The series is derived, and the derivation is simple: the average yield on CCC-rated corporate bonds minus the average yield on AAA-rated ones, both all-sector rating buckets from the Federal Reserve data named in the source line above, plotted daily over the last three years. Because both sides are yields taken on the same day, the gap strips out the level of interest rates and isolates the risk premium itself.
The gap is quoted in percentage points: how much more a CCC-rated borrower pays than a AAA-rated one for the same money. A flat or narrowing gap means credit is cheap for everyone; a widening one means the market is repricing risk from the bottom up, borrower by borrower.
Two limitations, stated plainly. This is not a technology-sector measure: the buckets cover all sectors, and they work as a proxy only because AAA is roughly where the cash-rich hyperscalers borrow and CCC is roughly where the most leveraged datacentre builders borrow. And it is not a credit default swap — single-name CDS quotes are licensed data and cannot be republished. A credit scare anywhere else in the economy can move this series without anything changing in AI, so read it alongside the builders' own filings, not alone.
That repricing is the transmission channel into US equities. The AI build-out runs on continued access to borrowed money — tracked company by company on the capex-vs-debt page — and a widening spread raises the cost of exactly that money. A widening gap is the financing crack arriving in the data before it arrives in the headlines; a narrowing one is the credit market saying the build-out can keep rolling.
Nothing here is a forecast: the chart shows the premium itself, daily, not a probability of default or a timing. Pair it with the builders' leverage to see both sides — who owes, and what the market charges for owing.