Chipmaker credit protection now costs more than hyperscaler protection

For the first time, it costs more to insure against chipmakers defaulting than against the hyperscalers that buy their chips. The semiconductor credit default swap basket now trades at wider spreads than the hyperscaler basket.

In credit markets, that ordering usually tells you who looks shakier. Credit traders are effectively saying the arms dealers of the AI boom carry more risk than their biggest customers.

What the numbers show

First, a quick primer. A credit default swap works like an insurance policy on a company’s debt. The buyer pays a recurring premium, and the seller pays out if the company defaults.

That premium is quoted as a spread in basis points, where one basis point equals one hundredth of a percentage point. A wider spread means pricier protection, which means the market sees more risk.

Barclays data on its AI portfolio shows five-year CDS levels as of October 28, 2026. On the hyperscaler side, Microsoft sat at 53 bps, Alphabet at 80 bps, Amazon at 81 bps and Meta at 114 bps.

The chipmakers priced noticeably higher. Intel came in at 104 bps, AMD at 121 bps, Nvidia at 125 bps and Broadcom at 168 bps.

Every chipmaker on that list costs more to insure than Microsoft, Amazon or Alphabet. Broadcom’s protection runs more than three times the cost of Microsoft’s.

Meta is the outlier among the hyperscalers. At 114 bps, its protection costs more than Intel’s.

The broader AI group is also lagging the market. The AI basket’s underperformance against the CDX.IG investment-grade benchmark widened from around 10 bps six months earlier to 76 bps by October 2026.

Trading volume went vertical

It is not just prices moving. The amount of protection changing hands has exploded.

A late-October 2026 Barclays report found hyperscaler CDS volumes jumped approximately 740% year-on-year to a four-week average of $6 billion. Semiconductor CDS volumes climbed approximately 1,900% year-on-year to $3 billion.

Weekly CDS trading volumes for AI-related firms reached about $9 billion in September 2026. In July, that figure was under $2 billion.

Part of the volume surge reflects hedging demand from equity and multi-asset managers. These investors hold large positions in AI stocks and are using credit protection as a way to guard against a downturn without selling their shares.

Why the chipmakers flipped

The backdrop is a borrowing spree. AI-related firms have issued approximately $466 billion of debt in 2026 to date, compared with $101 billion over the same stretch the previous year.

Much of that money is funding AI infrastructure: data centers, power and the chips that fill them. Debt-financed capital spending, rising leverage and negative free cash flow at several firms are all feeding into wider spreads.

Chipmakers face a specific pressure point: vendor financing. When a supplier helps fund its own customers’ purchases, it takes on some of their credit risk alongside the sale.

That links balance sheets together. Interlinked financing structures mean trouble at one node can travel along the chain, and credit markets tend to price that kind of contagion risk early.

Oracle offers a preview of how far things can stretch. Its CDS spreads hit record levels near 257 to 305 bps in recent months, and certain semiconductor names have also seen particularly sharp moves.

Widening spreads at these levels point to potential ratings pressure. Some AI-linked names are trading near or into weaker investment-grade or high-yield territory.

What this means

For equity holders, credit markets often act as an early warning system. Bond and CDS traders focus on downside protection rather than upside, so persistent widening can flag stress before it shows up in stock prices.

The things to watch are straightforward. Watch whether the AI basket’s 76 bps gap to CDX.IG keeps growing, whether issuance continues at this year’s pace, and whether rating agencies act on the pressure that spreads are already signaling.

Also watch the relationship between the two baskets themselves. A single crossover can be noise, but a sustained period with chipmakers priced as riskier than hyperscalers would mark a real repricing of where risk sits in the AI supply chain.