratio of fusion energy produced to laser energy delivered — to more than 25 times ignition. Today, the National Ignition Facility’s best-performing laser shot has yielded a gain of just over four times what it took to start the reaction. Goal 10 then zooms out to the ultimate objective: integrating all these technologies into a commercially viable power plant that can deliver either electricity or industrial heat to end customers.
To reach that point, Inertia has embarked on an industrial engineering hiring spree, recruiting folks with experience taking complex hardware systems from prototype to mass production, “not unlike the processes that are used in the semiconductor or consumer electronics world,” Lawson explained. The company has been making progress on its component development goals since the beginning of the year, he told me, and expects to announce the successful demonstration of a few of these milestones in the coming months. Lawson ultimately expects Inertia to complete the core components of its laser and target manufacturing systems by the middle of next year.
The team will spend the next two to three years integrating these individual pieces into two fully operational subsystems, a prototype laser system and a target manufacturing line. Around 2030, the company will begin combining those subsystems into a first-of-a-kind fusion power plant, which will also serve as the proving ground for the target chamber, tritium fuel breeding system, and power conversion system that turns fusion heat into electricity. By the middle of the next decade, Inertia aims to be generating power from this first plant, setting the stage for the company to build and connect additional grid-scale commercial power plants.
There are plenty of engineering trade-offs that the company will have to solve for. Take the decision around how to size the target chamber, for example. “If you make it bigger, your walls have an easier time and survive longer, but it’s more expensive. If you make it smaller, your walls have a tougher time because they’re closer to all the heat and energy that the fusion reaction is creating, but now your power plant costs less to build.”
But to Lawson, this represents exactly the type of problem Inertia was built to solve: complex engineering issues that come to the fore once scientists have demonstrated the fundamental physics are sound. He thinks other fusion companies may someday reach this stage, as well — though he’s unwilling to hazard a guess on exactly what approach or startup is best positioned to do so.
“There have been generations of scientists who’ve made their predictions about fusion energy and gotten it wrong,” he told me. “I’m not going to pretend to be smarter than them. All I’m here to say is, just knowing that one did work, we can commercialize it.”
Current conditions: After forming into Tropical Storm Bertha late Monday, the system is barreling toward the Florida Panhandle as it makes landfall as far west as Texas • In the Pacific, Hurricane Fausto has strength as it heads toward Hawaii but remains a Category 1 storm • Temperatures in Ouargla, Algeria’s southern city in the Sahara desert, are soaring to nearly 120 degrees Fahrenheit this week.
Emissions from the United States’ electrical sector spiked 4% last year as demand for power drove up generation from coal. That’s according to the latest annual assessment published Tuesday morning by the U.S. Energy Information Administration. The report, which has tracked annual emissions data from all power sources since 2010, found that U.S. energy-related carbon dioxide emissions increased by 2%, or about 115 million metric tons, in 2025. But the power sector specifically saw a surge of 4%, or 58 million metric tons, due to a spike in fossil fuel use. Coal-fired generation rose by 13%, even as natural gas-fired power fell 4%. Renewables helped avoid more coal use. While wind generation increased 3%, solar skyrocketed by 34%. Generation from all other sources — including nuclear and the category of “other renewables” that includes hydropower and geothermal — were essentially flat last year.
The coal surge isn’t unique to the U.S., as my colleague Matthew Zeitlin wrote last year. Worldwide, rising demand for electricity and shrinking supply of natural gas coming through the Strait of Hormuz made for a good year for coal.
Watershed, the software platform focused on corporate sustainability, just published what it called its first comprehensive open framework for estimating the greenhouse gas emissions from companies’ use of AI programs. The framework has three elements: A comprehensive system that includes all phases of a data center’s use, from model training to inference to hardware production; a function unit of kilograms of carbon dioxide equivalent per million tokens; and a three-tier calculation approach “that aligns with companies’ data quality.”
In a statement to my colleague Emily Pontecorvo, Watershed’s science chief John Bistline said he had “heard from companies that they’re already being asked about AI emissions from investors, from auditors, from regulators, and right now most of them are guessing. We wanted to give them something that was more defensible.”
Oil prices spiked again Tuesday after President Donald Trump publicly weighed taking “a nice big fat shot” at Iran’s Pickaxe Mountain, where Israeli intelligence suggests the Islamic Republic moved its uranium-enriching centrifuges last fall. Brent crude, the main European benchmark for the price per barrel of oil, rose nearly 3% to over $91. West Texas Intermediate, the U.S. price signal, saw a 3% hike to just nearly $85. Murban crude — out of the United Arab Emirates, therefore the most sensitive to Persian Gulf disruptions — soared nearly 5% to just under $86 per barrel.
Shakeups among smaller producers, meanwhile, appeared to cancel out each other’s effects on the market. The shot: Kazakhstan, which falls just outside the top 10 oil-producing nations, is halting crude shipments to the Russia ports it relied on to get its hydrocarbons to market now that Ukraine is consistently attacking the Kremlin’s energy infrastructure, according to the Financial Times. The chaser: Norway’s oil output just beat forecasts, per Oil Price.
Sign up to receive Heatmap AM in your inbox every morning:

Unlike the last man Trump put in charge of the Environmental Protection Agency during his first term in office, Lee Zeldin hadn’t formally worked for the coal industry before serving in government. But the EPA administrator sure made it sound like the industry’s executives are high-priority constituents. At a National Coal Council event in Washington, D.C.’s Willard Hotel that E&E News covered, Zeldin said “many of the items that were on your wish list are now done.” In the coming months, he added, the agency would get to “the remainder of those items,” but said he wouldn’t “prejudge” any rulemaking outcomes. “Between now and your next meeting, I’m excited to be able to share with great optimism, hope, and enthusiasm that you all, again, not prejudging the outcome of any rulemaking, we’ll have a lot to celebrate the next time you all get together again in January,” Zeldin said. One thing the EPA can’t do: Keep the coal plants the Trump administration wants open actually running. As Matthew wrote last year, the big problem with aging coal stations is that they keep breaking down.
Mergers and acquisitions within the global nuclear industry totaled more than $7 billion in value in the first half of 2026, doubling that same figure from a year earlier. That’s according to new data the law firm White & Case LLP shared Tuesday with World Nuclear News. The number of individual deals increased 10%, from 40 to 44. “At the current pace of dealmaking activity, 2026 is set to surpass all years aside from 2024 when a record $29 billion of M&A activity was registered,” the law firm said. More proof that the nuclear dealmaking boom, as Heatmap’s Katie Brigham wrote last year, “is real.”
It’s not just automobiles going hybrid-electric. The startup Electra, which has promised to build a nine-passenger hybrid-electric plane that can take off in as little as 150 feet, is now pumping $850 million into its first aircraft factory in Ohio. The plant, announced Tuesday, will build up to 800 aircraft per year at full capacity. But as Electrek put it, “that’s a big commitment for a plane that hasn’t flown yet.”
Frontier model developers still keep their energy use largely a secret, but Watershed is proposing a new formula that will at least get you close.
With companies now rapidly adding artificial intelligence into their products and using it across their workstreams, it stands to reason that all that extra energy use might show up in their climate accounting. But to any business that wants to get a sense of how big its AI-related emissions footprint is becoming — and, god forbid, maybe even try to reduce it — I say well, good luck. AI providers mostly keep the data required to make such calculations a secret.
Now Watershed, a startup that helps companies track and estimate their carbon emissions, is proposing a workaround. The firm published a white paper on Wednesday laying out a method for companies to produce rough estimates of their carbon impact from AI, while also encouraging them to demand better data from AI developers.
“We’ve heard from companies that they’re already being asked about AI emissions from investors, from auditors, from regulators, and right now most of them are guessing,” John Bistline, Watershed’s head of science, told me. “We wanted to give them something that was more defensible.”
For most frontier AI models, including those developed by OpenAI and Anthropic, there’s very little information to work with. Google is the only proprietary AI developer that has published a transparent estimate of its model’s operational energy use and related emissions. In a paper last August, researchers at the company found that “the median Gemini apps text prompt consumes 0.24 watts,” which is “less energy than watching nine seconds of television,” and released 0.03 grams of CO2-equivalent. These numbers may be out of date by now, however. In the paper, the authors note that this already represented a 33-fold reduction in energy consumption compared to the previous year.
That’s one challenge with estimating AI-related emissions — tech companies are both growing and innovating rapidly, expanding their energy footprints while also finding greater efficiencies, which may be one reason they don’t disclose this information yet.
Another obstacle is that the exercise involves making a number of carbon accounting decisions, and there’s no consensus yet on best practices. For instance, where do you draw the line on which emissions to include? You could just look at the energy required to operate the model, or you could include the energy used to train the model, or even the emissions related to fabricating and manufacturing the hardware it’s running on. Training a model tends to be more energy-intensive than running it to respond to queries, but it only happens once. If you’re going to include training emissions, the next question is, how should responsibility for those be allotted across the lifetime of the model and its use by hundreds of thousands of customers?
Another decision is how to account for differences in user behavior. A model’s energy intensity can vary widely depending on whether the user is asking a simple question, requesting complex research, generating images, or dispatching agents to conduct multiple tasks simultaneously. Models capable of “reasoning” use an estimated 30 times more electricity than those without that ability, according to research by HuggingFace, a company that creates tools for AI developers. A per-prompt emissions average would not capture these differences, and therefore would not give companies actionable information to help them reduce their emissions.
A “per token” average might be more useful in that sense. When AI models process queries, they break the sentence or code down into smaller components called tokens. One token might be just the first few letters of a word. When the model generates a response, it also processes it in terms of tokens. Estimating emissions per token is not a perfect system either, however, since a token’s value can vary across AI providers. Input tokens, i.e. user questions, also tend to be less energy-intensive than output tokens, or user responses, and a single per-token average will conceal that difference.
Then there’s the question of how to get from a model’s energy intensity to an emissions estimate. Should you use the real-world average carbon intensity of the electric grid? What about any clean energy agreements the AI company may have signed? And how should you factor in companies that decide to bypass the grid entirely and build their own on-site generation, which tends to use natural gas?
The Watershed paper proposes some answers to these questions, and also offers guidance for how companies can develop emissions estimates based on the data available to them.
While most of the published research on AI emissions to date has calculated energy intensity on a per-query basis, Watershed advocates for a per-token approach. — i.e. “kilowatt-hours per thousand tokens.” The authors reason that electricity use scales more directly with the number of tokens used than the number of queries submitted. AI application customers are also often billed based on their token usage, so there’s a business case for tracking tokens and trying to use them more efficiently.
For those companies working with essentially zero data — not even the number of tokens they’re using per year — Watershed recommends they approximate their AI emissions using a “spend-based” method. This means simply multiplying the amount they spend per year on AI services by an emissions factor of 0.134 kilograms of carbon dioxide equivalent per U.S. dollar, which is based on U.S. Bureau of Economic Analysis numbers for the data processing sector of the economy.
The Watershed paper concedes that whatever number this method spits out will be wrong, noting that it “can misestimate true AI emissions by several times in either direction,” and advising companies to treat this as a “provisional placeholder.” But publishing these numbers, even though they are wrong, could help push AI companies toward more transparency if they want to correct the record.
For companies that do track their token volumes, Watershed has a more rigorous solution. The paper proposes a formula companies can use to calculate their AI emissions, accounting not just for inference energy use, but also training emissions, embodied emissions of the data processing equipment, and a figure known as “power usage effectiveness.” This captures the energy consumed by cooling systems, power conversion, and other data center infrastructure that’s not directly serving AI processing. Since model-specific values for the various inputs to the formula are mostly not available today, Watershed has provided default values gathered from previous studies, including papers by Microsoft and Google. Companies can substitute the actual numbers disclosed by AI providers into the formula as that information becomes available.
I reached out to Google, Microsoft, Anthropic, and OpenAI to ask why they didn’t share token carbon intensity, and whether they planned to in the future. Only Microsoft responded to my inquiry, pointing me to its blog post and peer-reviewed paper estimating general AI energy use across frontier models.
To get the most accurate estimate, companies would also need to know where, geographically, their AI queries are being serviced, since emissions from the electric grid varies by region. In some cases, companies may be able to actually choose where their queries are being processed, offering another lever by which they could potentially reduce their emissions.
The right data, disclosed in sufficient detail, will unlock companies’ ability to reduce their AI-related emissions, Watershed argues. Employees would have more reason to choose the most appropriate model for a given task, for example, like avoiding using energy-intensive reasoning models for basic questions.
“I think about a John von Neumann test here,” Bistline said, referring to the mathematician and proto-computer scientist. “You wouldn’t ask an advanced model like Fable anything that you would be embarrassed to ask John von Neumann, or Marie Curie, right? You wouldn’t want to ask ‘how many R’s are there in Strawberry?’ or ‘which restaurants would you recommend I go to in Miami?’”
Of course, companies can already implement this recommendation today, but there will be no way to account for and prove that they are reducing their emissions as a result until AI providers disclose distinct model-based energy estimates.
As Bistline mentioned, this information isn’t just nice-to know — companies are already being asked for it. Upcoming regulations in California and the European Union will require large companies to disclose their total direct emissions, and will eventually require them to disclose indirect emissions like AI energy use. The EU’s AI Act will also require AI companies to disclose a breakdown of the energy consumption of its general purpose AI models.
“There are customer-side disclosure rules and provider-side ones developing in parallel,” Bistline said, “and right now there’s no agreed methodology connecting the two, which is the gap we’re trying to address with our AI emissions framework.”