Produced by Zhineng Technology

Great companies usually die in one of two ways. One is like Nokia, standing helplessly in the twilight of its empire, watching the sunset. The other is like xAI on May 6, 2026, dying on an extraordinarily ordinary afternoon.

Musk announced on X that xAI would no longer exist as an independent company and would become an AI product under SpaceX (SpaceXAI). Its core computing asset, Colossus 1 (220,000 GPUs), was leased to Anthropic (the parent company of Claude).

This left us puzzled: Elon, why the betrayal!?

With just a few lines, Elon Musk stuffed xAI into SpaceX's storage closet.

The most powerful computing cluster on Earth, Colossus, became a 'low-rent apartment' leased to a rival.

xAI, once a 'world-changing AI player,' is dead. What remains is a 'shrewd landlord of electricity and computing power.'

01

Collapse:

When the 'World Challenger' Decides to Go Home

Many say xAI lost to OpenAI. That’s too lazy an explanation.

xAI wasn’t short on money or GPUs. Its consistent problem was defining what 'winning' meant.

By early 2026, the elite team recruited from Silicon Valley’s top firms, driven by a sense of mission, began leaving en masse in 'silent resignations.'

The pre-training lead departed. The engineering vanguard withdrew. By late March, when the last two core members walked out, only the hum of server cooling fans remained in xAI’s office.

When even the top talent realized they couldn’t win under Musk—when Grok failed to escape X’s (formerly Twitter’s) ecosystem to become a true productivity tool—the game was over.

Grok didn’t fail, but the market didn’t need Grok to succeed. In AI’s deep waters, if you’re not a layer others must rely on, you’re disposable.

The most absurd part of this story: SpaceX leased the 220,000-GPU Colossus cluster to Anthropic.

Who is Anthropic? OpenAI’s most direct rival and a frequent target of Musk’s online mockery. Leasing your core strategic weapon to a rival is called 'aiding the enemy' in wartime and 'stopping losses' in business.

For xAI, this was a transfer of dignity.

Data centers are cash-flow black holes. Electricity, cooling, maintenance, depreciation—when you realize you can’t build a world-dominating model, the best move is to convert your 'mine' into an 'office building' and lease it to dreamers. Musk acted pragmatically and ruthlessly.

02

Terafab: Musk’s Future

On the same day xAI vanished, applications for Texas’s Terafab chip factory surfaced. Phase one: $55 billion. Final scale: $100 billion.

If the application layer is too crowded, go straight to the physical layer.

Nvidia is now the world’s most valuable company, capturing 80% of industry profits and defining the underlying physical rules.

Musk’s plan:

◎ From chip design to building factories for manufacturing.

◎ To train models while controlling every step, from power intake to computing output.

From chip fabrication to liquid cooling systems to computing leases, xAI once stood in the gold mine wielding a pickaxe. Now it’s the giant excavator.

To make this 'landlord plan' less dull, SpaceX floated the 'space computing' fantasy.

◎ The logic is grand: Earth’s energy is insufficient, cooling is too hard—we must compute in orbit.

◎ Physically, it’s beautiful.

◎ Financially, it’s a disaster.

Launching servers into space, solving latency, mitigating cosmic rays—each cost alone could bankrupt a trillion-dollar company.

But 'space computing’s' true purpose isn’t solving computational problems—it’s solving 'valuation problems.'

When xAI lost its independent value, it had to attach itself to a grander narrative: SpaceX’s cosmic ambition, where it need only prove itself as the 'only brain capable of operating in Martian orbit.'

Conclusion

May 6 marked the end of the era where 'a few geniuses and a pile of GPUs could subversion (disrupt) the world.'