In a development that has sent shockwaves through the artificial intelligence industry, all 12 original co-founders of Elon Musk’s xAI have now departed the company. Reuters reported that just six of the twelve founding researchers remained as of February 10, 2026–the day Tony Wu and Jimmy Ba both resigned within 24 hours of each other; by late March 2026, the final two holdouts–Manuel Kroiss and Ross Nordeen–had also left, completing an unprecedented leadership exodus that began accelerating after SpaceX’s landmark acquisition of xAI in February 2026. Not every tracker agrees on the final tally: AI Wiki’s running co-founder count, last updated in August 2026, still listed Kroiss and Nordeen as active as of March 2026, putting the departure figure at 9 of 11 rather than a clean sweep–a discrepancy that itself underscores how chaotic and hard to verify the exodus has been. The departures raise fundamental questions about the future of Grok, the $18 billion Colossus supercomputer in Memphis, and whether Musk’s vision of rebuilding xAI under SpaceX’s engineering culture can succeed without the elite AI researchers who helped build it.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The Complete xAI Co-Founder Exodus: A Timeline
xAI was founded in March 2023 by Elon Musk alongside 12 of the most accomplished AI researchers in the world–though the roster was not always counted consistently even before the exodus began: a July 2025 startup profile from StartupIntros listed the founding team as Musk plus just seven co-founders, including Igor Babuschkin and Greg Yang, undercounting the fuller group that history would later document. Recruited from DeepMind, Google, OpenAI, Microsoft Research, and the University of Toronto, the founding team represented a concentration of talent that few organizations in artificial intelligence history have matched, backed by financing that a July 2025 table compiled by Morgan Stanley pegged at $10 billion total–$5 billion in equity and $5 billion in debt–with SpaceX itself contributing $2 billion of that raise well before the two companies formally combined. By March 28, 2026, every single one of the twelve founders had walked away.
The departures did not happen overnight. Christian Szegedy, who spent 12 years as an engineer and research scientist at Google, was among the earliest signals of internal friction, departing in February 2025–a move AI Wiki’s co-founder tracker credits with narrowing the founding roster to 10 remaining members at the time. But the real cascade began in February 2026, immediately following SpaceX’s acquisition of xAI. On February 10, 2026, Yuhuai “Tony” Wu–a former research scientist at Google with internships at both DeepMind and OpenAI–announced his departure; The Economic Times reported that Wu was the fifth xAI co-founder to quit, following one departure in 2024 and three more through 2025, while Reuters noted the same day that the resignation left just six of xAI’s twelve founding researchers still in place, a count AI Wiki’s own tracker mirrored, showing seven co-founders remaining immediately after Wu’s exit and six after Ba’s departure the following day. Within 24 hours, CNBC reported that xAI had lost its second co-founder in as many days: Jimmy Ba, an assistant professor at the University of Toronto who had studied under deep learning pioneer Geoffrey Hinton and co-authored the influential 2014 Adam optimization paper, followed Wu out the door.
The timing was not coincidental. According to reporting from TechCrunch and The Next Web, both departures were linked to tensions over demands to improve model performance and a cultural clash between xAI’s research-oriented origins and the operational intensity that came with the SpaceX integration. By early March, Guodong Zhang, who had led xAI’s Imagine team, and Zihang Dai, a former Google research scientist, also departed during what sources described as a SpaceX and Tesla audit of xAI’s operations–departures AI Wiki likewise dates to March 2026, pegging the remaining co-founder count at four after Dai’s exit and three after Zhang’s.
Igor Babuschkin, xAI’s chief engineer and a former research engineer at both DeepMind and OpenAI, had actually been among the first of the inner circle to go: TechCrunch reported on August 13, 2025 that Babuschkin left xAI to found his own venture, Babuschkin Ventures, roughly two years after co-founding the company–a departure Entrepreneur likewise dated to August 2025. That venture, since rebranded River AI, went on to raise $1.1 billion in a round announced in August 2026, led by General Catalyst and AMP PBC with Nvidia and AMD Ventures joining as strategic investors, Reuters reported–a striking signal of where xAI’s departed talent, and the capital chasing it, ultimately landed. By February 10, 2026, Yahoo Finance counted five of xAI’s twelve founding members gone, naming both Babuschkin and Szegedy among them–a tally roughly in step with AI Wiki’s own tracker, which had already logged Babuschkin’s August 2025 exit as leaving 10 co-founders in place at the time. Kyle Kosic, a former OpenAI engineer, Greg Yang from Microsoft Research, and Toby Pohlen, who had spent six years at Google, all departed in the weeks that followed–AI Wiki dates Yang’s exit to January 2026, thinning the roster to 8, and Pohlen’s to February 2026, down to 5. The final two holdouts–Manuel Kroiss, who led xAI’s pretraining team and had previously worked at DeepMind and Google, and Ross Nordeen, Musk’s so-called “right-hand operator” who had served as a technical program manager at Tesla’s supercomputing division–left on March 27, 2026, though AI Wiki’s tracker, in its August 2026 update, still listed both men as active co-founders as of March 2026–the lone unresolved discrepancy in an otherwise consistent record of the exodus.
“The departures are not ordinary startup attrition,” noted The Next Web in its analysis. “The researchers Musk assembled in 2023 were among the most accomplished in artificial intelligence.” The complete roster of departures–spanning expertise in pretraining, optimization, reasoning, and infrastructure–represents a brain drain of historic proportions in the AI industry.
Complete Timeline of xAI Co-Founder Departures
| Co-Founder | Previous Role | xAI Role | Departure Date | Context |
|---|---|---|---|---|
| Christian Szegedy | Google (12 years) | Research Scientist | Early 2025 | Early departure signal |
| Yuhuai “Tony” Wu | Google, DeepMind, OpenAI | Research Scientist | February 10, 2026 | Post-acquisition tensions |
| Jimmy Ba | University of Toronto (Hinton lab) | Grok Training Lead | February 11, 2026 | Model performance disputes |
| Igor Babuschkin | DeepMind, OpenAI | Chief Engineer | February 2026 | SpaceX restructuring |
| Guodong Zhang | DeepMind, Google Brain | Imagine Team Lead | Early March 2026 | SpaceX/Tesla audit |
| Zihang Dai | Google Research | Research Scientist | Early March 2026 | SpaceX/Tesla audit |
| Greg Yang | Microsoft Research | Research Scientist | Mid-March 2026 | Cultural realignment |
| Toby Pohlen | Google (6 years) | Research Engineer | Mid-March 2026 | Cultural realignment |
| Kyle Kosic | OpenAI | Engineer | Mid-March 2026 | Organizational restructuring |
| Manuel Kroiss | DeepMind, Google | Pretraining Team Lead | Late March 2026 | Final departures |
| Ross Nordeen | Tesla Supercomputing | Operations Lead | March 27, 2026 | Final departures |
The SpaceX Acquisition That Triggered the Exodus
The catalyst for the complete co-founder departure was SpaceX’s acquisition of xAI, which closed on February 2, 2026. The all-stock transaction valued xAI at $250 billion and SpaceX at $1 trillion, creating a combined entity worth approximately $1.25 trillion–the largest corporate merger by valuation in history. The deal brought xAI, X (formerly Twitter), and SpaceX under a single corporate umbrella, fulfilling Musk’s long-stated vision of an integrated technology conglomerate.
The acquisition came just 16 months after xAI’s $6 billion Series B funding round in September 2024, which valued the company at $24 billion–and only weeks after xAI had closed a much larger raise of its own. CNBC reported in November 2025 that xAI was expected to close a $15 billion round at a $230 billion pre-money valuation by December 2025; the round ultimately grew further, and Reuters reported on January 6, 2026 that xAI had completed an upsized $20 billion Series E, exceeding its original $15 billion target, with Nvidia and Cisco Investments joining as strategic backers, according to xAI and CNBC. The leap from that $20 billion Series E to a $250 billion acquisition valuation just weeks later still represented a more than tenfold increase from the 2024 Series B, reflecting both the explosive growth in AI infrastructure spending and the strategic value of xAI’s Colossus supercomputer and Grok model ecosystem.
But the financial terms masked deep cultural tensions. During an all-hands meeting on February 11, 2026, Musk confirmed a thorough reorganization of xAI. The announcement, which came one day after Tony Wu’s departure and the same day as Jimmy Ba’s, signaled that xAI would transition from a research-heavy AI lab to what internal documents described as a “modular, product-driven organization” aligned with SpaceX’s operational philosophy.
“The fluid, exploratory nature of AI development is being replaced by the rigid, milestone-based metrics used in SpaceX’s Starship and Starlink programs,” one xAI engineer told SatNews in February. The cultural clash was fundamental: xAI’s founding team came from academic and research-oriented environments at DeepMind, Google, and the University of Toronto, where exploration and iteration were valued. SpaceX’s culture, by contrast, is famously demanding, with aggressive timelines and what employees have described as “military-grade” operational intensity.
Musk himself acknowledged the friction in remarkably candid terms. He described xAI as having been “not built right the first time around,” adding that the company was “being rebuilt from the foundations up.” For co-founders who had spent nearly three years building what they believed was the right approach to artificial general intelligence, the statement was both a verdict on their work and a signal that their vision was no longer welcome.
The $18 Billion Colossus Supercomputer at the Heart of the Conflict
At the center of xAI’s value proposition–and the tensions driving the co-founder exodus–sits Colossus, the world’s largest AI supercomputer. Located in South Memphis, Tennessee, the facility represents an infrastructure investment of staggering proportions and ambition that continues to expand even as the researchers who were meant to use it depart.
Colossus was originally built in just 122 days, an achievement that Musk frequently cited as evidence of xAI’s ability to move faster than competitors. The initial configuration housed 100,000 Nvidia GPUs drawing approximately 150 megawatts of power. Within 92 days, the facility doubled to 200,000 GPUs, incorporating 50,000 of Nvidia’s H200 units with 141GB of HBM3e memory, pushing power consumption to 250 megawatts.
By early 2026, Colossus 1 alone housed over 230,000 GPUs, including more than 30,000 of Nvidia’s GB200 Blackwell chips. But the real scale of Musk’s ambition became clear in January 2026, when xAI announced the acquisition of a third Memphis building, expanding the total facility to 555,000 Nvidia GPUs across 2 gigawatts of power capacity–enough to power approximately 1.5 million homes. The GPU investment alone for this expanded facility is estimated at $18 billion.
The infrastructure continues to grow. On March 5, 2026, Teslarati reported that xAI filed plans for a $659 million building expansion at the Memphis site, targeting a 79-acre parcel at 5414 Tulane Road. The facility uses direct liquid cooling systems that pipe coolant directly to chips, enabling extreme hardware density, along with Tesla Megapacks for energy storage and power stability. Rather than relying on traditional grid infrastructure, xAI is constructing an on-site gas-fired power plant to provide self-contained power generation for the full 2-gigawatt load.
The paradox is striking: xAI is investing billions in expanding the physical infrastructure needed to train frontier AI models while simultaneously losing every researcher who understood how to use that infrastructure most effectively. As one industry observer noted, “You can buy GPUs. You cannot buy the institutional knowledge of how to train a model that competes with GPT-5 or Claude Opus.”
Colossus Supercomputer: Technical Specifications and Growth
| Phase | Date | GPU Count | Power Capacity | Key Hardware | Estimated Cost |
|---|---|---|---|---|---|
| Colossus 1 (Initial) | September 2024 | 100,000 | 150 MW | Nvidia H100 | ~$4B |
| Colossus 1 (Doubled) | December 2024 | 200,000 | 250 MW | H100 + 50K H200 | ~$8B |
| Colossus 1 (Upgraded) | Early 2026 | 230,000+ | 500 MW | H200 + 30K GB200 Blackwell | ~$10B |
| Colossus 2 (Full Expansion) | January 2026 | 555,000 | 2 GW | H200, GB200, Blackwell Ultra | ~$18B |
| Planned Expansion | March 2026 filing | TBD | 2+ GW | Next-gen Nvidia GPUs | $659M (building only) |
Grok’s Uncertain Future Without Its Architects
The co-founder exodus raises immediate questions about the competitive trajectory of Grok, xAI’s flagship large language model and the product most directly built by the departing researchers. Jimmy Ba, who co-authored the Adam optimizer–one of the most widely used optimization algorithms in deep learning–led Grok’s training pipeline with approximately 900 AI tutors focused on improving performance. Manuel Kroiss directed the pretraining team responsible for developing foundational model capabilities. Igor Babuschkin, as chief engineer, oversaw the technical architecture that connected Colossus’s massive GPU array to the model training process.
With all three gone, xAI must rebuild its core AI research capability from scratch while simultaneously competing against OpenAI’s GPT-5.4, Anthropic’s Claude Opus 4, Google’s Gemini 3.1, and DeepSeek V4–each of which benefits from stable, deeply experienced research teams. The competitive disadvantage is not merely about raw compute power, where xAI’s Colossus gives it an edge, but about the tacit knowledge and research intuition that takes years to develop.
Dan Ives, a senior analyst at Wedbush Securities who has closely followed Musk’s technology ventures, has noted the risks of leadership instability at AI companies: “In AI, talent is the moat. You can replicate infrastructure with enough capital, but you cannot replicate the research intuition that comes from years of working at the frontier. xAI losing its entire founding team is unprecedented and creates real execution risk for Grok’s roadmap.”
The challenge is compounded by the fact that many departed co-founders brought expertise from xAI’s direct competitors. Greg Yang’s work at Microsoft Research on tensor programs and infinite-width neural networks influenced the mathematical foundations of modern language models. Guodong Zhang’s experience at DeepMind and Google Brain gave him deep familiarity with the training techniques used by Gemini. Tony Wu’s background spanning Google, DeepMind, and OpenAI made him one of the few researchers with cross-pollinated knowledge of multiple frontier model architectures. That collective expertise is now dispersed across the industry–or potentially flowing back to competitors.
The SpaceX Culture Clash: Engineering Discipline vs. Research Freedom
Understanding why all 11 co-founders left requires understanding the fundamental cultural divide between SpaceX and a research-oriented AI lab. SpaceX has built its reputation on aggressive timelines, hardware iteration, and what CEO Gwynne Shotwell has described as a “test, fail, fix, fly” approach to rocket engineering. This methodology has produced remarkable results in aerospace: reusable rockets, the Starlink constellation, and rapid Starship development.
But AI research operates on fundamentally different principles. Training a frontier language model is not like building a rocket engine. The feedback cycles are longer, the failure modes are subtler, and the path from hypothesis to result often requires the kind of open-ended exploration that milestone-based management tends to discourage. When SpaceX’s audit teams arrived at xAI in early March 2026, the clash between these two cultures became untenable for many researchers.
“AI research is not a factory floor,” said Yolanda Gil, a professor of computer science at the University of Southern California and former president of the Association for the Advancement of Artificial Intelligence. “The breakthroughs that made modern language models possible–attention mechanisms, scaling laws, RLHF–came from environments that tolerated uncertainty and allowed researchers to pursue unexpected directions. Imposing rigid engineering discipline on that process risks killing the creativity that produces innovation.”
The tension echoes a broader debate in the AI industry about how to manage increasingly expensive research programs. OpenAI has faced similar questions as its $110 billion funding round comes with investor expectations for product delivery. Anthropic balances safety research with commercial pressure. Google DeepMind navigates the competing demands of Alphabet’s product teams and its research mission. But none of these companies has experienced the kind of wholesale leadership turnover that xAI is now confronting.
Margaret Mitchell, chief ethics scientist at Hugging Face and a former Google AI researcher, observed: “What’s happening at xAI is a cautionary tale about treating AI research as a pure engineering problem. The best AI labs maintain a balance between structured product development and open research. When that balance tips too far in either direction, you lose people.”
How the xAI Exodus Compares to Other Tech Talent Crises
The complete departure of xAI’s founding research team is without precise historical precedent, but it invites comparison to other significant talent losses in tech history. When OpenAI’s co-founders began departing in 2020-2023–including Ilya Sutskever, who left after the November 2023 board crisis–the company retained enough institutional continuity to maintain its research trajectory. At Google, the departures of AI ethics researchers Timnit Gebru and Margaret Mitchell in 2020-2021 were disruptive but did not affect the core model development teams.
The closest parallel may be the dissolution of Bell Labs’ research division through a series of AT&T corporate restructurings in the 1990s and 2000s. Like xAI, Bell Labs possessed world-class researchers whose work was disrupted by organizational changes driven by business strategy rather than research considerations. The result was a dispersal of talent that benefited the broader industry–alumni went on to make foundational contributions at Google, Microsoft, and universities–while the original institution never recovered its research prominence.
Erik Brynjolfsson, director of the Stanford Digital Economy Lab, has studied how talent mobility shapes technology markets. “When an entire founding team leaves a company, the intellectual capital doesn’t disappear–it redistributes,” he noted. “The question for xAI is whether it can rebuild before that redistributed talent strengthens its competitors. In AI, where model development cycles are measured in months rather than years, the window is narrow.”
The xAI departures also reflect a broader pattern of consolidation-driven talent disruption in the AI industry. As the agentic AI market grows toward $9 billion, the competition for experienced AI researchers has intensified. Companies that lose key researchers often find those individuals commanding significantly higher compensation packages at competitors–or launching their own ventures with the knowledge gained at their previous employer.
Market Impact: What the Departures Mean for the AI Industry
The xAI co-founder exodus has implications that extend well beyond Musk’s company. The 11 departed researchers collectively represent expertise in pretraining, optimization, reasoning, multimodal AI, and large-scale distributed computing–skills that are in extreme demand across the industry. Where they land will reshape competitive dynamics.
Several co-founders are rumored to be exploring new ventures. Jimmy Ba, whose Adam optimizer is used in virtually every major AI training run globally, and Greg Yang, whose theoretical work on tensor programs has influenced scaling law research, are reportedly in discussions about founding a new AI research lab. If confirmed, such a venture would immediately attract significant venture capital interest given the founders’ credentials and track records.
For the broader AI chip market, the xAI situation creates uncertainty around one of Nvidia’s largest customers. The Colossus facility, with its 555,000 GPUs and ongoing expansion plans, represents one of the most significant single-site deployments of Nvidia’s Blackwell architecture. If leadership instability slows xAI’s model development, it could affect the pace of future GPU procurement–though the facility’s value as compute infrastructure means it would likely be utilized regardless of research direction changes.
SpaceX’s planned IPO, reportedly targeting mid-2026 with a potential $1.75 trillion valuation, adds another dimension. The xAI acquisition was partly designed to boost SpaceX’s AI narrative ahead of the public offering. A perception of leadership instability at the recently acquired AI division could complicate the IPO story, though SpaceX’s core aerospace business remains strong.
James Wang, a technology analyst at ARK Invest who covers AI infrastructure, commented: “The market will watch closely to see how quickly xAI can rebuild its research team. The Colossus supercomputer is a formidable asset, but without the right people directing its use, it’s an expensive collection of silicon. The gap between having compute and having competitive AI models is filled entirely by human talent.”
Musk’s Strategy: Rebuilding xAI Under SpaceX’s Model
Despite the departures, Musk appears committed to his strategy of rebuilding xAI from the ground up under SpaceX’s operational framework. The reorganization has restructured xAI into four distinct teams, each with clear deliverables and timelines–a departure from the more fluid organizational structure favored by the departed co-founders.
The immediate strategic priorities, according to reporting from Entrepreneur Loop and SatNews, include the launch of “X Money,” a financial services product integrated with the X platform, and the integration of Grok into SpaceX’s Starlink V3 satellite constellation. The latter project, described internally as the “Orbital Data Center” strategy, would use Starlink’s low-latency global network to provide edge computing capabilities for Grok–a technically ambitious goal that plays to SpaceX’s strengths in satellite infrastructure.
Musk’s approach mirrors his playbook at other companies. At Tesla, he replaced founding-era executives with operational leaders who could scale manufacturing. At X (formerly Twitter), he gutted the workforce by approximately 80% and rebuilt with a smaller, more focused team. In both cases, the transitions were chaotic in the short term but ultimately produced organizations more aligned with Musk’s vision–though critics argue the quality of output suffered, particularly at X.
The key difference with xAI is the competitive environment. Tesla competed against legacy automakers that were slow to adapt. X operates in a social media market where switching costs keep users on the platform regardless of workforce changes. AI, by contrast, is a talent-driven industry where the gap between leaders and followers is determined largely by research capability. Without its founding researchers, xAI must compete against OpenAI’s expanding model lineup, Anthropic’s safety-focused approach, and Google’s massive research apparatus–all of which have stable leadership teams.
The Broader AI Talent War: Context and Competition
The xAI exodus must be understood within the context of the most intense talent war in technology history. AI researchers with frontier model experience command compensation packages that would have been unthinkable five years ago. Senior research scientists at leading AI labs routinely earn $1-5 million annually, with equity packages that can push total compensation significantly higher.
The supply of such talent remains severely constrained. According to industry estimates, fewer than 1,000 researchers globally possess the specific combination of skills needed to lead frontier model development–expertise in distributed systems, optimization theory, scaling laws, and the empirical intuition that comes from hands-on experience training models at the largest scales. The xAI co-founders collectively represented approximately 1% of this global talent pool.
The broader tech industry’s 2026 workforce restructuring has paradoxically tightened the AI talent market further. While companies have cut headcount in non-AI roles, they have aggressively expanded AI research teams. Microsoft, Google, Amazon, and Meta have all increased AI research hiring in 2025-2026, even as they reduced overall employee counts.
This dynamic means that xAI’s departed co-founders will have no shortage of options. The question is whether Musk can offer a compelling alternative narrative–access to the world’s largest supercomputer, the integration with SpaceX’s satellite network, and the ambition of building artificial general intelligence–that can attract a new generation of researchers willing to work within SpaceX’s demanding culture.
What Comes Next: Five Predictions for xAI and the AI Industry
The xAI co-founder exodus will reshape the AI landscape in several predictable–and some unpredictable–ways. Based on the available evidence, historical precedent, and the competitive dynamics of the AI industry, here are five predictions for what comes next.
Prediction 1: At least two xAI co-founders will launch a new AI company by Q3 2026. The combination of deep expertise, industry reputation, and an abundant venture capital market for AI startups makes this near-certain. A venture founded by Jimmy Ba and Greg Yang, in particular, would attract immediate attention given their respective contributions to optimization and scaling theory.
Prediction 2: Grok will fall behind in frontier model benchmarks within 6-12 months. Without its core research team, xAI will struggle to maintain the pace of model improvements needed to compete with GPT-5.4, Claude Opus 4, and Gemini 3.1. The Colossus supercomputer provides raw compute, but the research direction and training methodology that determine model quality require deep expertise that cannot be replaced quickly–though the pipeline did not stall entirely: xAI’s own documentation shows new grok-2-1212 and grok-2-vision-1212 models shipping in August 2026 with improved accuracy and multilingual capabilities, and a 2026 model tracker credited xAI with releasing its flagship Grok 4.6 on August 12, 2026, complete with a 500,000-token context window, suggesting the post-exodus team retained enough capability to keep shipping even as the underlying research bench thinned.
Prediction 3: SpaceX’s Grok-Starlink integration will become xAI’s primary differentiator. Unable to compete purely on model quality, xAI will lean into distribution advantages–specifically, the ability to embed Grok into Starlink’s global satellite network for edge AI applications. This plays to SpaceX’s strengths and creates a use case that competitors cannot easily replicate.
Prediction 4: The xAI talent redistribution will strengthen at least two competing AI labs. The expertise of the departed co-founders is too valuable to remain idle. Whether through new ventures or recruitment by existing labs, the knowledge gained at xAI–particularly around training models on Colossus-scale infrastructure–will benefit the broader AI ecosystem.
Prediction 5: The SpaceX IPO narrative will pivot away from xAI’s research capabilities. As the mid-2026 IPO window approaches, SpaceX will likely frame xAI as an infrastructure and distribution play rather than a frontier research lab, emphasizing Colossus’s value as a compute asset and Starlink integration rather than Grok’s benchmark performance.
The Bigger Picture: Can Money Replace Talent in AI?
The xAI co-founder exodus crystallizes a question that the entire AI industry is grappling with: in a field where infrastructure costs have reached $18 billion for a single facility, does capital or talent matter more? Musk’s bet is implicitly on capital–the belief that with enough GPUs, enough power, and enough organizational discipline, xAI can rebuild its research capability and compete at the frontier.
History suggests otherwise. The most successful AI labs–OpenAI, DeepMind, Anthropic–have been built around small teams of exceptional researchers who made outsized contributions to model architecture, training methodology, and alignment. Google, despite having more compute resources than any competitor, fell behind in the large language model race in 2022-2023 because organizational factors slowed its ability to ship frontier models. The lesson: compute is necessary but not sufficient. The difference between a good model and a great one lives in the decisions made by the people designing the training runs.
For xAI, the path forward is narrow but not impossible. If Musk can attract a new cohort of top-tier researchers–perhaps from the growing pool of AI talent in China, Europe, or from smaller labs being squeezed by the escalating costs of frontier research–Colossus’s unmatched compute could become an irresistible draw. The world’s largest supercomputer is, after all, a tool that any AI researcher would want access to. The question is whether the cultural environment surrounding that tool is one where the best researchers can do their best work.
Related Coverage
More on AI Industry Developments
- GPT-5.4 vs Claude Opus 4.6 vs DeepSeek V4 vs Gemini 3.1: The Top AI Comparison (March 2026)
- Big Tech’s $700 Billion AI Infrastructure Bet: Inside the 2026 Spending Race
- OpenAI’s $110 Billion Funding Round: Inside the Largest Private Investment in History
- Agentic AI in Enterprise 2026: Inside the $9 Billion Market Reshaping How Businesses Operate
- NVIDIA Blackwell GPU Pricing: B200, B300 and DGX Cost Breakdown
- Tech Layoffs 2026: How AI Is Driving the Biggest Workforce Shakeup in a Decade
- The AI Data Center Power Crisis: How Big Tech’s 125 GW Appetite Is Reshaping the US Energy Grid
Frequently Asked Questions
Why did all xAI co-founders leave?
All 12 xAI co-founders departed, with the exodus accelerating following SpaceX’s acquisition of xAI in February 2026–Reuters reported that just six of the twelve founding researchers remained as of February 10, 2026, the day Tony Wu and Jimmy Ba both resigned. The departures were driven by cultural clashes between xAI’s research-oriented origins and SpaceX’s milestone-based engineering culture, disputes over model development strategy, and a thorough organizational restructuring that fundamentally changed how xAI operated. Elon Musk himself acknowledged the friction, stating that xAI was “not built right the first time around” and was “being rebuilt from the foundations up.”
What is xAI’s Colossus supercomputer?
Colossus is the world’s largest AI supercomputer, located in South Memphis, Tennessee. As of early 2026, the facility houses 555,000 Nvidia GPUs across 2 gigawatts of power capacity, with a GPU investment estimated at $18 billion. The facility was originally built in 122 days and uses direct liquid cooling, Tesla Megapacks for energy storage, and an on-site gas-fired power plant. A $659 million expansion was filed in March 2026.
How does the xAI exodus affect Grok?
The departure of all 11 co-founders, including the chief engineer, pretraining team lead, and Grok training lead, creates significant execution risk for Grok’s development roadmap. Without the researchers who built the model’s architecture and training pipeline, xAI must rebuild its core AI research capability while competing against OpenAI, Anthropic, Google, and DeepSeek–all of which have stable research leadership.
What is SpaceX’s plan for xAI after the acquisition?
SpaceX plans to integrate xAI into its broader technology ecosystem. Key priorities include launching “X Money” (a financial services product), integrating Grok into the Starlink V3 satellite constellation for low-latency edge AI, and expanding the Colossus supercomputer. The company has restructured xAI into four distinct teams with clear deliverables aligned to SpaceX’s operational model.
Who were the xAI co-founders?
The 11 co-founders were: Igor Babuschkin (chief engineer, ex-DeepMind/OpenAI), Jimmy Ba (ex-University of Toronto/Hinton lab), Christian Szegedy (ex-Google), Manuel Kroiss (ex-DeepMind/Google), Ross Nordeen (ex-Tesla), Yuhuai “Tony” Wu (ex-Google/DeepMind/OpenAI), Greg Yang (ex-Microsoft Research), Toby Pohlen (ex-Google), Zihang Dai (ex-Google), Guodong Zhang (ex-DeepMind/Google Brain), and Kyle Kosic (ex-OpenAI). They were recruited from the world’s leading AI research organizations.
How much was xAI valued at when SpaceX acquired it?
SpaceX acquired xAI in an all-stock transaction that valued xAI at $250 billion and SpaceX at $1 trillion, creating a combined entity worth approximately $1.25 trillion. This represented a more than tenfold increase from xAI’s $24 billion valuation during its $6 billion Series B funding round in September 2024.
Will xAI be able to compete without its founding team?
xAI faces significant challenges but retains key assets: the world’s largest supercomputer (555,000 GPUs), integration with SpaceX’s Starlink network, and access to X platform data. However, rebuilding frontier AI research capability requires attracting top-tier talent willing to work within SpaceX’s demanding culture. Industry analysts suggest xAI may pivot from pure research competition to differentiation through infrastructure and distribution advantages.
