Unsubstantiated Hype: The Troubling Lack of Evidence Behind Big Tech’s AI Climate Claims

An investigation led by energy researcher Ketan Joshi has cast a critical light on the ambitious claims made by major technology companies, particularly Google, regarding artificial intelligence’s potential to combat climate change. Joshi’s recent report, supported by several environmental organizations, reveals that a significant portion of these high-profile assertions are backed by minimal, if any, verifiable evidence, raising concerns about potential "greenwashing" at a critical juncture for global climate action.

The genesis of Joshi’s inquiry dates back a few years when he encountered a striking statistic concerning artificial intelligence and climate change. In late 2023, Google began widely circulating a claim that AI could contribute to cutting global greenhouse gas emissions by an impressive five to 10 percent by 2030. This bold prediction was initially disseminated through an op-ed co-authored by Google’s chief sustainability officer and subsequently gained traction, being quoted extensively across various media outlets, including Forbes and S&P Global, and even appearing in some academic papers.

Joshi, a seasoned energy researcher, found himself taken aback by the sheer magnitude of the numbers Google was promoting. A five to 10 percent reduction in global emissions by 2030 would be equivalent to offsetting the entire annual emissions of the European Union—a truly monumental achievement. "I found [the emissions claim] really compelling because there’s very few things that can do that," Joshi remarked, highlighting the extraordinary nature of such a potential impact. This curiosity propelled him to delve deeper, determined to trace the origin of this powerful statistic.

His investigation revealed a chain of citations leading to a surprisingly weak foundation. The five to 10 percent figure, Joshi discovered, originated from a paper jointly published by Google and BCG, a prominent global consulting group. This paper, in turn, referenced a 2021 analysis conducted solely by BCG. Crucially, the BCG analysis offered a remarkably thin basis for its estimate of massive emissions reductions from AI, citing merely its "experience with clients." Joshi critically described this source as "flimsy," especially given the gravity of the claim. It’s also noteworthy that this analysis was published a full year before the public introduction of ChatGPT in late 2022, an event that ignited the current intense race among tech companies to build out the energy-intensive infrastructure now deemed essential to power the burgeoning AI revolution.

The narrative surrounding AI’s climate benefits grew even more complex when, just a few months after initially endorsing the five to 10 percent estimate, Google quietly acknowledged a contradictory reality. In its 2023 sustainability report, the tech giant admitted that the aggressive buildout of AI infrastructure was, in fact, significantly driving up its own corporate emissions. Despite this internal admission, Google has continued to champion the reduction figures provided by BCG. Most recently, the company reiterated these figures last year in a memo addressed to European policymakers, suggesting AI’s vital role in achieving climate goals.

Joshi expressed his astonishment at this continued promotion. "One of the most powerful tech companies in the world using this metric to make policy recommendations to one of the biggest regions in the world—I thought that was remarkable," he stated. "That instance was what got me immediately very interested in the structure of this claim and the evidence behind it." The use of such a metric in high-level policy discussions underscores the importance of scrutinizing its underlying data.

When approached for comment, Google spokesperson Mara Harris defended the company’s position, telling WIRED in an email, "We stand by our methodology, which is grounded in the best available science. And we’re transparent in sharing the principles and methodology that guide it." Harris provided a link to Google’s general methodology for calculating emissions reductions from its products and partnerships but did not offer specific details on how these standards were applied to validate the particular numbers supplied by BCG. BCG itself did not respond to WIRED’s inquiries, leaving a significant gap in the transparency surrounding the original data.

The current landscape sees tech companies deeply entrenched in a fierce competition to develop AI as rapidly as possible, a race with profound and potentially far-reaching implications for global climate change. In the United States, which boasts the world’s largest data center market, the escalating energy demands driven by this buildout are having tangible effects. Reports indicate that these demands have led to coal plants remaining operational longer than planned and are necessitating the addition of hundreds of gigawatts of new gas power to the national grid. Nearly 100 gigawatts of this projected new power capacity are specifically earmarked to fuel data centers, highlighting the immense energy appetite of the expanding AI sector.

Tech executives frequently justify this substantial energy and data center expansion by emphasizing the transformative possibilities AI presents for the planet. At last year’s annual Climate Week event in New York City, the Bezos Earth Fund, Jeff Bezos’s sustainability-focused nonprofit, hosted a series of discussions centered on the theme of "AI will be an environmental force for good." Echoing similar sentiments, former Google CEO Eric Schmidt declared in late 2024 that since the world is unlikely to achieve its climate goals, the focus should shift to what AI can accomplish. "I’d rather bet on AI solving the problem, than constraining it and having the problem," he famously stated. OpenAI’s CEO, Sam Altman, has gone further, boldly promising that AI will "fix" the climate altogether.

However, as Joshi’s investigation and new report underscore, many of these sweeping claims often lack robust empirical backing. The report, released on Monday, meticulously examines over 100 high-profile assertions made by tech companies, energy associations, and other entities about AI serving as a "net climate benefit." Joshi’s comprehensive analysis revealed that a mere quarter of these claims were supported by academic research. Disturbingly, more than a third of the examined claims failed to cite any public evidence whatsoever, leaving their credibility severely undermined.

Jon Koomey, an independent energy and technology researcher not involved in Joshi’s report, offered a cautionary perspective. "People make assertions about the kind of societal impacts of AI and the effects on the energy system—those assertions often lack rigor," Koomey observed. "It’s important not to take self-interested claims at face value. Some of those claims may be true, but you have to be very careful. I think there’s a lot of people who make these statements without much support." His comments reinforce the need for critical evaluation of industry-driven narratives.

Another critical aspect explored in Joshi’s report is the ambiguity surrounding the kind of AI being discussed when tech companies champion its climate-saving potential. Many forms of AI are considerably less energy-intensive than the generative, consumer-focused models that have dominated recent headlines—tools like ChatGPT, Claude, and Google Gemini—which demand massive computational power and energy to train and operate. Machine learning, for example, has been a foundational tool in numerous scientific disciplines for decades, often with relatively modest energy footprints. However, Joshi’s analysis found that nearly all the claims he examined conflated these more traditional, less energy-intensive forms of AI with the highly energy-intensive, consumer-focused generative AI that is currently driving the rapid expansion of data centers globally.

David Rolnick, an assistant professor of computer science at McGill University and chair of Climate Change AI, a nonprofit advocating for machine learning to address climate issues, shares some of Joshi’s concerns, albeit from a slightly different angle. Rolnick acknowledges the difficulty of quantitatively proving impact in this field and is perhaps less focused on the precise provenance of Big Tech’s numbers. For him, the distinction between different types of AI is paramount. "My problem with claims being made by big tech companies around AI and climate change is not that they’re not fully quantified, but that they’re relying on hypothetical AI that does not exist now, in some cases," he explained. "I think the amount of speculation on what might happen in the future with generative AI is grotesque."

Rolnick points out that existing deep learning techniques are already being effectively deployed in various sectors worldwide to reduce emissions and combat climate change right now. These applications range from optimizing grid efficiency to developing models that aid in the discovery of new species. "That’s different, however, from ‘At some point in the future, this might be useful,’" he clarified. Furthermore, Rolnick identifies a significant "mismatch between the technology that is being worked on by big tech companies and the technologies that are actually powering the benefits that they claim to espouse." Companies might showcase examples of algorithms that improve flood detection, for instance, using them to advertise the climate benefits of their large language models—even though the flood prediction algorithms are fundamentally different types of AI than consumer-facing chatbots.

Sasha Luccioni, another prominent AI and sustainability researcher, echoed these sentiments. "The narrative that we need big AI models—and quasi-infinite amounts of energy—tries to sell us the idea that this is the only kind of AI we need, and the only future that’s possible," she contended. "But there are so many different, smaller and more efficient models that can be deployed for a fraction of the cost, both to people and the planet."

In a separate piece of research also released on Monday, Luccioni, in collaboration with Yacine Jernite, head of sustainability at AI company Hugging Face, investigated the costs associated with training a wide array of AI models. Their findings challenged the notion that massive, proprietary models, which demand access to vast amounts of data and energy, are the only viable option for powerful AI solutions. Often, smaller, more efficient models demonstrated comparable performance to their more expensive counterparts in various AI applications.

Luccioni critically analyzed the implications of this "bigger-is-better" AI race. "The only companies that can compete in this bigger-is-better AI race are the ones with the deepest pockets, who have hoovered up our data—consensually or not—over the last decades, and continue to do so," she argued. "Now they are selling this data back to us by convincing us that we need these mammoth models, the planet be damned." This perspective highlights a potential economic and environmental agenda driving the promotion of large-scale AI.

A fundamental obstacle to accurately assessing AI’s true impact on climate, experts consistently tell WIRED, is the pervasive lack of basic, publicly available information. Without crucial data, it remains challenging to fully comprehend the capabilities and environmental footprint of AI. Currently, we largely rely on rough estimates for how much energy AI—let alone the rapidly expanding generative AI sector—consumes within data centers. While Google did release some estimates of its AI prompts’ energy usage last year, many other companies continue to lag behind or refrain from disclosing key environmental data about their models. Moreover, despite the widespread integration of generative AI into consumer experiences, concrete examples demonstrating how large-scale generative AI can more effectively tackle climate issues than less energy-intensive models are still largely absent.

For Ketan Joshi, the path forward is clear and straightforward: companies driving the accelerated development of AI must be compelled to disclose more comprehensive information about the climate costs associated with their innovations. "If [tech companies] are worried that people are overstating or exaggerating the climate impacts of generative AI, then there should be nothing stopping them from saying, ‘Well, okay, our energy growth this year was six terawatt-hours, and two of them were for generative AI,’" he asserted. "That’s information that we push for more disclosure of in the report. I think that would ultimately be a very good thing for them," suggesting that greater transparency would not only benefit public understanding but also potentially enhance the tech industry’s credibility.