Six weeks ago, Elon Musk merged xAI with SpaceX in the largest corporate deal in history - $1.25 trillion in combined valuation. This week, he publicly admitted the AI company was broken from the start.
“xAI was not built right first time around, so is being rebuilt from the foundations up,” Musk posted on X, following the departure of two more co-founders, Zihang Dai and Guodong Zhang. That brings the exodus count to nine of twelve original founding members who have walked away since xAI launched in July 2023.
Only three co-founders remain: Manuel Kroiss, Ross Nordeen, and Toby Pohlen. The people who designed xAI’s architecture, led its research teams, and understood its codebase have scattered to competitors, venture capital firms, and their own startups.
The Coding War Musk Is Losing
The immediate trigger for the latest departures was Musk’s frustration with xAI’s AI coding tools. According to the Financial Times, Musk complained that xAI wasn’t competing effectively against Anthropic’s Claude Code and OpenAI’s Codex - the AI-powered programming assistants that have become critical infrastructure for software development.
Guodong Zhang, who led xAI’s Imagine team, was reportedly blamed for the coding product’s poor performance and relieved of his primary duties. His response was to leave entirely. “Last day at xAI. Wild journey past three years but excited about next chapter,” Zhang posted.
The coding tools market has become intensely competitive. Claude Code and Codex have captured significant market share among professional developers, while Cursor - a standalone AI coding environment - recently hit a $2 billion valuation. Grok’s coding capabilities have trailed behind, and Musk’s response has been to blame the people who built them.
Raiding Cursor for Replacements
Rather than fix the internal culture problems driving people out, Musk went shopping at a competitor.
xAI hired Andrew Milich and Jason Ginsberg, two senior leaders from Cursor’s parent company Anysphere, to rebuild Grok’s coding capabilities from scratch. Both will report directly to Musk. Separately, xAI also poached Brandon Borzelli (head of revenue) and Tushar Makhija (business operations) from Anysphere.
It’s a significant talent grab. Milich and Ginsberg co-led product engineering during Cursor’s explosive growth period. The implicit message: xAI is scrapping whatever it built internally and starting over with people who know how to ship a successful AI coding product.
But raiding competitors doesn’t solve the underlying problem. If the executives who built xAI from zero couldn’t succeed in Musk’s environment, why would external hires fare better?
The Burn Rate Problem
xAI’s financial position makes the rebuild even more precarious. The company burned through $7.8 billion in the first nine months of 2025 while generating just $107 million in quarterly revenue. That works out to roughly $1 billion per month in cash burn against revenue that wouldn’t cover the electric bill on a single data center.
The most recent quarterly loss was $1.46 billion, widening from $1 billion in Q1. Management has told investors profitability won’t arrive until 2027 - which means xAI needs to burn through tens of billions more before it breaks even.
Tesla invested $2 billion in xAI’s Series E round at a $230 billion valuation. When SpaceX acquired xAI, Tesla’s stake converted to a minority SpaceX holding. Tesla shareholders are now exposed to both xAI’s losses and its reputational damage - without having voted on the investment.
The Grok Problem Hasn’t Gone Away
While the coding team was collapsing, Grok’s other problems continued to compound. The chatbot remains under investigation in multiple jurisdictions for generating child sexual abuse material and non-consensual deepfakes. The EU, UK, France, India, Malaysia, California, and other regulators have opened probes.
These aren’t abstract compliance issues. They’re existential threats to a company preparing for the most anticipated IPO in history. Institutional investors asking about xAI’s contribution to the combined SpaceX entity will want to know why its flagship product keeps generating illegal content and whether that liability transfers to the parent company.
Musk’s Apology and Re-recruitment
In an unusual move, Musk publicly apologized for xAI’s hiring decisions. “Many talented people over the past few years were declined an offer or even an interview @xAI,” he posted.
He said he and Baris Akis, who handles engineering talent at xAI, are “going through the company interview history and reaching back out to promising candidates.”
The admission is remarkable. Musk is saying that xAI’s original hiring process was so flawed that he needs to personally review years of rejected applications. It’s also a tacit acknowledgment that the people he did hire - including nine co-founders who have since left - either weren’t the right choices or couldn’t succeed in his management environment.
What This Means
xAI is now a company being rebuilt in public view. Nine of its twelve founders are gone. Its coding tools lost to competitors. Its chatbot generates content that attracts criminal investigations. It burns a billion dollars a month. And six weeks after a historic merger, its CEO admits it “was not built right.”
The SpaceX IPO is still planned for mid-2026. When that roadshow begins, xAI will be presented as a key component of a $1.25 trillion combined entity. Institutional investors will need to decide whether a company in the middle of a “foundations up” rebuild, run by executives poached from a competitor, represents a reasonable bet.
The AI industry has seen spectacular failures before. But most of them happened quietly, in private funding rounds and shuttered offices. xAI is different. It’s failing in front of everyone, with its CEO narrating each collapse in real time.
Musk has rebuilt companies before - Tesla nearly died multiple times before becoming the world’s most valuable automaker. But Tesla’s near-death experiences happened when Musk was focused on a single company. Today, he’s running Tesla, SpaceX, xAI, Neuralink, The Boring Company, and X simultaneously. The question isn’t whether xAI can be rebuilt. It’s whether anyone will stick around long enough to try.