The AI industry faces an unprecedented infrastructure crisis as OpenAI's flagship $500 billion 'Stargate' project grinds to a halt in Britain, with its planned 8,000-chip facility remaining shuttered months after its promised 2026 opening. This paralysis comes amid Oracle's massive 30,000-worker layoff, revealing a sector-wide bottleneck that has stalled half of Earth's AI projects and forcing industry leaders to urgently pivot toward orbital computing solutions as terrestrial limitations become insurmountable.

The problems aren’t limited to a few companies

Research group Sightline Climate found that between 30% and 50% of large AI facilities planned for the United States this year will face delays or get scrapped entirely. The group looked at 140 building projects representing at least 16 gigawatts of capacity scheduled to open by the end of 2026, but only around 5 gigawatts are actually being built right now.

Last year saw similar issues, with 26% of announced capacity getting delayed and another 10% pushed back to later dates. For 2027, plans call for more than 25 gigawatts, but less than 10 gigawatts is currently under construction.

Power supply is the biggest problem. Global electricity use by these facilities hit roughly 415 terawatt-hours in 2024, and the International Energy Agency thinks it could top 1,000 terawatt-hours by 2026. In Virginia, these facilities already consume 26% of all electricity. Ireland could reach 32% by the end of this year.

Getting basic equipment like batteries and transformers has also become difficult, according to Bloomberg.

Some companies are proposing wild solutions

SpaceX filed paperwork on January 30 asking to launch up to one million satellites to create what it called a constellation with “unprecedented computing capacity to power advanced artificial intelligence models.” Seven weeks later, Blue Origin filed for 51,600 satellites in its Project Sunrise.

Starcloud, a startup, raised $170 million in March at a $1.1 billion value and filed for permission to launch 88,000 satellites. Aethero, another new company, raised $8.4 million for similar work.

But scientists say the physics don’t work. Getting rid of heat in space requires about 1,200 square meters of radiator for every megawatt of power. Radiation damages chips.

Communication delays make training AI models nearly impossible. IEEE Spectrum calculated a one-gigawatt space facility would cost upward of $50 billion, roughly three times what a similar ground facility would cost.

Even Altman has called the idea “ridiculous” for this decade, pointing out that launch costs relative to regular power costs “simply does not work yet.”

The crypto card with no spending limits. Get 3% cashback and instant mobile payments. Claim your Ether.fi card.