AMD AI Deals Could Add 320m Shares: Who Pays?

AMD reported 1,632,475,042 shares outstanding on July 29, 2026, against warrants that could add 320m shares through its OpenAI and Meta agreements. At maximum issuance, maintaining the same earnings per share would require roughly 19.6% more net profit, before other changes to the share count.

That is the useful test behind the AI demand headline: how much additional profit reaches each existing share after the customer incentives? The contracts were announced in October 2025 and February 2026. They matter now because the original announcements scheduled initial deployments or shipments for the second half of 2026. Those schedules do not prove that delivery or warrant milestones have been completed.

Have OpenAI and Meta already received 320m AMD shares?

No such conclusion follows from the filings used here. AMD’s Q2 10-Q, Note 12 reported no warrant shares vested or exercisable as of June 27, 2026, and no warrant impact on that quarter’s or half-year’s financial statements. This is a June status; the October 11 status is unverified.

The maximum is a conditional entitlement. Vesting establishes a right under the contract; exercise and share issuance are additional steps. Treating all possible shares as issued today collapses those stages and overstates what the disclosure establishes.

What has to happen before the maximum shares can be issued?

GPU purchases alone do not satisfy every condition. The two original agreement filings disclose these headline terms:

Source: the respective Form 8-K Item 1.01 disclosures. GW refers to the specified GPU deployment or purchase measure, not revenue. Stated expirations are subject to the warrant documents’ terms.

OpenAI also has stock-performance thresholds. Both contracts impose other technical and commercial conditions before exercise. The $600 figure is a contractual condition, not our price target or evidence that AMD currently satisfies it. Nor is an initial binding purchase commitment the same as an unconditional purchase of the entire deployment contemplated by an agreement.

The detailed vesting and exercise schedules in Exhibits E and F of both warrant documents are redacted. We therefore cannot reconstruct a precise tranche-by-tranche issuance calendar or assume that each additional gigawatt earns an equal number of shares. Initial shipment schedules are disclosed in the OpenAI announcement and Meta announcement.

Is the dilution 19.6% or 16.4%?

Both numbers are correct, but they answer different questions. We use the dated July 29 outstanding share count as the base and assume the full contractual maximum is issued, with no buybacks, employee share issuance or other changes.

Source: July 29 share count from the 10-Q cover; maximum shares from the agreement filings. Results are our conditional calculations, not an October share-count observation.

New shares = 160,000,000 + 160,000,000 = 320,000,000
Share-count increase = 320,000,000 / 1,632,475,042 = 19.60%
Ownership fraction reduction = 320,000,000 / 1,952,475,042 = 16.39%

An investor who keeps the same number of shares would own a smaller fraction of the company. That does not establish a corresponding fall in the share price or the value of the holding: the business may also become more profitable. The economic tradeoff is additional customer business against a larger denominator dividing the resulting profit.

How much profit growth would offset maximum issuance?

About 19.6% in this simplified model. If shares increase by that percentage, net profit must grow by the same percentage to keep the per-share result unchanged. Revenue growth cannot substitute for net profit growth in that calculation.

The starting per-share index is 100. All rows assume 320m new shares against the July 29 base; profit growth is a scenario input, not a forecast.

Per-share index = 100 × (1 + assumed profit growth) × S / (S + new shares)
S = 1,632,475,042; new shares = 320,000,000
Flat profit: 100 × S / (S + new shares) = 83.61
EPS-neutral profit growth = new shares / S = 19.60%

The model isolates the denominator. It does not predict the earnings produced by these contracts, their timing, margins or accounting effects. With flat profit, the per-share result falls by the same fraction as an existing investor’s ownership. With sufficient profit growth, the enlarged business can offset the additional shares.

Reported diluted EPS is more complicated: it uses weighted-average shares and applicable rules for potentially dilutive instruments. This snapshot model should not be added mechanically to AMD’s reported diluted EPS denominator. Cashless exercise is permitted, so actual net shares issued may also differ from the face maximum.

What could make this trade worthwhile—or costly?

The result depends on incremental net profit and the shares actually issued. Large deployments can create demand, but delivery costs, research spending, financing, taxes and any warrant accounting effects all matter before revenue becomes earnings for existing owners.

AMD’s Q2 note says the warrants will be classified as liabilities until conditions for equity classification are met. Its reported June financial-statement impact was zero. That does not establish zero future accounting cost. Future disclosure must be read alongside any change in legal vesting, exercise and issuance.

The opposite risk is treating the maximum dilution as inevitable while ignoring the business earned in return. Conditions can limit issuance, and future buybacks or other share changes can alter the denominator. Without a verified issuance schedule and contract-level profitability, a precise net benefit remains unverified.

Financing provides a separate pressure on profit: our AMD note-offering analysis examines coupon obligations rather than equity dilution. Intel’s escrowed-share analysis shows why share issuance and EPS treatment need distinct checks. And CoreWeave’s EBITDA-to-earnings bridge illustrates why demand and a headline profit measure do not establish net earnings.

For the next AMD filing, check completed shipments, vested and exercisable tranches, actual shares issued and the warrant accounting note. Then compare net profit with the corresponding share denominator. The agreements can expand AMD’s business; the unresolved question is how much of that expansion reaches each existing share.