The AI Capex Supercycle

The Thesis

Nvidia sells chips to CoreWeave. Nvidia also owns 5.96% of CoreWeave and put another $2 billion into it in January 2026. CoreWeave uses a Meta services contract as collateral to borrow $8.5 billion from banks, then builds the data center that houses the chips Nvidia sold it. Nvidia separately commits $30 billion in equity to OpenAI, which pays Oracle a reported $300 billion over several years for compute built largely from Nvidia chips. OpenAI, for its part, now holds roughly $2.6 billion in CoreWeave and Cerebras stock — payment in kind for committing to buy their chips and cloud time, and in Cerebras’s case, for lending it money too.

That is the AI capex supercycle: not a spending number but a financing shape. Four groups — chipmakers, hyperscalers, “neoclouds” like CoreWeave, and model labs — sell to each other, take stakes in each other, and borrow against each other’s contracts. TEXXR’s coverage record shows this loop tightening every quarter since 2024, and shows the trade press itself naming it a closed loop: Bloomberg’s October 2025 chronicle of “a wave of circular deals” among Nvidia and OpenAI drew 54 pieces of related coverage in the archive, from a Bank of England bubble warning to an outlet titled, without much subtlety, “OpenAI’s Trillion-Dollar Circle: When Vendors Finance Their Own Payday.”

Every ticker on this page sits somewhere in the loop. Nvidia is the chip. CoreWeave is the neocloud building the capacity. Salesforce is a step removed — a software incumbent whose balance sheet and pricing are now tied to the same cycle through an equity stake and a credit market that is starting to price in AI-linked risk. This page tracks what the record shows about the loop: how fast it grew, who is financing it, and where the record itself flags strain. It does not forecast where any of these stocks go next.

The Evidence

The clearest number in TEXXR’s archive isn’t a dollar figure — it’s a coverage count with a baseline. Quarterly volume on AI data-center capex ran at 85 articles in 2024 Q1. By 2024 Q4 it was 132. By 2025 Q4, 304. By 2026 Q2, 534 — a 6.3x jump in nine quarters, out of 2,879 articles total across the topic’s full 45-quarter history back to 2015. The sharpest single-quarter shift in what that coverage was actually about, measured by semantic drift between periods, lands earlier than the current boom’s timeline suggests: 2020 Q4, when Nvidia’s early supercomputer commitments and pandemic-era cloud spending first pulled the topic away from ordinary data-center announcements. The current acceleration is real, but it’s the second inflection, not the first.

The Stargate venture shows how unstable a single mega-deal can be even while the totals climb. OpenAI, SoftBank, and Oracle unveiled the $500 billion project in January 2025. By July, the Wall Street Journal reported it had “struggled to get off the ground” and had sharply scaled back its ambitions. By September, it had expanded again — five new US sites, pushing planned capacity toward 7 gigawatts. By February 2026, The Information was reporting that OpenAI had scrambled for compute as Stargate stalled amid clashes with SoftBank. The headline number survived; the plan under it kept changing.

The financing web is the part that repeats across every name in this cycle. Nvidia announced up to $100 billion for OpenAI in September 2025; by January 2026 that commitment had stalled and failed in its original form; by May it reappeared as a $30 billion equity stake, one piece of more than $40 billion in equity Nvidia committed across the AI stack in 2026 alone, per its own disclosures. CoreWeave sits downstream of that money and upstream of the compute: its $11.9 billion, five-year contract with OpenAI came bundled with $350 million of CoreWeave stock issued to OpenAI at IPO — the customer as shareholder. Its $14.2 billion contract with Meta, expanded to $21 billion in April 2026, became loan collateral the same year: reporting in February said CoreWeave was seeking bank financing backed explicitly by that Meta paper, and by March 31 the deal closed at $8.5 billion, which CoreWeave itself called “the largest chip-backed debt deal of its kind.” CoreWeave has also moved $2.6 billion of data-center construction debt off its own balance sheet using special-purpose vehicles — a capex-accounting maneuver with a direct precedent, covered below.

The money increasingly isn’t equity — it’s debt, and it’s compounding. Morgan Stanley estimated in August 2025 that hyperscalers would fund $1.4 trillion of a projected $2.9 trillion in AI infrastructure need, leaving more than half to bonds, private credit, and vehicles like CoreWeave’s. By June 2026, the same firm had raised its forecast again: global AI-tied debt issuance more than doubling to nearly $570 billion in 2026. TSMC, sitting upstream of the whole chain, keeps raising its own number too — its 2026 capex guidance moved from $52–56 billion to $60–64 billion in July, citing the “AI megatrend” directly. Even names a step removed feel it in their borrowing costs: Salesforce sold $25 billion in bonds in March 2026 at what the Financial Times called a “significant premium,” a sign, its sources said, of Wall Street jitters that have moved past the direct participants.

The record also carries the counter-evidence. The Wall Street Journal’s November 2025 accounting of the buildout put 2025’s US data-center capacity — built, underway, planned, or stalled — above 80 gigawatts, running into what it called the “ground truths of physical” limits: power, water, chips, and local opposition that don’t scale as fast as balance sheets do. One economist’s estimate, cited in a January 2026 Financial Times piece, put the whole AI investment boom at roughly 1% of US GDP — comparable in scale to prior investment booms, which is a benchmark, not a verdict on how this one ends.

The Companies

Nvidia sits at both ends of the loop. It supplies the chips every other name here is spending to acquire, and it increasingly funds the customers buying them: the OpenAI equity stake, the CoreWeave stake, the more than $40 billion in 2026 equity commitments across the stack. When Jensen Huang said an investment “might be the last time” Nvidia writes a check that size, TEXXR’s record treated the remark as news, not reassurance.

CoreWeave is the neocloud in the middle — Nvidia-backed since before its 2025 Nasdaq IPO, running 43 active data centers as of April 2026 (the same month it signed a multiyear Nvidia-chip supply deal with Anthropic), and holding a $99.4 billion revenue backlog against contracts with OpenAI, Meta, and Anthropic. It is also the clearest single case of the loop’s financial engineering: debt moved off its balance sheet through special-purpose vehicles, then more debt raised against the very hyperscaler contracts that are supposed to prove its demand is real rather than circular.

Salesforce isn’t building data centers. It’s the second-order name: a roughly $5 billion equity stake in Anthropic ties its balance sheet to the same cycle, its Agentforce pricing is shifting from flat subscriptions toward usage-based AI fees as compute costs eat into software margins, and its March 2026 bond sale — priced at a premium — is the record’s clearest sign that AI-linked credit risk has reached investment-grade enterprise software, not just the chipmakers and clouds.

The Lenses

Benjamin Graham built his discipline around distrust of what Security Analysis calls new-era reasoning — the argument, common to every boom, that old standards for valuing a business no longer apply because the future has turned too promising to measure by them. Graham’s answer wasn’t to predict the ending; it was to separate the price of a thing from what it can be shown to earn, and to treat any argument that skips that step as speculation wearing an investment’s clothes. Applied here, the test isn’t whether AI capex is justified — it’s whether each deal in the loop can be priced on its own terms, independent of the next deal in the chain that’s meant to validate it. The Intelligent Investor’s own historical case study is almost too on the nose: Global Crossing, laying fiber-optic cable under the ocean in the late 1990s, booked its construction spending as a long-lived capital asset rather than an expense, right up until the debt caught up with it in 2002. The book’s read on the company, even accounting for GAAP’s tolerance of the maneuver, is blunt — it was, in its words, “spending money like a drunken sailor.” CoreWeave’s own off-balance-sheet debt vehicles are the same accounting question in 2026 dress.

Robert Shiller supplies the mechanism for why coverage of a financing structure becomes part of the structure itself. Narrative economics treats a story about an asset as a contagious thing, spreading investor to investor the way an epidemic spreads host to host, with the story and the price reinforcing each other until something breaks the loop. The “circular deals” framing that Bloomberg put into wide circulation in October 2025 fits that pattern exactly: once “circular” became the word regulators, banks, and rival outlets reached for, the deals no longer needed to be individually fraudulent to become individually suspect — the label did work the balance sheets alone hadn’t done yet. Whether that narrative cools the cycle or simply becomes background noise the market absorbs is an open question the record hasn’t answered. It is, so far, only asking it.

What Moved

Sources