It’s the worst time in years to invest in AI credit markets: ‘Almost no upside and plenty of downside’

Andromeda Capital Management says AI credit is the “worst time in years” to enter, citing historically tight spreads that leave “almost no upside and plenty of downside.” With hyperscaler AI spending expected to reach $1.2T by 2027 (per Bank of America’s Vivek Arya), the incremental funding need is shifting toward borrowing by non-public AI firms and likely increasing reliance on private credit.

Analysis

The market is likely underpricing the shift from equity-financed AI buildout to debt-financed AI infrastructure. That transition is usually good for the biggest balance sheets and bad for lenders that need perpetual spread compression to justify returns: the first-order winner is the hyperscaler cohort with true cash generation, while the second-order loser is private credit, where returns depend on refinancing optionality and stable residual values. For BAC, the opportunity is mainly fee capture and balance-sheet intermediation, not a durable uplift in net interest income; that makes the upside more cyclical than structural.

The more important mechanism is that debt funding forces a higher hurdle rate on marginal AI projects. Over 1-3 months, that should pressure smaller AI vendors, data-center adjacent borrowers, and any lender stretching for yield into opaque credits; over 6-18 months it widens the moat for MSFT/GOOGL/AMZN because they can self-fund or borrow cheaply while competitors cannot. If AI economics disappoint, the repricing will show up first in new-issue concessions and covenant language, not in headline defaults.

Contrarian view: spreads being tight is not itself a buy signal if issuance is still being absorbed by performance-chasing capital. The consensus is assuming AI credit behaves like investment-grade utility financing; the hidden risk is that these assets are quickly obsolete, so duration and refinancing risk are materially higher than the market is pricing. The thesis is falsified if large AI financings clear with minimal concession for multiple quarters and hyperscaler capex guidance keeps rising without any deterioration in free cash flow conversion.

Key Decisions for Investors

• Prefer BAC as a relative beneficiary of AI financing activity versus direct-credit exposure, but size it modestly; treat it as a fee-flow trade, not a core earnings re-rating.
• Avoid or reduce exposure to private-credit / direct-lending proxies (e.g., ARES, BX, KKR, BIZD) if AI-originated deals are being priced at historically tight spreads; the asymmetry is poor until spreads widen 50-100 bps.
• Pair trade: long BAC / short BIZD or a private-credit proxy for 1-3 months. Thesis: banks capture syndication/arrangement economics while direct lenders absorb the downside if AI capex growth slows.
• Set an alert on new-issue concessions, ICE BofA HY/IG spread widening, and rating outlook changes tied to AI infrastructure borrowers; if spreads gap wider while issuance volume falls, that is the signal to add risk back.
• If we see hyperscaler capex funded entirely out of operating cash flow for another quarter, stand down on the bearish credit view; that would extend the runway and reduce the chance of a near-term repricing.