Three trading days after AMD's Tuesday print and Meta announcement, the JPMorgan Neutral / Goldman Sachs Neutral $240 / Bank of America Buy $280 spread has not converged. JPMorgan's Harlan Sur kept Neutral with a warning that the Meta warrant could "represent $30 billion to $100 billion in equity value" while reducing gross margins by "200 to 400 basis points." Goldman's Toshiya Hari raised his price target to $240 from $210 but kept Neutral. BofA's Vivek Arya held Buy at $280 with a $20 EPS-by-2030 model on the combined Meta and OpenAI warrant cohort. [1] [2] The spread between the three desks ran into Friday's close.

The May 7 paper opened the file on the three-way sell-side spread as the warrant structure becoming the news. Today's standard reads the spread as durable — and the warrant cohort as a documented two-counterparty structure with implications for the SpaceX S-1 expected at T-7. The AMD-Meta warrant, signed February 23, gives Meta a $0.01-strike right to purchase up to 160 million AMD shares — about 10% of the company — vesting in tranches as Meta's cumulative purchases scale to six gigawatts and as AMD's stock price hits performance thresholds running up to $600 per share. [3] [4] The AMD-OpenAI warrant, signed October 5, 2025, is structurally identical: the same 160 million share count, the same penny strike, the same six-gigawatt vesting structure, the same stock-price threshold ladder. [5]

Together, the two warrants represent up to 320 million shares — roughly 20% of AMD's existing share count if both counterparties exercise fully. The dilution is visible in the 8-K disclosures and the warrant agreements filed as Exhibit 4.1 to each. [4] The warrants are not contracts; they are equity-issuance instruments. JPMorgan's Sur is reading the structure as economic compensation that has to be subtracted from gross-profit dollars before earnings flow through. Sur's 200-to-400-basis-point margin haircut, applied across an AMD revenue base running toward $40 billion in 2026, is a $1.6-billion-to-$3.2-billion annual gross-profit deduction the consensus model does not yet carry. [1] BofA's Arya is reading the same structure as customer-alignment incentive that pays for itself through revenue scale. The two readings cannot both be right.

The warrant cohort changes the institutional read on hyperscaler-chip economics. AMD has now committed up to 320 million shares of equity issuance to two AI buyers in exchange for purchase commitments scaled in gigawatts. Meta has Nvidia commitments running parallel; the AMD-Meta deal is not exclusive. OpenAI has the Cerebras 750-megawatt contract worth $20 billion-plus, the Broadcom XPU collaboration, and the AMD warrant; the AMD-OpenAI deal is also non-exclusive. [4] The structure that has emerged — chip-buyer equity for chip-maker silicon, with the equity vesting on stock-price performance — is now the standard architecture across at least three named transactions and three named issuers. The sell-side has not yet, in print, agreed on how to model that architecture.

What the spread tells the next quarter is that AMD is no longer trading on its own results. It is trading on the warrant structure's interaction with hyperscaler capex and a stock-price target list that runs to $600. The Q2 guide of $11.0-$11.6 billion in revenue is the fundamental number; the warrant disclosure in the 10-Q, due within the next several weeks, is the document the three desks will reconcile against. [4] The 10-Q's filing date is the next institutional event the spread converges or persists past.

The cohort precedent matters for SpaceX's expected S-1 at T-7. SpaceX has been reported to be readying a public-offering filing for late May or early June; the prospectus will include the company's $573 million in related-party transactions disclosed under Reg S-X. If the S-1 inherits the warrant-cohort architecture — equity to a strategic counterparty in exchange for purchase commitments — it does so on a precedent now established by AMD-Meta and AMD-OpenAI on the same documentary basis. Whether the SpaceX prospectus footnotes the AMD precedent explicitly is the watch item.

CEO Lisa Su's framing on the AMD Q1 call sets the operative magnitude. Each gigawatt of compute, in Su's language, is "worth double-digit billions." [6] At six gigawatts of Meta-purchased silicon and six gigawatts of OpenAI-purchased silicon, the cumulative revenue runs into the hundreds of billions. The warrant dilution at 320 million shares maxes out at roughly $192 billion of equity value at $600 per share — the threshold price for the final tranche. The arithmetic of revenue against dilution is the BofA case in compressed form. The arithmetic of warrant-as-margin-haircut is the JPM case in compressed form. Both arithmetics use the same inputs.

The MI450-based GPU and the Helios rack-scale architecture deploy in the second half of 2026 for the Meta deal's first gigawatt. [4] Vesting against that gigawatt unlocks the first tranche. Whether AMD's stock price at deploy-date is above the first-threshold price determines whether the Warrant Shares are exercisable. Both axes — silicon shipment and share price — gate exercise. The structure is contingent. The cohort it represents is not.

Two warrants. Two prospectuses on the same May tape. Three sell-side desks unable to agree on the price. The 10-Q is the next reconciliation document.

-- THEO KAPLAN, San Francisco