Hyperscaler Emissions on Track to Jump 60 Percent in 2026

Five hyperscalers will emit a combined 215 million metric tons of CO2 equivalent in 2026 — a 57 percent jump in a single year, according to the first output from a new capex-based forecasting model built by startup Bricola Group.

The forecast covers Alphabet, Amazon Web Services, Meta, Microsoft and Oracle. That combined footprint exceeds the national emissions of a large majority of countries, Bricola says. And the driver is unambiguous: new data-center construction, equipment and power, funded by capital expenditure that Bricola expects to top $1 trillion across the five companies in 2025 and 2026 combined.

AWS and Alphabet carry the largest share. Bricola puts both at just over 53 million tCO2e each for 2026.

The takeaway: the AI buildout is now showing up in corporate carbon accounts at a scale that rivals nation-states.

What does the capex model actually measure?

Bricola's founder, Jimmy Gillan, built the forecast by comparing each company's historical capital expenditure against its historical emissions. The logic is simple. Companies report capex in their annual financial results — the 10-K filings — months before they publish sustainability reports. But most of the change in a hyperscaler's future footprint comes from data-center spending.

So Gillan's model estimates emissions as soon as capex figures or guidance land, instead of waiting for the official carbon disclosure. Each company's forecast comes as a range, set partly by what financial data is available. Oracle and Microsoft, whose fiscal years end in May and June respectively, already have confirmed capex numbers in the model. Alphabet, AWS and Meta, with December fiscal year-ends, run on recent guidance.

The gap it addresses is real. Hyperscaler sustainability reports typically arrive five to nine months after the 10-Ks. A full year's emissions picture can land close to a year after the money was spent.

How big is the jump after carbon instruments?

Gross emissions aren't the whole ledger. After deducting renewable energy certificates and other instruments the companies purchase, the net total falls to 140 million tCO2e for 2026 — still up sharply from 87 million tCO2 in calendar year 2025.

That net figure rising by roughly 60 percent in one year matters for anyone tracking these companies against their climate targets. Bricola calls it the largest increase since the race to build AI products first triggered an emissions surge — a surge that has already upended progress toward climate goals at Microsoft, Alphabet and others in the group.

In plain terms: buying clean-energy paper isn't keeping pace with concrete, servers and electricity demand.

Why does faster forecasting matter?

Gillan, a former Amazon economist, built the model for sustainability teams, industry groups and anyone who needs to act on emissions data before it's officially published.

"My hope is to accelerate what we think is going to come out of the disclosure process so that we can identify the areas where decarbonization is most needed," Gillan said. The point, he added, is that resources can be mobilized quickly and teams are "not waiting for a number to come in."

For sustainability directors at these firms and their suppliers, that shifts the timeline. Instead of reacting to a footprint report that lands mid-year, they can work from a capex-based estimate within weeks of financial results.

The next test is whether the model's 2026 numbers hold when the official sustainability reports arrive — disclosures that will land months from now and settle whether 215 million tCO2e was the ceiling or the floor.