The AI infrastructure race has created a reverse crowding out phenomenon where private-sector borrowing for AI infrastructure competes with sovereign debt for the same pool of long-term capital. According to an ICICI Bank Research report, AI debt issuance could be larger than the US government's comparable long-term debt supply, with AI debt supply estimated at around $500 billion compared to the US Treasury's $424 billion net issuance of long-term debt in 2025. This represents a fundamental shift from traditional crowding out, where government borrowing reduces funds available for private investment, to a situation where private AI borrowing creates additional pressure on government bond markets.
The AI infrastructure race is pushing top tech firms Alphabet, Amazon, Meta, Microsoft and Oracle to have a combined spending of over $1 trillion annually from 2027 onward. As reported by ICICI Bank Research, these firms are expected to spend $729 billion on capital expenditure in 2026, rising to $1.069 trillion in 2027. Hyperscaler capex is projected to remain above $1 trillion through 2030, reaching $1.184 trillion in 2028, $1.172 trillion in 2029 and $1.213 trillion in 2030. The spending is increasingly stretching companies' operating cash flows, with hyperscalers' capex projected to reach 90% of operating cash flow by 2026 and 104% in 2027, up from a third of operating cash flows between 2018-2021.
The surge in private-sector AI borrowing is putting additional pressure on government bond markets through unprecedented debt issuance patterns. US hyperscalers had already borrowed $220 billion through debt instruments in 2026, with borrowing expected to increase as AI infrastructure requirements grow. Around 80% of hyperscalers' bond issuance has a maturity of more than five years, with 23% categorised as 'very long term'. AI-related debt is estimated to account for around 15% of US investment-grade debt, while the five hyperscalers alone are projected to exceed 5% of the investment-grade index by the end of 2026. This creates direct competition with long-term US Treasury bonds, with AI debt issuance for similar maturities estimated at around $500 billion.
Despite the mounting debt concerns, Singapore's Temasek remains confident about massive AI infrastructure spending despite growing scrutiny of funding methods. As reported by Reuters, Temasek's North America Head Jane Atherton stated that companies funding the AI buildout believe large-scale investments will yield risk-adjusted returns with growing AI adoption. However, recent developments show increased scrutiny of funding methods, with Alchemy Capital Management Deputy CIO Alok Agarwal expressing caution about the AI investment cycle as hyperscalers increasingly rely on debt financing. The top five hyperscalers raised close to $240 billion in debt in the first eight months of the year, more than double the previous full-year figure, raising questions about sustainability.
The funding pressure extends beyond the five major hyperscalers, with ICICI Bank Research estimating total AI debt issuance at $489-570 billion in 2026, compared with $220-250 billion for the core hyperscalers. The report anticipates that AI borrowing could peak in 2027 and moderate thereafter as AI investments begin generating revenue and operating cash flows fund a larger part of capital expenditure. This would reduce competition for sovereign debt and eventually reduce pressure on bond yields. The funding requirement stems from AI infrastructure requiring large upfront investments in hardware infrastructure, data centres, power generation, GPUs and networking equipment, creating a mismatch between initial investment requirements and return generation timelines.
