Meta Platforms lost $124 billion in market value in a single session on Thursday. The trigger was not a product failure but a financing decision: the company said it would issue a new class of non-voting shares to pay for artificial intelligence (AI) and lifted the low end of its 2026 capital spending plan to $135 billion.

That is the AI buildout's new arithmetic. Meta reported negative free cash flow of $2.12 billion last quarter – its first cash burn in more than a decade. Across the largest cloud operators, annual capital spending has climbed to roughly $750 to $850 billion and is heading toward a trillion, a scale that no longer fits inside operating cash flow.

So the funding mix is shifting. The first phase ran on debt: large technology companies have raised about $250 billion in global bond markets this year, with total issuance projected near $320 billion, up roughly 60%. Now it runs on equity – and equity means dilution, with existing shareholders absorbing the cost.

The credit channel is where the strain shows first. CoreWeave's unsecured bonds yield around 15%, and its five-year credit default swaps have approached 1,200 basis points. Much of this paper sits in private credit funds and annuity portfolios, where losses surface slowly and quietly.

The bear case runs deeper than interest costs. Accelerators depreciate in three to four years, and each generation can strand the last; part of the sector's growth has been financed through circular arrangements, with chip suppliers backing the very buyers of their chips.

The argument over returns is now public. Gartner's John Lovelock, whose forecast helped fuel this month's selling, describes the race as "a last-man-standing position" in which most AI markets will "collapse down to a duopoly." Neoclouds have already voted with their feet: South Korea's Elice fell more than 30% on debut; Firmus, AirTrunk and SB Energy have shelved listings.

Bulls concede the financing problem without accepting the demand one. "Nobody is worried about Meta or Amazon or Google getting the financing they need," said Matthew Maley of Miller Tabak. "But if it gets prohibitive for smaller companies, then you don't have to have an implosion of things to make people nervous."

The market's judgment arrived quickly but narrowly: the Nasdaq fell 2.6%, while the Dow ended flat – a rotation, not a verdict.

What is being tested is not whether AI gets built. It is whether the money arrives before the returns do.

(With input from agencies)