The foundation of xAI is shifting as the company transitions from a nimble research lab into an industrial-scale subsidiary of the Musk empire, with exactly half of its founding team—six out of the original 12 members, excluding Elon Musk, who remains CEO—departing. The exodus highlights a growing friction between the experimental ‘scrappy’ culture of the original core and the industrial-scale demands of the post-merger era. With xAI advancing projects like Grok upgrades amid its formal SpaceX acquisition, these changes prompt scrutiny of talent dynamics in the race for AI leadership.
Foundational Strategy and Initial Leadership
xAI commenced operations in July 2023 under Elon Musk’s vision to rival Big Tech’s AI efforts and pursue a deeper understanding of the universe. The founding team of 12 drew from elite institutions: Igor Babuschkin from DeepMind and OpenAI; Yuhuai (Tony) Wu from Google Brain and Stanford; Jimmy Ba on research and safety; Christian Szegedy from Google Brain; Greg Yang from Microsoft Research; Kyle Kosic on infrastructure; Guodong Hu from Google Research; Toby Pohlen from DeepMind; Ross Nordeen with Tesla engineering experience; Manuel Kroiss from Microsoft; and Zihang Dai from Google. Six remain active, including Pohlen, Nordeen, Kroiss, Dai, and Hu, anchoring continuity amid expansion, while Hang Gao serves as a key engineer rather than the original founder.
This highly specialized assembly drove swift progress, launching Grok-1 as an open-source model with real-time X data integration, distinctive humor, and fewer content restrictions than GPT-4. Early on, xAI’s lean structure fueled innovation, with Grok variants excelling in benchmarks like simulated financial modeling. The emphasis on first-principles thinking and accelerated timelines mirrored Musk’s playbook from Tesla and SpaceX. As headcount surpassed 1,000 and ambitions targeted trillion-parameter models via the “Colossus” supercluster in Mississippi—scaling to 2 gigawatts of compute powered by approximately 555,000 NVIDIA GPUs (a mix of H200s, GB200s, and GB300s)—the foundational group faced mounting strains.
Timeline of Leadership Changes
What began as a slow trickle of exits in late 2024 has transformed into a high-stakes exodus as the company scales. Kyle Kosic left in mid-2024 for OpenAI, Christian Szegedy in February 2025 after building core systems, Igor Babuschkin in August 2025 for a venture firm, and Greg Yang recently citing health issues—four gone by late 2025. February 2026 accelerated: Tony Wu resigned on the 9th, praising small AI-enhanced teams; Jimmy Ba confirmed his exit on the 11th, optimistic about advancements but seeking recalibration.
Several key engineers from the Grok Imagine and ML teams, including Hang Gao, Vahid Kazemi, and Shayan Salehian, departed around then, some eyeing joint ventures. By February 11, reports verified six of 12 founders—precisely half—had left, alongside broader attrition amid the SpaceX merger valuing the combined entity at $1.25 trillion.
Cultural Friction and Technical Hurdles
Farewells mixed gratitude with critiques. Tony Wu favored startup agility; Jimmy Ba invoked the Kardashev scale—gauging civilizations from Type I (harnessing a planet’s energy) to higher levels—and recursive self-improvement, where AI iteratively enhances its own capabilities yet chooses new paths. Kazemi questioned AI homogenization; Salehian valued Musk’s principles but sought fresh challenges. Pressures included boosting Grok against OpenAI’s o1 and Claude amid benchmark gaps, plus reorganization for speed.
Capital-intensive efforts like Colossus drove costs: Q3 2025 posted $1.46 billion USD losses on $107 million revenue, $7.8 billion USD spent over nine months, and $1 billion monthly burn—fueling massive compute for frontier models. In AI’s talent wars, such churn halved leadership during IPO planning.
Musk’s Perspective and Ongoing Resilience
Musk framed exits strategically at a February 10 all-hands, noting some early stages as best, and via X described reorganization as organic evolution for velocity, pledging aggressive hiring. Technical development remains aggressive: Grok handles 2 million-token contexts, shines in real-time tasks, and offers efficient APIs—key for U.S.-based creators in San Francisco exploring cloud AI. Open-sourcing and Colossus propel trillion-parameter aims, with Grok 4.20 targeting early 2026.
This echoes Musk’s restructurings, sustaining valuation post-acquisition.
SpaceX Synergies and Orbital Ambitions
The SpaceX acquisition, valuing the entity at $1.25 trillion, enables orbital data centers to overcome terrestrial power limits. Sun-synchronous satellites promise eightfold solar gains for AI training, launched via Starship with laser networks streaming terabit datasets from space and X. This secures orbital dominance. challenging Amazon’s Kuiper and transforming AI infrastructure tied to Colossus’s gigawatt scale.
Governance Crises and Legal Jeopardy
Scrutiny mounts: French raids hit X’s Paris offices on February 3, 2026, over Grok Imagine’s deepfakes, explicit outputs, and child safety lapses under EU rules, with potential fines up to 6% of global annual turnover under the Digital Services Act. The Paris prosecutor summoned Musk and former X CEO Linda Yaccarino (who served from May 2023 to July 2025) for questioning on April 20. “Unshackled” modes allegedly enabled nonconsensual images, drawing UK probes and testing xAI’s ethos.
Strategic Resilience and IPO Viability
The erosion of the founding core creates a leadership vacuum that xAI must fill quickly if it hopes to maintain its aggressive development velocity, but the ambitious pursuit of orbital compute and Kardashev-scale energy mastery represents a calculated bet that raw infrastructure can outpace the loss of human capital. As xAI nears its public debut, orbital infrastructure may forge a defensible moat against stable rivals.
