In a move that underscores the breathtaking capital intensity of the global artificial intelligence arms race, SpaceX is preparing to execute one of the largest corporate hardware acquisitions in tech history. According to reports first disclosed by the Financial Times and confirmed by CNBC, private equity powerhouse Apollo Global Management and a consortium of major investment banks are in preliminary discussions to structure a $40 billion debt package for SpaceX.
The proceeds from the debt vehicle are earmarked for a massive purchase of high-performance Nvidia graphics processing units (GPUs), cementing SpaceX transition from a pure-play aerospace launch contractor into one of the world largest commercial AI compute and data center operators.
Anatomy of the $40 billion financing package
The proposed financing structure represents an increasingly popular Wall Street mechanism: leveraging the physical microprocessors themselves as primary collateral for investment-grade debt. The total capital target is $40 billion USD, with approximately $30 billion in high-grade corporate bonds managed by Apollo alongside fixed-income institutional investors including Pimco, paired with $10 billion in syndicated bank loans.
The collateral anchor is the high-value Nvidia GPU clusters themselves, taking advantage of persistent secondary-market demand for enterprise AI silicon. The target closing window is 2027, aligning with Nvidia production and delivery schedule for its next-generation Blackwell and Rubin server architectures.
Expanding Colossus | Why SpaceX needs $40 billion in chips
The decision to pour $40 billion into semiconductor hardware follows the structural consolidation of Elon Musk technology ventures. Since merging xAI into SpaceX under the unified SpaceXAI operational canopy, the company has taken over primary development of the massive Colossus supercomputing hub in Memphis, Tennessee.
By expanding its physical server footprint beyond 200,000 active Nvidia chips, SpaceX is building out compute capacity to handle three distinct operational workloads. First, autonomous orbital and spacecraft telemetry, training real-time inference models to manage satellite routing across Starlink expanding LEO constellation. Second, SpaceXAI model training, underwriting the compute demands for next-generation multimodal models and autonomous coding tools. Third, neocloud compute leasing, renting dedicated high-density GPU clusters directly to third-party AI developers, a revenue stream that already includes major infrastructure leases with frontier labs like Anthropic.
The neocloud leasing model mirrors the deepening financial ties between Nvidia and frontier labs like Anthropic, where chip supply and equity stakes are increasingly intertwined. For SpaceX, the GPU fleet becomes both an internal compute asset and a revenue-generating infrastructure business.
Market reaction | Split sentiments on Wall Street
The news triggered an immediate split-screen reaction across global capital markets. SpaceX shares trimmed roughly 1.2% in postmarket trading to $169.79. While investors acknowledge the massive recurring revenue potential of data center leasing, bond markets remain cautious about the rapid debt accumulation required to maintain mega-scale infrastructure.
Nvidia stock edged higher in after-hours trading. For Nvidia, a single $40 billion purchase order reinforces the longevity of its enterprise order books, confirming that major hyper-scalers are willing to issue multi-billion-dollar debt packages to secure chip allocations through 2027 and beyond.
The deal also signals how the governance debate over AI infrastructure is shifting from policy to balance sheets, as credit markets now underwrite the physical compute layer of the AI economy. The asset-backed structure echoes the escalating rivalry between Musk ventures and OpenAI, where capital deployment is now part of the competitive playbook. The same dynamic is visible in the agentic AI race between OpenAI and its rivals, where compute access is the binding constraint.