GS Papai IG Credit Down the Stretch 7 Sep 2026

GS Papai - IG Credit Down the Stretch 7 Sep 2026 Jeffrey Papai · Goldman Sachs · IG Credit / FICC Mon 7 Sep 2026, 1:25pm ET

As we close an eventful summer in IG credit (especially given the lack index level vol) the path forward feels increasingly complicated. The techncials that have largely dominated this year (AI vs. Non-AI supply/demand) look set to flip in September. At the same time the demand function for IG, which has been bulletproof, potentially has some clouds forming in the horizon with an increasingly uncertain Fed path combined with the potential for macro vol into the midterms. The near-term setup seems to be for a reversal of some of the AI vs. non-AI dispersion, however, would treat such moves as temporary and buy any supply led dip in generic IG and right size risk in the AI factor ahead of an even busier 2027.

• After ~$300bn of AI supply YTD the rest of the year looks to materially slow, especially in hyperscalers. This creates the conditions for short-term relief in AI spreads (GSUCIABK), but only temporarily. • With ~$230bn worth of mostly non-AI IG supply to come in September the rest of IG looks relatively vulnerable – but with the yield based demand function for IG robust and supply expected to be light on duration buying a kneejerk dip in the next few weeks in non-AI IG (GSIG30NH) seems prudent. • The amount of risk transfer within credit markets is materially increasing. And it’s not just one product as CDS is BACK, TRS continues to grow at pace, and cash bond volumes are showing significant growth across IG and HY.

As we approach the 2026 home stretch spreads have trended wider over the summer but have not broken the range (IG index at 80bps vs. 77 average and GS100 at 99bps vs. 96), while AI spreads (GSUAIBK), despite the early Aug wave of short covering, are also again drifting wider (now +50bps in past yr).

AI CAPEX – A Breather… For Now…

With already ~$300bn of supply YTD AI has easily been the most important topic in the IG market. The rest of the year though should likely be relatively calm, with just one more senior hyperscaler deal + some datacenters expected. This would represent that lowest amount of senior hyperscaler/chip debt issuance across any period since the start of the build out by almost 50%.

However, this is just a temporary reprieve… some very rough estimates on senior hyperscaler/chip issuance in 2027 show a potential rise of ~40% from this cohort to ~$340bn (using the below assumptions and please note this is not an official GS view but rather a rough outline based on extrapolating this year’s trends forward).

• Hyperscaler CAPEX of ~$930bn in 2027 (midpoint of GIR/BBG estimates) • 37.5% debt issuance as % of CAPEX (vs. ~30% in 2026 – the 2027 increase is due to an expected reduction in equity issuance from this cohort) • Hyperscalers 70% funded in USD (down from ~80% in ‘26) • $35bn of senior chip debt in ’26 becomes ~$50-$75bn in ’27 as chip financing build out expands. • Only includes names that issued in 2026

~$340bn would make the average quarter in 2027 equivalent to the busiest quarter we have seen thus far.

If we include harder to estimate portions of AI issuance like datacenters and other AI adjacent issuance with a modest growth rate YoY we don’t appear to be in the 5th inning of the buildout yet. This is also excluding structured chip financing, which could potentially be the biggest growth area in AI financing