Apollo's Slok Warns AI Hyperscaler Debt Echoes 2008

Torsten Slok of Apollo warned that debt tied to hyperscaler AI expansion is showing stress similar to early signals before the 2008 housing collapse. The warning centers on credit default swap spreads that have widened sharply for credit linked to AI infrastructure funding. CDS spreads measure the annual cost to insure bonds against default, so a sharp move points to rising concern among credit investors. Slok framed the shift as a parallel to housing credit stress in 2007 and 2008, not as a call for identical losses. The remark was reported on September 16, 2026 and spread quickly across macro and crypto trading desks.

What Slok flagged

Slok pointed to hyperscaler borrowing used to fund large scale AI data centers and compute capacity. Hyperscalers are the largest cloud platform operators, and they sit at the center of current AI capital spending. The concern is that rapid growth in borrowing has increased exposure if cash flow from AI services arrives more slowly than expected. Widening CDS spreads suggest that some credit investors now want more compensation to bear that issuer risk.

Such warnings carry weight because Slok tracks corporate credit, rates and business cycle risk for Apollo. Apollo is a large alternative asset manager with deep activity in private credit and corporate lending. Its economists often comment on credit availability, spreads and refinancing conditions. In this case the message was simple: watch AI related credit the way markets once watched housing credit.

How hyperscaler AI debt grew

Demand for AI training and inference has driven a wave of spending on chips, servers, power and buildings. Data centers need land, electricity contracts, cooling systems and network links before they produce revenue. Many of those projects are funded with a mix of cash flow, bonds, bank loans and lease structures. When several large operators expand at once, total issuance can grow fast even if each borrower starts from a strong balance sheet.

Credit buyers initially welcomed that supply because the issuers had large revenues and long operating histories. Investment grade technology debt often offers yield with perceived safety relative to smaller borrowers. Low defaults in big tech reinforced that view through several credit cycles. The risk builds when future earnings are expected to cover present borrowing, and those earnings depend on continued AI adoption.

Why does widening CDS matter right now?

It matters because CDS spreads price the cost to insure against default and react fast when sentiment changes. A sharp widening means protection buyers will pay more, which usually reflects weaker liquidity, higher perceived default risk or forced hedging. Equity prices can stay calm while credit starts to tighten. That gap is why macro traders treat credit as an early signal.

Timing adds to the focus because rates, growth forecasts and AI spending plans are all under review. If borrowing costs stay high, refinancing and new project finance become more expensive. If AI revenue growth slows, coverage ratios look thinner. CDS moves do not prove distress, but they show where professional credit desks are paying to reduce exposure.

Parallels with 2008 housing stress

The 2008 housing collapse followed a long rise in mortgage borrowing tied to expectations for ever higher home prices. Loans were packaged, rated and sold across the financial system, which spread the exposure. Early stress appeared in credit gauges tied to mortgages before headline defaults surged. Traders who watched those gauges saw conditions tighten ahead of broader market recognition.

Slok compared that pattern to current AI infrastructure credit, not the asset itself. The shared feature is concentration around one growth story funded with rising debt. In housing it was home price growth and mortgage credit. In AI it is compute demand and data center finance. Concentration does not guarantee a crisis, but it can speed contagion if confidence turns.

There are also clear differences that traders cite when testing the comparison. Housing risk was widely distributed through securitized products held by banks, insurers and funds. Hyperscaler debt is more directly tied to named operating companies with diversified revenue. Corporate disclosure, bank supervision and post 2008 clearing rules also changed how credit risk is tracked. The open question is whether those safeguards offset higher absolute spending.

What does this mean for crypto traders?

It means tighter credit conditions could spill into risk assets including BTC. Bitcoin often trades with technology equities and reacts to liquidity, rates and credit appetite. When CDS spreads widen across corporate credit, market makers tend to cut risk and raise cash buffers. That response can reduce bids across equities, high yield bonds and crypto at the same time.

Crypto has its own structure, with no direct link to data center loans. Yet funding conditions still matter for exchanges, lenders, miners and funds that use debt. Wider spreads can raise the cost of working capital and hedging. Past stress episodes showed that Bitcoin can fall with risk assets even when the shock starts in traditional credit, then recover on its own path once liquidity stabilizes.

What to watch next

Traders will watch whether CDS spreads stabilize, keep widening or spread to related issuers. Fresh bond sales, earnings calls and capital spending updates will show how operators plan to fund the next phase of AI buildout. Credit rating actions, bank loan terms and lease disclosures can add detail on costs and maturities. Macro releases on inflation, jobs and rates will shape the refinancing backdrop.

Another set of signals concerns actual use of AI capacity, including cloud growth, chip orders and power contracts. If demand absorbs new supply, credit concern may fade even after a sharp spread move. If projects are delayed or utilization lags, concern may persist. Key risks include a broader credit repricing, cuts to capital spending and weaker appetite for long duration corporate debt. Stabilization in spreads would suggest the warning marked a scare rather than a turn.

For now the debate is whether hyperscaler AI debt becomes a crowded short or proves resilient. Slok placed it in the lineage of 2008 housing stress because early credit moves mattered then. Skeptics will ask for defaults, downgrades or failed deals before accepting the parallel. The next round of credit trading, issuance and company guidance should give a cleaner read.