SpaceX’s IPO drew global attention, not only refreshing the record for a first raise at US$75 billion but becoming the world’s fourth-largest listed company on its third trading day, sending Musk past a trillion-dollar fortune.
The prospectus stated a stark fact: over 90 per cent of its future market sits in AI. More importantly, AI growth is severely constrained by Earth’s power shortage.
Musk’s answer is to put AI data centres in space, using free solar, free of the grid. A grand goal, but as distant as landing on Mars. Capital is cooling. SpaceX shed US$400 billion in two days, the second-largest single-day wipeout in US history. To solve the AI power crunch now, a Chinese company may already be ahead.
That company is Envision Group, building an AI power system, an AI infrastructure fusing energy and intelligent systems. At a recent National Energy Administration AI plus energy meeting, Envision chairman Zhang Lei said the power system is becoming the main project of AI. Only by solving full-chain energy management of intelligent production can the new industrial revolution be powered.
AI calls for a new power system
AI data centres are true power hogs. IEA data shows that by 2030, AI data-centre power use hits 950 TWh, roughly double 2025’s 485 TWh, nearing Japan’s annual generation.
Use is surging. Data centres even squeeze residential power. In Virginia and Texas, outage time grows and some areas saw electricity up more than 12 per cent per kWh year on year.
Early this year, President Trump convened seven tech giants, Microsoft, Google, Amazon, Meta and OpenAI, to sign deals making them bear AI data-centre power costs, not pass them to the public.
Worse, European and US grids are ageing. Main lines average 40 to 50 years. EPRI notes grids and prior policy were not designed for AI infrastructure. There is a physical bottleneck. Rebuilding takes 5 to 10 years. Data centres build in months to a year.
In China, with strong grid control, the challenge remains. Data centres clustered in power-tight east. The east data, west computing strategy helps but long-haul transmission loses energy and cost. Under dual carbon, thermal is unsustainable. New rules require new national-hub AI data centres to use at least 80 per cent green power.
SpaceX’s fix is to leave the ground: orbit-based data centres, solar at over five times ground potential, avoiding residential competition. But space data centres may not be the rational commercial choice. Per SpaceX’s AI1 blueprint, space PV needs higher material and structure reliability. Add satellite R&D, launch, space cooling, space-chip and operating costs.
SemiAnalysis estimates the space-versus-ground cost gap exceeds four times in 2026, not closing until about 2040.
Envision’s Mission Gobi
Envision’s answer is Mission Gobi, building AI data centres in the Gobi desert, where solar and wind are abundant and land cheap, paired with an AI power system.
The first layer is the brain of the AI power system, solving generation and cost. Envision uses a Tianji weather model to predict weather and help wind turbines and PV panels generate precisely. The second layer integrates storage and flexible load. The third layer is green power trading and carbon management.
Through an AI power system dynamically integrating each link of power transport, Envision achieves 100 per cent green supply and low-carbon production, cutting comprehensive energy cost by over 40 per cent and reducing annual carbon by a meaningful margin at its pilot.
Musk sells a 2040 vision. Envision is shipping 2026 capacity. For the AI power crunch, the bottleneck was never the idea of free solar but the system to deliver it on the ground, and that is what Envision is building in the desert.
More from the original report
Editor’s note: This is an adapted translation of the original LeiFengWang report. It has been trimmed and restructured for readability for an international business audience.
Translated and adapted from LeiFengWang (https://www.leiphone.com/category/transportation/xYypGspPg2OhtNK7.html).