AI And Capital Allocation

Executive Summary

Artificial intelligence has triggered the largest concentrated capital-spending cycle in modern corporate history. In 2026 alone, the largest technology companies are on pace to spend a combined $650–690 billion on AI infrastructure — roughly a 60–74% increase from 2025 — and cumulative hyperscaler capital expenditure is projected to exceed $3 trillion between 2026 and 2030. This paper examines what that spending means for how capital is being allocated, financed, and ultimately valued across the technology ecosystem and the broader market.

Three shifts define the current period. First, the source of funding is changing: companies that historically self-funded growth from free cash flow are now issuing record volumes of corporate bonds and increasingly turning to private credit, off-balance-sheet special purpose vehicles, and asset-backed securitization to bridge a widening gap between spending and internally generated cash. Second, a web of circular investment arrangements - chipmakers investing in their own customers, cloud providers cross-holding equity in AI labs - has re-emerged in a form reminiscent of, though not identical to, the vendor-financing structures that inflated and then helped collapse the late-1990s telecom buildout. Third, an increasingly public debate over GPU depreciation schedules has emerged as a genuine swing factor in reported technology-sector earnings, with estimates of potential earnings overstatement running into the hundreds of billions of dollars over 2026–2028.

None of this means the AI infrastructure cycle is a bubble in the sense the dotcom telecom buildout was. The companies funding the buildout are, for the most part, more profitable, less leveraged, and facing more clearly demonstrated demand than their 1990s counterparts. But the financing structures, accounting judgments, and capital-allocation choices being made today will materially shape investor returns across equities, credit, and private markets for the remainder of the decade — and the risks are concentrated enough that a small number of companies and financing vehicles could have outsized effects on the broader market if the buildout's economics disappoint.