The News: A Credit Risk Spike at the Edge of the AI Boom
According to the Financial Times, credit risk indicators tied to SpaceX have jumped sharply amid concerns over the company's expanding borrowing. The alarm comes as Bloomberg reports that SpaceX is in talks to borrow as much as $40 billion, largely to fund purchases of Nvidia chips for AI infrastructure build-out unrelated to its core launch business. This is a company best known for rockets taking on tens of billions in debt to buy semiconductors — a detail that should give any credit analyst pause.
The arrangement is not unique. Across the AI infrastructure complex, a layered structure of vendor financing, chip-backed loans, and circular investment commitments between chipmakers, cloud providers, and compute buyers has become standard practice. Nvidia invests in companies that buy its chips; those companies borrow against future contracts to fund the purchases; the debt sits across multiple balance sheets, private credit vehicles, and special purpose entities, with no single regulator holding a complete picture of aggregate exposure. That opacity is the detail worth dwelling on, because it is structurally identical to a problem regulators in Beijing failed to solve eleven years ago.
Today's AI financing stack has the same two-tier character. Bank loans and public bond issuance are visible, rated, and disclosed. But vendor financing, chip-as-collateral lending, and circular equity-and-supply agreements between a handful of interlinked firms are harder to size from the outside. The dot-com era's infrastructure overbuild offers one comparison; China's 2015 shadow margin structure offers a sharper one, because it demonstrates what happens when a credit structure's true size is unknown until the moment it must be repaid.
Why the Structure, Not the Sector, Matters
Bubbles are rarely defined by the asset class alone — rockets, shares, chips, housing — but by the credit structure underneath. A market can absorb significant equity-funded overbuild without systemic damage. What it cannot absorb without damage is debt-funded overbuild where the true leverage ratio is distributed across entities that do not report to a common regulator. For investors assessing current equity valuation and market risk, the AI infrastructure financing chain deserves the same scrutiny once reserved for China's shadow banking sector, a dynamic explored in more detail through macro context on historical credit cycles. The parallel is not that SpaceX or Nvidia will collapse — it is that nobody currently has a consolidated figure for how much debt sits behind the AI build-out, which is exactly the condition that preceded China's 2015 unwind.
Sentiment indicators are flashing too. Bloomberg reported that investor Mike Novogratz has joined a growing chorus describing AI as a bubble, while simultaneously advising investors to keep buying into it — a posture common in late-stage manias, where participants acknowledge overvaluation but calculate the unwind is not yet imminent. Chinese retail investors in early 2015 expressed similar reasoning: aware valuations were stretched, yet unwilling to exit a rally endorsed by state media and official policy.
The Historical Parallel
Read the full Chinese Stock Market Bubble (2014-2015) entry in the Bubblepedia Encyclopedia for the complete account of how Beijing's active promotion of equity investment, combined with loosened margin rules and an invisible shadow lending layer, produced one of the fastest boom-bust cycles in modern market history. When the Shanghai Composite cracked, the government intervened directly — suspending trading in over half of listed stocks, directing state funds to buy shares, and banning major shareholders from selling — measures that slowed but did not reverse the decline. The lesson transfers directly: once leverage is distributed and undisclosed, no single intervention point exists to stop a cascade once it begins.
The defining feature of a credit-driven bubble is not the size of the debt but the number of people who can accurately state its total. In China in 2015, regulators themselves underestimated aggregate margin exposure. In 2026's AI infrastructure financing web, the equivalent figure does not yet exist in any single public filing.
Readers tracking these dynamics across sectors can follow ongoing coverage at Bubble Watch and the broader Bubblepedia encyclopedia.