The LinkedIn co-founder and investor in both Anthropic and OpenAI provides the sharpest public assessment yet of Elon Musk’s AI ambitions and warns of the government’s treatment of Anthropic’s pull model
As a founder, lead investor, and 10-year member of Microsoft’s board of directors, Reid Hoffman has seen the AI industry from almost every angle. So when he calls SpaceX’s AI strategy “buying its way to relevance” and describes xAI as “a complete train wreck,” it’s not a heated observation from the sidelines, but a verdict from one of Silicon Valley’s most respected voices.
“SpaceX is not an AI company,” Hoffman said in a conversation with Rana El Karioby. AI pioneer Podcast. “XAI, as Elon himself has said, is a complete train wreck in terms of building foundational models and other types of things.” He also said that all of the founders have left and that the company is currently on its “third fresh start.”
The exodus of xAI co-founders is well documented. By May 2026, all 11 of xAI’s original co-founders had left the company, but the chain of events began in earnest in February when Tony Wu, said to be one of the most operationally central co-founders, announced his resignation. In response, Musk reorganized xAI’s team, but turnover continued. The company’s flagship Grok model has come under persistent criticism for lagging behind competitors from Anthropic and OpenAI in benchmark performance.
The timing of Mr. Hoffman’s remarks has been pointed out. SpaceX went public on June 12, putting AI at the center of its IPO story. Within days, the company announced it was acquiring Cursor, an AI coding tool. “That’s not proof of AI’s ability, it’s proof of its non-existence,” Hoffman read. “You can almost think of this as the IAC of AI,” he said, citing Barry Diller’s Internet-era conglomerate acquisition rollup strategy. “They will use their market capitalization to acquire AI companies and try to gain relevance.”
His assessment of SpaceX’s core computing business was similarly tepid. The company is positioning its revenue from leasing AI infrastructure, including Anthropic, as validation of its AI credentials. “You’re CoreWeave at a premium price,” Hoffman said. “Okay, we’re not an AI company.”
human condition
If Mr. Hoffman’s commentary on SpaceX was skeptical, his reaction to the U.S. government forcing Anthropic to remove its fabled and mythical models from the market bordered on alarm.
The directive was issued by the U.S. government on June 11 as an export control order, halting all access to both models by foreigners. According to reports from luckwas when Amazon CEO Andy Jassy sounded the alarm about a jailbreak discovered in the Fable 5 model, a vulnerability that Anthropic itself was trying to address. Cybersecurity experts widely criticized the government’s response as disproportionate and inadequate in scope.
Hoffman lands in a similar spot. “There doesn’t seem to be any particular principle. This is how we get through things, applying some sort of law and predictability,” he said. “It’s more like, ‘Anyway, we had a controversial interaction with this company, so let’s hit them with a stick.'” And that doesn’t include any principled explanation, he added.
He called the approach “autocratic and arbitrary” and “very suboptimal,” but acknowledged there may be a legitimate cybersecurity basis. What bothers him most is the asymmetry in which humans are penalized while OpenAI is not. Notably, Anthropic itself has flagged security concerns with the model, details of which el Kaliouby raised in the conversation.
For companies preparing for what is expected to be the largest IPO in history, unpredictable regulatory intervention creates a new category of investor risk, illustrated by an episode of fable and myth.
Room for both – but not for everyone
Hoffman, who is an investor in both Anthropic and OpenAI, was adamant that the two companies are in a zero-sum competition. “We tend to like to tell these stories as cage matches where two companies come in and only one company leaves,” he said, “but the reality is that both companies have an incredible amount of room to win.
He sketched out a clear competitive lane. Anthropic is strong in code and has expanded into design and legal. OpenAI and ChatGPT function like consumer search front ends and are “not talked about enough” given the strength of their Codex coding products. The only pointed question he posed was whether Cursor, which was just acquired by SpaceX, has already peaked. “The cursor seems to have had a bright star a few months ago and is disappearing over the horizon,” he said. Cursor has faced increasing pressure since early 2026, with the rise of Claude Code and Codex, and developers increasingly questioning whether standalone coding IDEs still warrant a premium. But of course, Hoffman has a vested interest in debating each of these positions.
On the topic of bubbles, he provided a framework for understanding market speculation. It would be wrong to say all valuations are crazy, but some are not, he said. “Which matters?” His anchor for bullish claims about OpenAI and Anthropic: If AI becomes as widespread as electricity, these two will become major power companies. And revenue models don’t need to be fully visible today. After all, Google’s early monetization theory was Enterprise Server, followed by AdWords, which is, to date, “the best business model ever invented in human history.”
Tell Gen Z to stop booing AI
When it comes to how young people should navigate an AI-saturated job market, Hoffman’s advice was blunt and contrary to conventional wisdom.
“I was thinking of writing an essay about the mistakes college graduates make when they boo or diss AI,” he says. As if speaking to all of Gen Z, he added: “You have an opportunity to become the AI generation, where you enter the workforce saying, ‘I know more than you. Hire me to help you become an AI-native organization.’ And that should be an opportunity, not a threat.”
Goldman Sachs publishes a semi-regular AI tracker that finds that AI will eliminate about 16,000 net U.S. jobs per month in April 2026, and 11,000 jobs as of the beginning of this month. Gen Z bears a disproportionate share of the impact, as entry-level knowledge jobs face the highest risk of turnover. According to another study, the unemployment rate for graduate students rose from 3.6% in 2019 to 5.6% in 2026. By mid-2026, 35% of entry-level jobs required at least three years of experience, and 45% of companies were using automatic rejection systems in the early stages of recruitment.
Hoffman’s counterargument is that most of that pain is caused by the wrong cause. He argued that the entry-level slump has more to do with other factors, with “AI washing” doing much of the work in the story. “It’s really because of global turmoil and companies not knowing how to invest and plan,” he said. “Remote work may be really successful during the pandemic because of basically overemployment. Oh, yeah. It’s pretty hard to work remotely.”
His prescriptions excerpted from his book super agencyis an agency mindset that treats AI not as a threat to your career, but as a means to it. “AI is my tool, my buddy, my car, etc., as I go about things,” he said. “AI can do a lot of amazing things on its own, but it’s not perfect, and humans can add a lot of important and important things.”
Microsoft chapter ends
Hoffman’s departure from the Microsoft board (he chose not to run for re-election and will remain on the board through the end of the year) caps a chapter that included facilitating the LinkedIn acquisition (“one of the epic M&A deals in history”), steering the GitHub acquisition, and helping broker trust between Microsoft and OpenAI in the early days of the partnership. He explained his decision succinctly. “I’d rather be a founder than a governance person.” He noted that he and Satya Nadella still discuss strategy. “Actually, Satya and I were just on the phone today.”
Next comes drug discovery. His company, Manas AI, with co-founders Ujjwal and Sid, has a small molecule proposal that computational chemists say is really promising, which Hoffman says was the catalyst for his decision to go all-in. The pitch materials he disclosed describe Manas as an “AI drug discovery factory to create monopolies,” but he points out that this is legally permissible because pharmaceutical IP functions as a purposefully granted monopoly. This may be the longest bet ever for the man who helped build the technology era at LinkedIn and Microsoft.
Regarding this story, luck Journalists used generative AI as an investigative tool. Editors verified the accuracy of the information before publication.
