Green AI Was a Promise. The Emissions Data Says Otherwise.

The announcements came with the confident language of corporate inevitability. Microsoft pledged to be carbon negative by 2030. Google committed to operating on 24/7 carbon-free energy by the same year. Meta promised net-zero emissions across its entire value chain by 2030. These were not vague aspirations buried in footnotes — they were headline commitments, celebrated by sustainability advocates, cited in ESG filings, and woven into the brand identity of the world’s most powerful technology companies.

Then came the data.

The Emissions Gap No One Wants to Talk About

Since each of these companies made their landmark climate pledges, their actual greenhouse gas emissions have moved in precisely the wrong direction. Microsoft’s emissions have climbed approximately 30% above its baseline. Google’s are up roughly 50%. Meta’s have surged by nearly 70%. These are not rounding errors or accounting anomalies. They are the measurable, reported consequence of a single dominant force: the rapid, resource-intensive buildout of artificial intelligence infrastructure.

Data centers consume enormous quantities of electricity. Training a single large language model can emit as much carbon as five average American cars over their entire lifetimes. Inference — running the model to answer queries, generate images, write code — compounds that demand at scale, continuously, around the clock. The more people use AI, the more energy the servers require. And as of early 2026, hundreds of millions of people use AI every day.

The uncomfortable truth is that the emissions trajectories of Big Tech’s most climate-ambitious companies are now being driven, in material part, by the same product lines those companies are actively promoting as tools for solving the climate crisis.

Stargate and the $75 Billion Acceleration

If the existing emissions gap was concerning, the investment pipeline makes it structural. The Stargate project — the $500 billion AI infrastructure initiative backed by OpenAI, SoftBank, and Oracle, with Microsoft as a key partner — represents the largest single commitment to AI buildout in history. Google, not to be outdone, has announced over $75 billion in capital expenditure on data center infrastructure in 2025 alone.

The critical detail that rarely makes the press release: a significant portion of the energy powering these facilities, particularly the new capacity being brought online rapidly to meet AI demand, is sourced from fossil fuels. Renewable energy procurement takes time. Grid interconnection queues stretch years. Utility-scale solar and wind projects face permitting, siting, and transmission bottlenecks that no corporate sustainability commitment can dissolve by executive memo.

So when Microsoft signs a power purchase agreement for wind energy to be delivered in 2028, its new data center in Virginia or Texas may be drawing on natural gas today. The accounting can look clean; the atmosphere sees the difference.

The AI-Saves-the-Climate Narrative, Examined

In parallel with the infrastructure expansion, the industry has developed a counter-narrative: yes, AI uses energy, but AI will optimize the energy system. It will accelerate drug discovery, improve agricultural yields, reduce industrial waste, design better batteries. The net effect, the argument goes, will be deeply positive for the climate.

This argument received a significant challenge at the 2026 AI Impact Summit, where researchers and independent climate scientists examined the available evidence. Their finding was unambiguous: there is currently no verifiable evidence that generative AI is producing measurable reductions in global emissions. The efficiency gains AI enables in some sectors are real but remain diffuse, slow to scale, and — critically — are not yet offsetting the direct emissions from AI’s own infrastructure demands.

More troubling is the function the narrative may be serving. When a technology company points to AI’s hypothetical future climate benefits, it creates a permission structure for present-day emissions growth. It reframes an accountability problem as an investment thesis. The pledges don’t disappear; they get deferred — absorbed into a story about the long game that conveniently never arrives at a moment of reckoning.

This is not to say researchers working on AI-assisted climate modeling or grid optimization are operating in bad faith. Many are doing vital work. But the industry-level deployment of that work as a rhetorical shield against emissions scrutiny deserves exactly the skepticism it has so far largely avoided.

What a Credible Path Forward Actually Requires

The question worth asking is not whether AI and climate goals can coexist — they can — but whether the current trajectory reflects a genuine effort to make them do so. Based on the emissions record, the answer today is no. What would a credible path look like?

Researchers at Cornell and energy consultancy Schneider Electric have outlined the key requirements:

• Grid decarbonization timelines must precede, not follow, data center expansion. Building capacity before clean power is available is a choice, not a necessity.
• Additionality requirements for renewable energy procurement — meaning companies must fund new clean generation, not simply claim credit for existing renewables — would close the accounting loopholes that allow emissions to rise while sustainability scorecards stay green.
• Mandatory, standardized Scope 3 emissions reporting for AI infrastructure would make the true climate cost of model training and inference visible and comparable.
• Regulatory intervention is likely unavoidable. Voluntary commitments, the emissions data now clearly shows, are insufficient. The EU’s AI Act and emerging U.S. data center energy disclosure rules are early steps, but enforcement mechanisms remain weak.

The companies that made net-zero pledges did so knowing they would be held to them. The emissions data is the holding. Microsoft, Google, and Meta are not bad actors in a simple sense — they are companies optimizing for growth inside a system that has not yet priced carbon honestly or enforced accountability consistently.

But the gap between the pledge and the trajectory is no longer a minor discrepancy. It is the defining story of Big Tech’s relationship with the climate — and it deserves to be treated as such.