Apollo Global Management’s chief economist has put hyperscaler credit default swaps at the centre of a fresh debate about how much financial risk the AI infrastructure boom is quietly accumulating. In a 16 September Daily Spark note, Torsten Slok, Apollo’s partner and chief economist, argued that rising CDS prices on bonds issued by the giant cloud computing platforms powering artificial intelligence are not a technical artefact of dealer hedging. They reflect something more fundamental.
‘What the market is repricing is hyperscaler credit fundamentals, namely a debt-financed AI capex cycle with rising leverage, negative free cash flow and uncertain payback on depreciating assets,’ Slok wrote.
The tell, in Slok’s framing, is where the widening has shown up. If banks were simply hedging their exposure to a surge in bond underwriting, the CDS of those banks would move, too. They haven’t. The gap between hyperscaler CDS and bank CDS has widened to around 60 basis points from roughly zero since October 2025, isolating the risk to the cloud platforms themselves. Apollo followed up with a second note on 21 September, titled ‘Hyperscaler Credit Rests on One Consensus Assumption,’ suggesting the firm regards this as an evolving rather than settled argument.
Hyperscaler Credit Default Swaps and the Debt Build-up
The backdrop to the CDS moves is a debt issuance cycle of a scale the market has rarely absorbed. According to the OECD’s Global Debt Report 2026, hyperscalers raised $122 billion in bonds during 2025, with $88 billion of that arriving in roughly 54 days in the final months of the year. That single-year haul represented 45% of total bond issuance by all technology firms globally and more than three times the historical annual average since 2000.
The pace has not slowed. CNBC reported that Amazon’s long-term debt rose 81% to $119 billion between the end of 2025 and 31 March 2026, while Alphabet’s rose 111% to $98 billion in the first half of 2026. Alphabet also turned cash-flow negative in the second quarter of 2026 for the first time. Amazon guided to $200 billion in capital expenditure for the full year 2026 in February; Alphabet subsequently lifted the top end of its own 2026 capex forecast to $205 billion.
CreditSights projects the top five hyperscalers will collectively spend approximately $602 billion in capital expenditure in 2026, a 36% increase year-on-year, with roughly 75% of that directed at AI infrastructure. Capital intensity has reached 45% to 57% of revenue for some platforms. Evercore ISI, led by analyst Amit Daryanani, projects the aggregate capex-to-operating-cash-flow ratio for the Magnificent Seven plus Oracle to exceed 60% in both 2026 and 2027, up from around 50% in 2025.
The credit market has responded. J.P. Morgan Asset Management notes that hyperscaler spreads have widened by around 30 basis points year-to-date, versus just 2 basis points for the broader USD investment-grade index. Adjusted for duration, hyperscaler spreads are trading at around 105 basis points, wider than comparably rated bonds. Hyperscaler bonds represented only about 3.5% of the USD investment-grade index at the start of 2026, meaning a relatively small pocket of the market is driving a disproportionate share of spread movement.
The Fundamentals Behind the Widening
The FactSet forward figures disclosed in the original Apollo note illustrate the pressure on individual balance sheets. Alphabet carries a forward debt-to-equity ratio of 13% and forward free cash flow of negative $25.7 billion; Amazon shows debt-to-equity of 23% and forward free cash flow of negative $30 billion; Meta Platforms has a debt-to-equity ratio of 34% and forward free cash flow of negative $25.7 billion. FactSet’s own analysis shows that incremental debt as a share of capex for the hyperscaler group rose from 9% in FY2024 to 32% on a last-twelve-months basis by mid-2026. Only Microsoft, with a forward debt-to-equity of 7.34% and positive forward free cash flow of $33.4 billion, stands apart from the pattern.
Not everyone reads those numbers as cause for alarm. Paul Meeks, head of technology research at Freedom Capital Markets, argues the time horizon matters. ‘[Hyperscalers] don’t really add their capacity in a significant enough way into 2027 and 2028 to know how this is going to turn out,’ he said. ‘We’re starting to see a turn to the positive in their margins, and if we continue to see [this], there will be less concerns.’ J.P. Morgan similarly identifies the recovery in free cash flow after 2027 as the key test for bondholders.
The sceptics, however, point to what the CDS market itself is signalling. Dean Baker, founder of the Center for Economic and Policy Research, framed the position plainly: ‘They obviously think there is a substantial risk that the AI companies cannot make good on their commitments.’ The hyperscaler CDS widening, in that reading, is not noise. It is the most sophisticated part of the credit market pricing in the possibility that the revenue payback on hundreds of billions of dollars in AI infrastructure never fully arrives.
The next concrete test arrives when the large platforms report third-quarter results: free cash flow trajectories and any revision to full-year capex guidance will either vindicate the bulls or give the CDS market fresh cause to widen further.
