The sleek panel entitled "Confronting the data center energy crisis" at New York Climate Week was in full swing, with tech executives painting a rosy vision of an AI-powered, sustainable future. Then, the polished presentation was interrupted when Josh Parker, head of sustainability at NVIDIA, was expressing how much NVIDIA cares about climate change. A Greenpeace activist called out to him: "If you really care, why don't you focus on decarbonisation? If you really care, you should do more!", holding a banner that read, "STOP GREENWASHING."
Interestingly, besides the Greenpeace activist, three other disruptions occurred during this 35-minute discussion, with protestors calling on this company to stop the data centre boom accelerating the climate crisis and enabling warfare. As NVIDIA positions itself as the savior of the modern tech world, and Big Tech pushes Artificial Intelligence as the ultimate solution to climate change, there is an invisible elephant in the room: the dirty reality of AI itself.
The "unstoppable" NVIDIA
NVIDIA has been dominating the news headlines lately, sitting firmly at the center of the global AI frenzy. That prominence was on full display at the ALL-IN Summit a few days ago, a major gathering of tech leaders and investors. There, CEO Jensen Huang once again championed the rapid expansion of AI and the infrastructure powering it.
What we know about NVIDIA is that it is the most dominant chip design company in the world. Beyond cornering the GPU market, it provides the software, cloud services and infrastructure that power AI development-what CEO Jensen Huang calls "AI factories." To mobilize hundreds of billions of dollars for AI expansion, NVIDIA is even acting as a kind of central bank for the industry, financing its customers and partnering with major investment firms, which it calls an investment of "ecosystem".
Despite its massive expansion, NVIDIA has managed to cultivate a relatively innocent image, appearing less harmful than Big Tech peers like Meta and Amazon Web Service (AWS). That is primarily because, as a hardware company, NVIDIA lacked a public-facing presence, and has less direct contact with retail customers. Additionally, the company has strategically played the "good guy" in the industry, heralding narratives like "AI is a climate solution" to build and maintain a positive public perception.
But this is precisely why NVIDIA's power is so dangerous, and why the company must face far greater scrutiny. With its immense market power and cash flow, NVIDIA acts as an invisible hand, exerting enormous influence over how, where and how rapidly the AI industry expands.
The dirty reality behind the AI gold rush
AI does not exist in some weightless, invisible "cloud." Behind every content generated, rows of servers and chips are running at high speed inside data centers, consuming electricity, generating heat and relying on cooling systems, some of which also use vast amounts of water.
And this is only one part of AI's environmental footprint. Long before these machines are switched on, extracting raw materials, and manufacturing chips and servers, have already generated substantial greenhouse gas emissions.
NVIDIA designs its chips but, as a "fabless" company, relies on outsourced manufacturers in East Asia: Taiwan handles over 90% of advanced chips, while South Korea produces over 60% of memory chips. The massive manufacturing has already generated substantial greenhouse gas emissions. According to a recent briefing from lead author Ketan Joshi, Greenpeace International and a consortium of civil society groups[1], NVIDIA's disclosed Scope 3 greenhouse gas emissions[2], largely from upstream manufacturing, have risen by 725% since 2020.
This US-based company's own data exposes a critical blind spot: its sustainability efforts appear to be primarily confined to its own facilities and offices. While the company boasts a 100% renewable energy target, this is estimated to cover less than 1% of its total emissions, effectively obscuring the massive environmental impact of its supply chain elsewhere.
Beyond the emissions generated by its upstream supply chain, the briefing estimates that the undisclosed downstream emissions from the use of NVIDIA's sold chips could actually be nearly twice as high as those currently estimated upstream emissions. Yet NVIDIA has failed to provide a credible answer to this other side of its climate footprint. The company still does not disclose its absolute use-phase emissions, and its board recommended that shareholders reject a proposal calling for more comprehensive downstream disclosure.
How NVIDIA is playing around with greenwashing tricks
Based on an analysis of NVIDIA's public statements and promotional materials, the same briefing finds that the company repeatedly presents AI as a climate solution, even as it helps drive the rapid expansion of energy-intensive AI infrastructure.
The briefing identifies three central narratives in NVIDIA's climate messaging.
1. "AI will solve climate change"
NVIDIA argues that AI can improve energy efficiency, optimize power systems, accelerate climate innovation and ultimately be a "net benefit" for climate.
But many of these promised benefits remain hypothetical or weakly supported. Some are also disproportionately focused on narrow applications of traditional AI, rather than the energy-intensive generative AI responsible for much of today's infrastructure growth.
By presenting potential climate benefits as likely, NVIDIA distracts attention from a more immediate reality: manufacturing and operating AI infrastructure is already consuming vast amounts of energy and generating significant emissions.
2. "Flexible data centres will help the grid"
Through initiatives such as its partnership with Emerald AI and Google, NVIDIA portrays flexible data centres as "shock absorbers" for the power grid. The idea is that data centres can temporarily reduce or shift their electricity use during periods of high demand, helping ease grid pressure and integrate more renewable energy.
However, flexible data centres can be useful, but its benefits are limited. It generally applies only during a small number of peak-stress hours and does not remove the enormous, around-the-clock electricity demand created by new data centres.
Research cited in the briefing suggests that demand response may cover less than 1% of the year in some high-growth scenarios. In certain electricity markets, it could even increase emissions if data centres shift their consumption toward hours when cheaper fossil-fuelled power is available.
In other words, making data centres slightly more flexible does not cancel out the impact of adding massive new loads to the grid.
3. "More efficient chips mean more sustainable AI"
NVIDIA frequently highlights improvements in "performance per watt," more efficient AI infrastructure and the efficiency advantages of newer systems such as Blackwell[3].
But efficiency per computation is not the same as lower overall energy consumption.
As AI becomes cheaper and more capable, companies deploy it more widely and use it more intensively. This rebound effect (often called the Jevons paradox) can outweigh the energy savings achieved by each individual efficiency improvement.
The briefing finds that despite dramatic improvements in chip efficiency, the total power capacity of NVIDIA chips sold has continued to grow. The result is that absolute electricity consumption and associated emissions are still rising.
The pattern across all three narratives is clear: NVIDIA spotlights potential climate benefits, short-term grid flexibility and relative efficiency gains, while diverting attention from the absolute growth of AI's energy demand and emissions. In doing so, climate action becomes a justification for expanding AI infrastructure, rather than a meaningful constraint on how that infrastructure is built and powered.
What meaningful climate action would actually look like
Well, there is only one real answer: direct emissions reduction at source. Anything less is greenwashing.
If NVIDIA wants to demonstrate genuine climate leadership, it cannot shirk its responsibility by claiming that it is merely a "fabless" company and that the emissions happen outside its business. We all know that is not how it should be done.
A few days ago, Jensen Huang once again made his attitude toward concerns about AI's impacts clear. NVIDIA will stand firmly on the side of expansion, leveraging every available resource to accelerate its position in the AI race. Huang appears to believe that the market will eventually solve these problems by itself.
However, it is difficult to believe that such a successful businessman could be so naive (or should we call it complacent?). His comments send a worrying signal: NVIDIA intends to remain "unstoppable," even if that means throwing more fuel on fires burning in other people's homes.
AI infrastructure is already creating serious problems around the world. We see communities resisting data centre expansion in the United States and all over the globe. In East Asia, the manufacturing of AI chips is contributing to air pollution and water stress. People living with these impacts are no longer waiting for promises about what AI might achieve someday. They are waiting for meaningful action now.
Greenpeace International demands that NVIDIA moves towards real public good by taking the initial step of investing substantially in new and additional renewable energy worldwide. We also call on the company to expand its 100% renewable energy target beyond its own corporate operations to cover its entire manufacturing supply chain by 2030, including key production regions such as Taiwan and elsewhere in East Asia. Besides, NVIDIA must disclose a comprehensive Scope 3 emissions including use of sold chips in data centres.
This is not a big ask. It is the bare minimum that one of the greatest architects – and beneficiaries – of the AI frenzy CAN and MUST do.
Avex Li is Project lead for AI supply chain campaign in Greenpeace East Asia.
Notes:
[1] The briefing, written by analyst and lead author Ketan Joshi and a consortium of civil society groups, including Greenpeace International, Beyond Fossil Fuels, Friends of the Earth, Green Screen Coalition, Stand.Earth, Kairos Fellowship, and Climate Action Against Disinformation, is the first analysis of greenwashing-style statements commonly used by NVIDIA.
[2] Scope 3 emissions (Category 1-8) relate to measured greenhouse gas impacts outside the formal boundaries of the company's activities, such as the emissions from manufacturing raw materials used in manufacturing, or the use of products sold by a company.
[3] Blackwell is a new generation of NVIDIA's AI architecture, which NVIDIA claimed that "delivers the highest performance per watt to maximize revenue and the lowest token cost to maximize profit margins."
