xAI Dies at $230B Valuation as Musk Folds It Into SpaceX

Musk Kills xAI, Hands Its GPUs to Anthropic, and Bets on Chips — All in One Day

On the afternoon of May 6, Elon Musk announced on his social platform X that 'xAI will no longer exist as an independent company. It will simply be SpaceXAI — an AI product of SpaceX.' With that single post, the AI startup once valued at $230 billion ceased to exist as a standalone entity, marking one of the most dramatic pivots in recent tech history.

But the announcement was only the first of three seismic moves that day. SpaceX simultaneously signed a compute-leasing agreement with Anthropic, handing over exclusive access to xAI's crown jewel — the Colossus 1 data center and its 220,000 NVIDIA GPUs — to OpenAI's fiercest rival. And in a third move, SpaceX filed a construction application in Texas for a semiconductor megafactory called 'Terafab,' with phase-one investment of $55 billion and a total buildout cost of up to $119 billion.

Three events. One day. A death certificate for an AI company, a weapons-to-the-enemy deal, and the birth of a chip empire.

Key Takeaways

• xAI is no more — the company has been absorbed into SpaceX under the new brand 'SpaceXAI'
• Anthropic gains exclusive access to 220,000 NVIDIA GPUs from xAI's Colossus 1 data center
• Terafab, a new semiconductor factory in Texas, will cost up to $119 billion when fully built
• The moves signal Musk is exiting the LLM race and pivoting toward AI infrastructure and hardware
• xAI's $230 billion valuation effectively evaporates as an independent entity
• Musk appears to be repositioning his AI strategy around vertical integration with SpaceX's mission

The Rise and Fall of xAI: From 'Understanding the Universe' to Absorption

Musk launched xAI in July 2023 with characteristic fanfare. The founding team of 11 members pulled talent from DeepMind, OpenAI, and Microsoft Research, representing some of the most elite AI researchers on the planet. The stated mission was nothing short of grandiose: 'to understand the true nature of the universe.'

The real motivation was more personal. Musk wanted to build a direct competitor to OpenAI, the company he co-founded and later left, accusing it of betraying its open-source origins. He poured not just capital into the venture but also exclusive resources — most notably, the firehose of more than 500 million daily posts from X, providing an unmatched real-time training corpus.

The company moved at breakneck speed. Within just 122 days, xAI built and deployed the Colossus 1 supercomputer cluster, assembling 220,000 NVIDIA GPUs into one of the world's most powerful AI training facilities. The company launched Grok, its flagship large language model, which was integrated directly into the X platform.

Yet despite the rapid technical achievements, xAI struggled to find a sustainable business model independent of Musk's other ventures. The AI model market became increasingly commoditized, with OpenAI, Google, Anthropic, and Meta all racing to the bottom on pricing while simultaneously pushing capability frontiers. For xAI, competing on model quality alone proved insufficient.

The Anthropic Deal: Arming Your Rival's Rival

Perhaps the most eyebrow-raising move of May 6 was the compute-leasing agreement between SpaceXAI and Anthropic. Under the deal, Anthropic receives exclusive usage rights to the entire Colossus 1 data center — all 220,000 NVIDIA GPUs.

On the surface, this looks like Musk is handing his most valuable AI asset to a competitor. But the strategic logic becomes clearer when you consider the broader competitive landscape:

• Anthropic is OpenAI's primary competitor — strengthening Anthropic weakens Musk's true nemesis, Sam Altman
• GPU leasing generates immediate revenue without the R&D costs of running a competitive LLM
• Musk exits the model-training arms race while still profiting from the infrastructure layer
• Anthropic gains massive compute capacity at a time when GPU access remains the industry's biggest bottleneck

This is classic Musk: if you cannot beat your enemy directly, fund and arm their strongest rival. The deal transforms xAI's compute infrastructure from a cost center into a revenue-generating asset while simultaneously undermining OpenAI's market position.

For Anthropic, led by former OpenAI executives Dario and Daniela Amodei, the deal is a windfall. Access to 220,000 GPUs dramatically accelerates training capacity for future versions of Claude, potentially closing the compute gap with OpenAI and Google DeepMind.

Terafab: Musk's $119 Billion Bet on Silicon Sovereignty

The third announcement may ultimately prove the most consequential. SpaceX filed a construction application in Texas for a semiconductor fabrication facility called Terafab, with a staggering investment profile:

• Phase 1 investment: $55 billion
• Full buildout cost: up to $119 billion
• Location: Texas (specific site not yet disclosed)
• Focus: Advanced semiconductor manufacturing

To put this in context, TSMC's Arizona fab — the largest semiconductor investment in U.S. history — carries a price tag of approximately $65 billion across multiple phases. Musk's Terafab would dwarf it at full buildout, potentially becoming the single largest private manufacturing investment ever made on American soil.

The strategic rationale is multilayered. SpaceX already consumes enormous quantities of custom chips for its Starlink satellite constellation, which now includes more than 6,000 satellites in orbit. Each satellite requires specialized processors for communications, navigation, and increasingly, edge AI inference.

Musk has long expressed frustration with dependence on external chip suppliers, particularly given geopolitical tensions surrounding Taiwan and TSMC. A captive semiconductor facility would give SpaceX vertical integration from silicon to orbit — a supply chain advantage no competitor could match.

Moreover, with the CHIPS Act providing substantial federal incentives for domestic semiconductor manufacturing, the timing aligns with favorable policy conditions. Musk appears to be positioning SpaceX to capture both government subsidies and long-term strategic independence.

What This Means for the AI Industry

Musk's triple move on May 6 sends a clear signal: the AI value chain is shifting from models to infrastructure. By dissolving xAI as a standalone AI model company while simultaneously investing in compute leasing and chip manufacturing, Musk is betting that the real money in AI lies not in building chatbots but in controlling the physical layer.

This aligns with a broader industry trend. NVIDIA's market capitalization has soared past $2.5 trillion precisely because it controls the GPU layer. Cloud providers like AWS, Azure, and Google Cloud generate massive margins from AI compute rental. Meanwhile, AI model companies face relentless pressure on pricing, with API costs dropping 90% or more over the past 18 months.

For developers and businesses, the implications are significant:

• Anthropic becomes a stronger alternative to OpenAI, with massively expanded compute resources
• GPU availability may improve as xAI's hoarded capacity enters the broader market via Anthropic
• Chip supply dynamics could shift if Terafab reaches production, reducing dependence on TSMC
• The 'AI company' model is evolving — pure-play model companies may struggle without infrastructure advantages
• Vertical integration is becoming the winning strategy, as demonstrated by Google, Apple, and now SpaceX

The Bigger Picture: Musk's Empire Consolidation

Viewed in isolation, dissolving a $230 billion company looks reckless. But Musk rarely operates in isolation. Every move connects to a larger architecture.

SpaceX needs AI for autonomous rocket landing, Starlink network optimization, and eventually Mars colonization planning. By folding xAI into SpaceX, Musk ensures AI development is directly aligned with these missions rather than chasing consumer chatbot market share.

Tesla already has its own AI division focused on autonomous driving and the Optimus humanoid robot. Having a separate xAI competing for talent, GPUs, and attention created internal friction across Musk's empire. Consolidation eliminates that redundancy.

The Terafab announcement suggests Musk is thinking in decades, not quarters. A fully operational semiconductor fab would take 5-7 years to reach production capacity. But once online, it would provide SpaceX, Tesla, and any future Musk ventures with a captive chip supply — insulated from geopolitical risk, supply chain disruptions, and competitor pricing pressure.

Looking Ahead: What Happens Next

The immediate aftermath of May 6 raises several critical questions. What happens to xAI's employees? The founding team of elite researchers may not be content building satellite optimization algorithms when they signed up to 'understand the universe.' Talent attrition to competitors like OpenAI, Google, or Anthropic seems likely.

What happens to Grok? The chatbot integrated into X could continue as a SpaceXAI product, but without dedicated model development resources, it risks falling further behind GPT-4o, Claude 3.5, and Gemini.

The Anthropic compute deal also raises questions about duration and exclusivity. If the lease is long-term, SpaceXAI effectively exits the frontier model training business entirely. If short-term, it could be a temporary monetization strategy while Terafab ramps up.

One thing is certain: Musk has made his bet. The future of his AI ambitions lies not in competing with OpenAI on chatbots, but in owning the atoms — the chips, the data centers, the satellites — that make AI possible. Whether that bet pays off will depend on execution at a scale that would challenge even the world's most audacious entrepreneur.

The AI industry just witnessed something rare: a billionaire admitting that building models is not his game, and pivoting to where he believes the real leverage lies. May 6, 2025 may be remembered not as the day xAI died, but as the day Musk redefined what it means to be an AI company.