Google and Amazon's net-zero climate goals are slipping out of reach because of their massive investments in artificial intelligence. Both companies released their annual sustainability reports last week, which showed soaring emissions from the AI datacenter boom, driven by the construction of new datacenters, fuel used for deliveries, and expanding electricity usage. Google's total carbon emissions climbed 25% year-over-year, and Amazon's shot up 16%.
Google and Amazon Face Rising Emissions
While Google touted its sustainability progress, it conceded that combating the energy and water-hungry needs of AI created a dilemma. “The environmental footprint of the data centers that power AI is growing, creating a dual challenge: managing that environmental footprint while simultaneously building infrastructure to meet growing demand and realize Al’s full potential,” reads Google’s report. Similarly, Amazon said: “We recognize that the path to being a more sustainable company is not a straight line. Though our emissions increased in 2025, we remain steadfast in our commitment to sustainability.”
Google, Amazon and Microsoft, which will release its sustainability report in the coming weeks, fashioned themselves climate leaders in the tech sector in the previous decade, when investors prioritized ESG (environmental, social, and governance). Each set ambitious net-zero carbon goals and has heavily invested in sustainable energy projects like wind and solar.
AI Arms Race Undermines Climate Pledges
However, as these companies fight to win the AI arms race, their stock prices depend on their ability to integrate AI, which needs energy. Their desire for huge amounts of power has outstripped their commitments to sustainability, so their emissions promises have softened. Microsoft’s 2025 sustainability report documented a 23% increase in emissions compared with a 2020 baseline. The company’s investment in AI infrastructure has expanded since last year’s report, so 2026’s disclosures seem likely to reveal a similar or even larger spike. Meta’s 2025 sustainability report showed that emissions jumped 64% year-over-year in spite of a pledge of net-zero emissions by 2030.
Google has noted increased emissions every year since 2023 – attributing the upward slope to datacenter energy consumption. By 2025, the company had stopped speaking in terms of concrete goals for 2030 and instead framed its emissions ambitions as “climate moonshots”, a term the company uses to denote speculative projects that may or may not come to fruition. All four companies have turned to fossil fuels to provide additional power to their AI datacenters, inking contracts for enormous amounts of gas-generated electricity in Texas, Indiana and Louisiana.
Fossil Fuel Reliance for AI Datacenters
Tech’s new need for fossil fuels doesn’t seem short-lived. The non-profit Environmental Integrity Project released a report last week reviewing plans for 74 gas-fired power plants that will cater to datacenters across the US. The group estimates that these facilities combined could emit upwards of 660m tons of greenhouse gas pollution per year – equivalent to the entire country of Australia.
Meta's Frantic Moves Amid AI Turmoil
It’s been a tumultuous year at Meta, even by the standards of a company historically mired in scandal and prone to massive shifts in strategy. The artificial intelligence boom increased the value of many major tech firms, but Meta’s stock is down about 9% year-to-date. The company laid off about 10% of its global workforce in May, including over 2000 workers at its Menlo Park headquarters.
A messy reorganization tied to AI led to a top executive internally admitting leadership did an “atrocious job explaining the vision” to employees, and last week, CEO Mark Zuckerberg conceded in a town hall meeting that its plan to build AI agents wasn’t progressing as quickly as hoped. Meanwhile, Meta is still reeling after losses in two landmark social media addiction trials earlier this year and facing an avalanche of future legal battles over whether its social networks harmed children’s mental health.
Meta Explores Cloud Computing and Prediction Markets
While Meta struggles to find a clear path forward in the AI era, it has feverishly pursued numerous other avenues to calm investor concerns and seek out new revenue. Bloomberg reported last week that the company would enter the cloud computing business, selling off AI computing power to developers in a similar arrangement to Amazon Web Services’ products. The New York Times reported that Meta is hoping to break into the highly profitable and controversial prediction markets industry. The company has built an app internally called “Arena”, the Times wrote, which would function similarly to Polymarket and Kalshi. The app would be independent from Meta products like Instagram or Facebook, as well as fit into a pattern of the company making its own version of pre-existing products.
Mike Proulx, market research firm Forrester’s VP research director, said the move came out of “Meta’s familiar copycat playbook”. Meta’s potential foray into prediction markets, meanwhile, steers the company back into the loosely regulated, ethically dubious territory that started its current legal battles: habit-forming products.
Meta's Future Uncertain
The developments highlight the challenging times one of tech’s biggest players finds itself in. It hasn’t convincingly communicated how it fits into the new AI-driven tech world. The simultaneous internal chaos and outward expansion give off the sense that the company is scrambling to find where it goes next. “Set aside the debate on whether prediction market apps are investing or gambling, they’re habit-forming. And Meta is already facing high-profile litigation tied to concerns about addictive product design,” Proulx said. “The irony here is hard to miss and not a great look for a company already under scrutiny.”
