NVDA $50B bond oversubscribed: SOXX RSI 78, risk decoded

NVIDIA's massive $50 billion bond offering on June 15, 2026, signaled both the confidence of the market in the company's AI dominance and the underlying fragility of the chip-capital-intensity thesis. The bonds were immediately oversubscribed, meaning demand exceeded supply by a wide margin. However, this success masks a broader structural risk: the AI capex cycle is being financed at an unprecedented pace, and credit markets are showing signs of strain.

On the same day, STMicroelectronics raised $1.5 billion in convertible bonds, and companies across the semiconductor supply chain issued $40+ billion in debt overall. Man Group, a major hedge fund operator, explicitly warned of "bubble risks" as bond sales boom in the race to build out artificial intelligence infrastructure. The velocity and scale of debt issuance suggest that markets are front-running capex decisions, with cheap financing available before potential recession concerns or rate-hike scenarios re-emerge.

The technical picture in semiconductors is stretched: the iShares Semiconductor ETFExchange-Traded Fund - a basket of securities trading like a single stock. (SOXX) reached an RSIRelative Strength Index - momentum oscillator on a 0-100 scale. (Relative Strength Index) of 78, a level typically associated with overbought conditions and potential near-term mean reversion. NVIDIA, Broadcom (AVGO), AMD, Applied Materials (AMAT), and Lam Research (LRCX) have all benefited from the capex frenzy, but valuations have compressed available margin-of-safety. Analysts note that unless capex growth sustains at current levels for 3-5 years, valuations will face significant downside.

The bubble debate centers on whether AI demand will justify the capex spend. Proponents point to long-dated contracts, power-grid constraints forcing data-center construction, and secular demand for inference chips. Sceptics counter that many of the bonds being issued are financing infrastructure that may not generate positive returns for 5-7 years, exposing investors to refinancing risk and potential credit events if the AI capex cycle stalls.