Elon Musk Is Rebuilding xAI: What Went Wrong and the Promise (and Peril) of ‘Orbital Data Centers’

Elon Musk admits xAI “wasn’t built right” as SpaceX pivots to orbital data centres

Elon Musk says he is rebuilding his AI startup xAI from the ground up, just weeks after SpaceX acquired the company. In a post on X, Musk drew a line under the first chapter of xAI and signalled a new direction that leans heavily on SpaceX’s capabilities.

“xAI was not built right first time around, so is being rebuilt from the foundations up.”

According to the Fortune report shared on Reddit, the stated purpose of the SpaceX acquisition is to build “orbital data centers” – Musk’s proposed route to cheaper AI compute. Yet on the ground, xAI is facing a classic startup problem: people. Nine of the original 11 cofounders (not including Musk) have reportedly left since 2024, alongside a wave of senior engineers. You can read the discussion thread here: Reddit post.

SpaceX acquisition and the promise of “orbital data centres”

The merger of xAI into SpaceX is framed as a way to unlock new infrastructure: data centres in orbit to power AI training and inference. In plain terms, that means putting compute clusters on satellites or space platforms, powered by solar energy and linked via high-bandwidth communications.

Why do this? The claimed benefits include lower energy costs (abundant solar), fewer land constraints, and tighter integration with global connectivity. Musk has argued this could be the most cost-effective route to scale AI compute.

What “orbital data centre” actually means (and the jargon)

• Compute: shorthand for the GPU/TPU-class hardware and energy needed to train and run AI models.
• Orbital data centre: a facility placed in Earth orbit rather than on the ground, hosting servers and storage, with power, cooling, and communications provided in space.
• Alignment: ensuring AI systems behave as intended and avoid harmful outputs.

Talent exodus: a strategic and execution risk

Per Fortune’s summary of reporting, two xAI cofounders left in the last week and two more the prior month. That reportedly leaves only two cofounders alongside Musk, and follows a broader departure of senior engineering talent.

For any AI company, leadership continuity and specialised systems engineering are critical. Rebuilding “from the foundations up” is a bold statement – but without a stable core team, timelines and product quality can slip. The move into orbital infrastructure also widens the skillset required, from model research to satellite systems, operations, and compliance.

Reported snapshot

SpaceX acquisition of xAI	Reported by Fortune as recently completed
Stated goal of acquisition	“Orbital data centers” (per Musk’s post, via Fortune)
Cofounders remaining	2 (reported)
Departures since 2024	9 of the original 11 cofounders (reported); plus a dozen senior engineers (reported)
Funding, roadmap, timelines	Not disclosed

Why this matters for UK developers and businesses

Compute scarcity, energy costs, and data centre planning constraints are live issues in the UK. If space-based compute ever becomes practical at scale, it could change the economics and geography of AI capacity. In the near term, though, the UK market needs clarity on availability, interoperability, and compliance.

• Compliance and data protection: UK GDPR expects appropriate safeguards for personal data. Space-based processing raises questions about jurisdiction, cross-border data flows, and vendor responsibilities.
• Reliability and latency: Enterprise adoption depends on predictable performance and SLAs. Any space-ground architecture must meet real-world latency, uptime, and recovery expectations.
• Procurement and vendor risk: Leadership churn can signal product instability. UK teams should adopt a multi-vendor strategy and avoid lock-in where practical.
• Sustainability: Space systems claim abundant solar power, but lifecycle impacts (launch, maintenance, deorbit) need transparent reporting to align with ESG commitments.

Orbital data centres: potential benefits vs. engineering risks

Potential benefits

• Scalable solar power without terrestrial grid constraints.
• Reduced land use and local planning friction.
• Tight integration with satellite connectivity for global services.

Key risks and open questions

• Heat management: rejecting heat in space is fundamentally different from on Earth; this is a known engineering challenge.
• Maintenance and upgrade cycles: on-orbit servicing is complex and costly compared to swapping GPUs in a terrestrial rack.
• Launch, debris, and safety: building and refreshing orbital fleets must account for orbital debris and regulatory oversight.
• Bandwidth economics: high-throughput, low-latency links to ground are essential; the cost and capacity trade-offs need evidence.
• Compliance and data sovereignty: clear guidance is needed on controller/processor obligations when data is processed in orbit.

What to watch next

• Team rebuild: who joins xAI for the second act, and what roles anchor model research, infrastructure, and safety.
• Technical milestones: credible demonstrations of orbital compute prototypes, including power, thermal, and communications performance.
• Product focus: clarity on xAI’s core offering post-rebuild – foundation models, inference APIs, or integrated services with SpaceX connectivity.
• Regulatory engagement: how xAI/SpaceX address data protection and safety concerns in major markets, including the UK.

Practical steps for UK teams today

• Adopt a provider-agnostic architecture: design your AI stack so you can swap models and infrastructure as needed.
• Document data flows: map where data is stored and processed to meet UK GDPR obligations; request DPAs and regional processing options from vendors.
• Focus on near-term productivity: prioritise use cases that deliver value on current infrastructure (RAG, fine-tuning, and workflow automation).
• Prototype with accessible tools: for example, streamline reporting or internal ops with LLMs connected to spreadsheets – see my guide on connecting ChatGPT to Google Sheets.

Bottom line

xAI’s reboot under SpaceX is a dramatic bet that AI’s bottleneck is infrastructure – and that the answer may lie in orbit. The vision is compelling, but success will be defined by execution: rebuilding the team, shipping reliable products, and proving the economics.

For UK organisations, treat this as a strategic signal, not an immediate procurement option. Keep your stack flexible, your data governance tight, and your expectations grounded in what delivers value today while the industry chases tomorrow’s compute.