Hoffman: SpaceX AI strategy is buying relevance, not building it

Reid Hoffman has spent years watching Silicon Valley from the inside — as a founder, an investor, and a decade-long Microsoft board member. So when he says SpaceX’s AI strategy amounts to buying relevance rather than building capability, and describes xAI as “a complete train wreck,” it lands with the weight of someone who actually knows where the bodies are buried.

Key takeaways

• Reid Hoffman says SpaceX is not an AI company and criticizes its acquisition strategy as an attempt to “buy your way into relevance.”
• All 11 of xAI’s original co-founders had left the company by May 2026, signaling deep organizational instability.
• SpaceX went public on June 12, 2026 with AI at the center of its IPO story, then quickly acquired Cursor — an AI coding tool Hoffman believes has already peaked.
• Reflection AI signed a deal to pay SpaceX $150 million per month starting July 1, 2026, for access to Nvidia GB300 chips at the Colossus 2 data center — a deal worth up to $6.3 billion through 2029.
• The U.S. government issued an export control order on June 11, 2026, suspending foreign access to Anthropic’s Fable and Mythos models — a move Hoffman called “autocratic willy-nilly.”

Reid Hoffman on SpaceX and xAI: buying the ticket, missing the ride

Speaking on Rana el Kaliouby’s Pioneers of AI podcast, Hoffman was direct: “SpaceX isn’t an AI company.” His argument isn’t about rockets or Starlink — it’s about what it takes to actually compete at the frontier of artificial intelligence. Leasing compute and snapping up AI startups, in his view, is not the same thing as building foundational models that move the field.

His comparison was pointed. “You could almost think of it as the IAC of AI,” he said, invoking Barry Diller’s internet-era acquisition machine — a company that used market capitalization to roll up digital properties without ever owning the underlying technology. “Use the market cap to buy AI companies and try to buy your way into relevance.” And for the compute leasing business specifically, Hoffman put it more bluntly: “You’re a premium-priced CoreWeave. I get it. Which is not an AI company.”

The critique lands harder when you factor in what SpaceX has actually been doing. The company went public on June 12, 2026, with AI woven through its IPO narrative. Within days, it announced the acquisition of Cursor, the AI coding tool. And its Colossus 2 data center — originally built by xAI for its own model training — is now functioning as a commercial compute platform, with Anthropic paying $1.25 billion per month, Google paying $920 million per month, and now Reflection AI committing $150 million per month starting July 1, 2026 through 2029. That deal alone is worth up to $6.3 billion if it runs to term, with either party able to exit with 90 days’ notice after the first three months.

That’s real revenue. But it’s infrastructure revenue — the kind that makes you a landlord in the AI race, not a player in it.

The xAI collapse Hoffman didn’t sugarcoat

xAI’s troubles go beyond strategic positioning. Hoffman called it “a complete train wreck for its kind of building of foundational models,” noting the company was on its “third restart.” The numbers behind that assessment are stark: by May 2026, all 11 of xAI’s original co-founders had departed the company. The wave began in earnest in February when Tony Wu, considered one of the most operationally central figures, announced his resignation. Restructuring followed, but the departures kept coming.

The organizational exodus matters beyond the human interest angle. Co-founder continuity is one of the strongest early signals of research culture and model quality. When every original architect has left, the institutional knowledge and technical direction they embodied leaves with them. Grok, xAI’s flagship model family, has faced persistent criticism for lagging behind Anthropic and OpenAI on benchmark performance — and the co-founder turnover makes a meaningful course correction harder, not easier.

Reflection AI’s Colossus deal and what it reveals about SpaceX

The Reflection AI compute agreement — confirmed by both TechCrunch and CNBC — adds another dimension to the SpaceX AI infrastructure story. Reflection, founded in 2024 by two former Google DeepMind researchers and last valued at $25 billion, is building open-weight AI models and working with government clients including the Department of Energy’s Genesis Mission and broader Pentagon AI programs.

Its pitch is deliberately timed. After the U.S. government suspended foreign access to Anthropic’s closed Fable and Mythos models, open-source AI gained a new and concrete argument: if a government can shut down a closed model overnight, maybe enterprises and nations should think harder about dependence on proprietary systems. Reflection has leaned directly into that narrative, positioning itself as “American open intelligence.”

For SpaceX, each new compute deal reinforces the same pattern Hoffman is critiquing — AI infrastructure as revenue, not AI capability as a competitive advantage. The company now has Anthropic, Google, Cursor, and Reflection as compute customers. That’s a formidable commercial position. Whether it constitutes an AI strategy is a different question entirely.

Anthropic, export controls, and what unpredictable regulation actually costs

On June 11, 2026, the U.S. government issued an export control order suspending all foreign national access to Anthropic’s Fable and Mythos models. The trigger, according to reporting from Fortune, was Amazon CEO Andy Jassy raising concerns about a discovered jailbreak in the Fable 5 model — a vulnerability Anthropic itself had been working to fix. Cybersecurity experts widely called the government’s response disproportionate.

Hoffman, who is an investor in Anthropic, landed in a similar place — but his concern went beyond the specific decision. “It doesn’t look like there’s anything that’s a particular principled, here’s-the-way-that-we’re-navigating-through-things, apply-kind-of-a-rule-of-law-and-predictability,” he said. The practical translation: the government appeared to be “hitting them with a stick” because of pre-existing friction with the company, not because of a clear and consistently applied framework.

He called the approach “autocratic willy-nilly” and “very sub-optimum.” The asymmetry — Anthropic penalized while OpenAI was not — is what troubled him most. And the implications reach beyond Anthropic’s immediate business. For a company preparing what is expected to be one of the largest IPOs in history, regulatory action of this kind creates a new and poorly-priced risk category for investors: not market risk, not technical risk, but the risk that a government intervention can suspend your product line with minimal notice and opaque reasoning.

That risk, now documented, will be priced into future AI IPOs in ways it wasn’t before.

Why Hoffman sees room for both Anthropic and OpenAI — and where Cursor fits

Hoffman pushed back firmly on the zero-sum framing that dominates AI coverage. As an investor in both Anthropic and OpenAI, he argued the two companies occupy meaningfully different competitive lanes. Anthropic, in his view, is strongest in code and expanding into design and legal applications. OpenAI and ChatGPT function more like a consumer-facing search interface, while its Codex coding product is “insufficiently talked about” given its actual capabilities.

The pointed question he raised about Cursor — the AI coding tool SpaceX just acquired — was telling. “Cursor seems to have had its bright star some number of months ago and seems to be fading over the horizon,” he said. Cursor has faced mounting competitive pressure since early 2026 as Claude Code and Codex gained ground, with developers increasingly questioning whether a standalone AI coding IDE still commands a premium. If Hoffman is right, SpaceX acquired a tool at or past its peak — another data point for the “buying relevance” critique.

Gen Z, the job market, and Hoffman’s counter-narrative

On the question of AI and employment, Hoffman’s advice to younger workers cut against the prevailing anxiety. “I’ve been thinking about writing an essay on the kind of mistakes that are made by college graduates booing or otherwise dissing AI,” he said. His prescription: treat the moment as an entry advantage, not a threat. “You guys have the opportunity to be generation AI — where you come into the workforce saying, ‘I know this a lot better than all of you.'”

The data he’s pushing against is real. A Goldman Sachs AI tracker found that by April 2026, AI was erasing roughly 16,000 net U.S. jobs per month, with Gen Z bearing a disproportionate share as entry-level knowledge roles face the highest displacement risk. Graduate unemployment rose from 3.6% in 2019 to 5.6% in 2026. By mid-2026, 35% of entry-level job postings required at least three years of experience, and 45% of companies were using automated rejection systems at early hiring stages.

Hoffman’s counter is that most of this pain is being misattributed to AI when the actual drivers are pandemic over-hiring, remote work normalization, and global economic turbulence. His book Superagency frames the prescription as an agency mindset: “The AI is my tool, companion, car, et cetera, as I navigate things. The AI can do a whole bunch of amazing things itself but is not complete — and humans can add in a lot of significant and important things.”

Manas AI and what comes next for Hoffman

Hoffman’s own next chapter is furthest from the infrastructure and compute debates dominating the current AI conversation. After stepping back from the Microsoft board — a tenure that included facilitating the LinkedIn acquisition, the GitHub purchase, and the early Microsoft-OpenAI partnership — he’s now focused on Manas AI, his drug discovery company co-founded with Ujjwal and Sid.

Manas is generating small molecule proposals that the company’s computational chemists describe as “genuinely promising” — Hoffman says that assessment was the trigger for his decision to go all-in. The company’s pitch deck, he revealed, describes it as “an AI drug discovery factory for creating monopolies” — legally permissible because pharmaceutical IP is a government-sanctioned monopoly by design.

It’s the longest-horizon bet in a career defined by them. And notably, it’s the one play in his portfolio that has nothing to do with whether SpaceX is or isn’t an AI company — a debate that, given the Colossus compute deals now accumulating, the market may resolve faster than Hoffman’s critique suggests.

FAQ

Why does Reid Hoffman say SpaceX is not an AI company?

Hoffman’s argument is that SpaceX’s AI involvement is primarily in leasing compute infrastructure and acquiring AI startups, not in building core AI models or advancing frontier research. He compares the approach to an acquisition roll-up strategy — buying market presence rather than earning it through technical capability.

What issues is xAI facing according to Reid Hoffman?

Hoffman describes xAI as “a complete train wreck” in terms of building foundational models, and notes the company has gone through multiple restarts. The most concrete indicator of instability: by May 2026, all 11 of xAI’s original co-founders had left the company.

What is the U.S. government’s stance on Anthropic’s AI models?

On June 11, 2026, the U.S. government issued an export control order suspending foreign national access to Anthropic’s Fable and Mythos models, citing security concerns including a discovered jailbreak vulnerability in the Fable 5 model. Hoffman criticized the action as lacking principled justification and called the regulatory approach “autocratic willy-nilly.”

What advice does Reid Hoffman give to Gen Z about AI and careers?

Hoffman urges Gen Z to embrace AI as a professional advantage rather than a threat. His view is that young workers who develop deep AI fluency can position themselves as essential guides for organizations trying to become AI-native — turning a moment of disruption into a career entry point.

{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"Why does Reid Hoffman say SpaceX is not an AI company?","acceptedAnswer":{"@type":"Answer","text":"Hoffman's argument is that SpaceX's AI involvement is primarily in leasing compute infrastructure and acquiring AI startups, not in building core AI models or advancing frontier research. He compares the approach to an acquisition roll-up strategy — buying market presence rather than earning it through technical capability."}},{"@type":"Question","name":"What issues is xAI facing according to Reid Hoffman?","acceptedAnswer":{"@type":"Answer","text":"Hoffman describes xAI as \"a complete train wreck\" in terms of building foundational models, and notes the company has gone through multiple restarts. The most concrete indicator of instability: by May 2026, all 11 of xAI's original co-founders had left the company."}},{"@type":"Question","name":"What is the U.S. government's stance on Anthropic's AI models?","acceptedAnswer":{"@type":"Answer","text":"On June 11, 2026, the U.S. government issued an export control order suspending foreign national access to Anthropic's Fable and Mythos models, citing security concerns including a discovered jailbreak vulnerability in the Fable 5 model. Hoffman criticized the action as lacking principled justification and called the regulatory approach \"autocratic willy-nilly.\""}},{"@type":"Question","name":"What advice does Reid Hoffman give to Gen Z about AI and careers?","acceptedAnswer":{"@type":"Answer","text":"Hoffman urges Gen Z to embrace AI as a professional advantage rather than a threat. His view is that young workers who develop deep AI fluency can position themselves as essential guides for organizations trying to become AI-native — turning a moment of disruption into a career entry point."}}]}