This article is an on-site version of our Moral Money newsletter. Premium subscribers can sign up here to get the newsletter delivered twice a week. Standard subscribers can upgrade to Premium here, or explore all FT newsletters.
Visit our Moral Money hub for all the latest ESG news, opinion and analysis from around the FT
Hello from Shanghai.
Four years away from the deadline for Microsoft’s 2030 “carbon negative” pledge, things aren’t looking good. The company’s carbon emissions increased by more than a quarter last year, it reported yesterday.
The AI data centre boom is sending tech sector emissions skywards — and raising fears that grid power prices will head in the same direction. In today’s newsletter, Chinese cleantech billionaire Zhang Lei argues for his radical solution: taking these facilities off the grid and into the desert.
I’m off from next week for my four-yearly sabbatical (one of the institutions that make the FT great). Stay tuned for weekly dispatches from some of my brilliant colleagues over the next month; I’ll be back in your inbox on Wednesday August 12. See you then.
Do data centres belong in the desert?
Big Tech has a big problem.
As it rolls out data centres at head-spinning scale to power the AI boom, it’s facing a growing wave of local and political opposition, fuelled largely by concerns about the effects of these facilities’ vast hunger for electricity.
Most of this demand is being met by power grids, which are coming under pressure across the US and beyond. Big US tech companies are also commissioning additional generation of their own — mostly in the form of gas-fired plants, which are driving up planet-warming carbon emissions and local air pollution.
Chinese clean energy billionaire Zhang Lei says he has the answer: build these data centres in the desert, running on dedicated microgrids powered entirely by wind and solar power.
“People haven’t found a solution to power AI. And it’s disrupting the current legacy grid system,” Zhang told me this week when we met at the Shanghai headquarters of his company, Envision. “So why not create a new energy system for AI?”
After an early career as an energy trader in London, Zhang founded Envision in 2007 to build wind turbines, a sector in which it now ranks among the world’s biggest producers.
Zhang has been using that as a foundation to expand into other low-carbon sectors, from batteries to green hydrogen and energy management software — in the process making himself one of Chinese cleantech’s most prominent names and building a personal fortune of $4.5bn, according to the latest Hurun China Rich List.
Last month, he announced his new AI-focused strategic push, codenamed Mission Gobi, named after the desert that sprawls from Mongolia into northern China. Envision last year opened a large-scale industrial park in China’s Inner Mongolia region, which is using the Gobi’s abundant wind and sunlight to produce green hydrogen for low-carbon fuels.
Now Envision has built at that site what it’s calling the world’s first AI data centre powered entirely by off-grid, directly supplied renewable energy.
Envision hasn’t disclosed the size of the facility, developed with Chinese tech giant Tencent. But Zhang is billing it as a proof of concept, demonstrating that — thanks largely to technological advances in large-scale batteries — data centres can now be run around the clock on renewable energy, without sacrificing reliability or cost-effectiveness.
At another site in Inner Mongolia, Envision is building a far larger AI data centre using this model, which it says will be a “gigawatt-scale” facility, referring to the power consumption at peak load. For context, this figure for OpenAI’s data centre capacity was 1.9GW at the end of last year. By 2030, Envision will have built 5GW of data centres powered by off-grid renewables, Zhang has vowed.
His agenda is a challenge to the energy strategy being pursued by big US tech companies.
Sam Altman — chief executive of OpenAI, which uses data centres operated by companies including Microsoft and Oracle — has been blunt about how he thinks the AI boom will be powered. “Short-term, natural gas,” he said late last year, adding that he thought nuclear fusion (a commercial energy source that doesn’t yet exist) and solar power would dominate in the longer term.
Even as they lean heavily on gas power to fuel their expansion, companies such as Microsoft and Amazon still claim to be pursuing green goals requiring huge cuts in their carbon emissions. They’re aiming to square that circle by entering financial contracts that enable them to take credit, in carbon accounting terms, for the “environmental attributes” of renewable power used by others, and use this to offset their emissions.
Zhang is openly scathing about the US tech companies’ energy strategies. “If you are not embracing renewables, you’re bullshitting,” he said, adding that AI companies risked pushing populations into “energy poverty” by consuming excessive amounts of grid power.
“They don’t have real responsibility,” he said. “That’s what disappointed me.”
Zhang’s desert vision faces some obvious questions of its own — not least around water. Many of the world’s data centres have been built using evaporative cooling systems to stop servers from overheating — resulting in heavy water consumption that’s raised concerns about the impact on local communities.
Envision plans to take advantage of technological developments in “dry” and closed-loop liquid cooling techniques, which have dramatically smaller water requirements (though they can entail higher electricity needs).
Then there’s the question of how much he’ll need to spend on battery storage capacity and renewable generation “overbuild” to ensure the near-100 per cent level of data centre reliability demanded by corporate AI customers.
Still, his vision seems to fit with the Chinese government’s broader strategy for AI infrastructure growth. Since 2022, Beijing has been pursuing an agenda to build more data centres in sparsely populated inland regions to exploit low-cost renewable energy, with huge investment in network infrastructure to link them with the country’s business hubs.
The international picture looks much more complicated. The US government is aiming to hold back Chinese AI growth, including by restricting sales of advanced chips and equipment. In Europe, there is a growing political focus on the security implications of Chinese involvement in critical infrastructure.
Zhang said such concerns were unwarranted, saying he saw particular potential for off-grid data centres in arid parts of Spain. Envision already has production sites in that country as well as France and the UK, where it also has contracts to develop large-scale battery storage systems. “This Chinese renewable technology is kind of a civilisational output for the rest of the world,” Zhang said.
Smart reads
Recommended newsletters for you
The Climate Graphic: Explained — Understanding the most important climate data of the week. Sign up here
The AI Shift — John Burn-Murdoch and Sarah O’Connor dive into how AI is transforming the world of work. Sign up here

