Any AMD stock price prediction that starts from the $913.9bn market capitalisation on the quote screen is already working from the wrong number. That figure uses the 1,632,475,042 shares on the cover of AMD’s most recent 10-Q. It ignores 320 million more that AMD has already promised, in two signed warrant agreements, to OpenAI and to Meta Platforms at an exercise price of one cent. Count those and the company that closed Friday at $559.82 is not approaching a trillion dollars. It passed that mark months ago.
Here is the arithmetic nobody runs. On the 10-Q share count, $1trn of market value arrives at $612.57 a share, which is 9.4% above Friday’s close and a shade below the $644.62 mean price target Nasdaq’s tracked analysts carry. Add the two warrants and the fully diluted count becomes 1,952,475,042. At $559.82 that is a $1.093trn claim on the business today, and the price at which AMD would first be worth a trillion on that basis is $512.17, a level AMD was still trading below as recently as Tuesday 15 September, when it closed at $504.20. So the interesting question for the next twelve months is not whether AMD joins the trillion-dollar club. It is whether the revenue that vests those warrants shows up fast enough to justify the share count that comes with it.
Key facts
- AMD closed at $559.82 on Friday 18 September 2026, 4.4% below its 52-week high of $584.73 and 3.7x its 52-week low of $149.85 — stockanalysis.com quote, pulled 19 September 2026.
- Second-quarter revenue was $11.54bn, up 50% year on year, with Data Center revenue of $6.7bn up 107% and accounting for 58% of the company — AMD Form 8-K Item 2.02, Exhibit 99.1, 4 August 2026.
- AMD guided third-quarter revenue to approximately $13bn, plus or minus $300m, with non-GAAP gross margin of about 56% — same filing.
- Two performance warrants, each for up to 160 million shares at $0.01, are outstanding: one issued to OpenAI OpCo on 5 October 2025, one to Meta on 23 February 2026. Both escalate to a $600 share-price threshold for the final tranche — 8-K of 6 October 2025 and 8-K of 24 February 2026.
- AMD closed a $4.75bn senior notes offering on 17 August 2026 across four tranches maturing 2029 to 2036, priced from 4.600% to 5.500% — 8-K filed 17 August 2026.
- Cash and cash equivalents stood at $5.09bn at 27 June 2026, down from $5.54bn at the end of 2025 — Q2 2026 balance sheet, same 8-K exhibit.
- Of the 35 analysts Nasdaq tracks, 34 carry a rating: 28 buy, six hold, none sell, with targets spanning $465 to $1,250 — Nasdaq analyst consensus, retrieved 19 September 2026.
The merchant accelerator finally has volume, and a balance sheet to match
For most of the past decade the bear case on AMD in AI was about software, not silicon. The chips benchmarked well and the deployments never arrived at scale. That argument is now closed by AMD’s own reported numbers rather than by anyone’s roadmap slide.
Data Center revenue of $6.7bn in the June quarter was up 107% year on year and represented 58% of company revenue, according to the Item 2.02 exhibit AMD filed on 4 August. Client and Gaming contributed $3.8bn, within which the Ryzen-driven client line grew 23% while gaming fell 31% on weaker semi-custom volumes. Embedded added $977m. Strip the mix apart and the picture is unambiguous: AMD has become a data-centre company with three profitable side businesses attached, and the transition happened inside about six quarters.
The financing tells the same story from a different angle. On 17 August AMD closed $4.75bn of senior unsecured notes in four tranches — $1.25bn at 4.600% due 2029, $1.50bn at 5.000% due 2031, $1.00bn at 5.250% due 2033 and $1.00bn at 5.500% due 2036 — with Barclays, BofA Securities, Citigroup, J.P. Morgan, Morgan Stanley and Wells Fargo as representatives of the underwriters. A company holding $5.09bn of cash at quarter end does not raise nearly as much again in term debt to fund a business that ships when customers happen to order. It does that when it has to pre-position substrate, memory and rack-scale inventory against delivery commitments it has already signed.
That is also the part of the thesis most exposed to somebody else’s pricing decisions. Rack-scale AI systems carry a great deal of high-bandwidth and conventional memory, and every dollar of DRAM and NAND inflation lands in AMD’s bill of materials before it lands in anyone’s margin. Set against that, the demand side of the AI capex debate has been argued at length elsewhere, including in FinanceFeeds’ own work on what a deliberate slowdown in AI spending would do to the trade; the point here is narrower and it is about share.
Lisa Su, chair and chief executive of AMD, put the quarter this way in the 4 August release: “We delivered an excellent quarter, with record revenue and profitability as Data Center revenue more than doubled year-over-year.”
Two customers, two warrants, and the same $600 line
The structure of AMD’s largest commercial relationships is unusual enough to be worth reading in the filings rather than the coverage.
On 5 October 2025 AMD issued OpenAI OpCo a warrant to purchase up to 160 million shares at $0.01. Vesting is tranched against Instinct GPU purchases, with the first tranche released on delivery of the initial gigawatt of MI450-series product and full vesting contingent on purchases reaching six gigawatts. The 8-K adds a condition that is easy to miss: vesting is “further subject to achievement of specified Company stock price targets that escalate to $600 per share for the final tranche”. The warrant runs to 5 October 2030.
On 23 February 2026 AMD did it again. Meta Platforms received a warrant on the same terms — up to 160 million shares at a cent, tranched against a six-gigawatt purchase commitment, with price thresholds that “escalate to $600 per share for the final tranche” and an expiry of 23 February 2031. The accompanying release describes a multi-generation deal built on the Helios rack architecture, with first-gigawatt shipments of a custom MI450-based part beginning in the second half of 2026 alongside sixth-generation EPYC CPUs codenamed Venice.
Mark Zuckerberg, founder and chief executive of Meta, was direct about the motive in that release: “This is an important step for Meta as we diversify our compute. I expect AMD to be an important partner for many years to come.”
Read the two documents together and something specific emerges. AMD has handed its two largest announced accelerator customers an equity payoff that is contingent on AMD’s own share price reaching $600. Friday’s close of $559.82 sits 7.2% below that line. The people with the strongest commercial incentive to see AMD’s stock above $600 are the same people writing the purchase orders that would put it there. That is alignment by design, and it is also a reflexive loop that no discounted cash-flow model handles gracefully.
The competitive alternative to buying merchant accelerators is designing custom ASICs with a partner such as Broadcom, and several of these same hyperscalers do both; that path trades flexibility for unit economics and is a separate argument. AMD’s bet is that a standard part with an open software stack wins the workloads that change fastest. Comparable share questions are running through the whole sector right now, from Nvidia’s position inside the S&P 500 to the licensing model at Arm.
The re-rating that did not happen
AMD is up roughly 254% over twelve months, from a close of $157.92 on 18 September 2025 to $559.82 on Friday. The received reading of a move like that is multiple expansion: a market paying more for the same earnings because the story improved.
The filings say otherwise, and this is where two sources combine into something neither states on its own.
Non-GAAP diluted earnings per share went from $0.48 in the June 2025 quarter to $1.66 in the June 2026 quarter. That is a 3.46x increase. Over the same twelve months the share price went from $157.92 to $559.82, a 3.54x increase. The multiple barely moved. Almost the entire 254% is earnings delivery, not sentiment. One honest caveat belongs on that comparison: AMD’s own release footnotes that the June 2025 quarter carried $800m of inventory and related charges tied to US export controls on the MI308, which flattered the growth rate. Even allowing generously for that, the direction holds. Investors who assume they are paying a bubble multiple for AMD are, on the company’s reported figures, mostly paying for profit that already exists.
The chart carries a second lesson in how the market now prices these announcements. AMD closed at $164.67 on Friday 3 October 2025 and at $203.71 on Monday 6 October, the day the OpenAI agreement was published: a 23.7% single-session gain. The Meta agreement of 24 February 2026 was structurally the same deal, same six gigawatts, same 160 million warrant shares, same $600 terminal threshold. AMD closed at $196.60 on 23 February and $213.84 on 24 February, a gain of 8.8%.
| Measure | The bull reading | The bear reading |
|---|---|---|
| Data Center growth | $6.7bn, up 107% year on year, 58% of revenue | Concentrated in a handful of named customers |
| Non-GAAP gross margin | 56% in Q2 2026, guided to about 56% in Q3 | Accelerator mix has historically diluted margin as it scales |
| Warrant shares | Align two anchor customers with shareholders | Up to 19.6% dilution at a $0.01 strike |
| $4.75bn notes | Funds the ramp without issuing equity | First material interest burden in a cyclical business |
| Analyst spread | 28 buy ratings, a $1,250 high target | A $465 low target sits 17% below Friday’s close |
The fading announcement premium is the number to watch. A second identical commitment moved the stock a third as far as the first. Whatever the third one is worth, the market has told you it will be less.
Export controls, customer concentration and a foundry in one jurisdiction
Three structural constraints sit underneath every scenario below, and only one of them is about demand. Any AMD stock price prediction that treats them as background rather than as inputs is incomplete.
The first is export control. AMD’s own second-quarter release attributes $800m of inventory and related charges in the prior-year quarter to “the U.S. Government’s export control on AMD Instinct MI308 data center GPU products”. Licensing for advanced accelerators is administered by the Commerce Department’s Bureau of Industry and Security, and the policy has been revised repeatedly. A company whose fastest-growing product line can be made unsaleable into a major market by an administrative rule change carries a risk that no multiple fully prices. AMD has already booked the cost of that once.
The second is concentration. Announced commitments now run to six gigawatts with OpenAI, six with Meta and, per the Q2 release, a partnership with Anthropic covering up to two gigawatts of MI450-series product in Helios racks. Fourteen gigawatts of announced demand from three counterparties is a magnificent order book and a governance problem in the same breath. Two of those three hold warrants that could convert into 19.6% of the company. If either relationship deteriorates, the revenue and the register move together.
The third is manufacturing. Leading-edge Instinct and EPYC silicon is fabricated by TSMC, which concentrates AMD’s supply into one foundry relationship and, for the most advanced nodes, largely one jurisdiction. We have tracked that dependency across the sector for some time, and it is the same constraint that shapes the outlook at TSMC itself. Institutional ownership of the whole complex is heavily overlapping, which is why a single macro shock tends to reprice every name at once — a pattern visible in how large asset managers hold Nvidia, Micron and AMD together.
Jean Hu, executive vice-president, chief financial officer and treasurer, framed the near-term shape of the business in the August release: “We expect Data Center sales to accelerate in the second half of 2026, driving stronger overall revenue growth and continued earnings expansion.”
The call: $615 base, $725 bull, $385 bear
Start from the guide. AMD expects about $13bn of third-quarter revenue at roughly 56% non-GAAP gross margin. Holding operating expense growth near the 8% sequential pace of the past two quarters gives non-GAAP operating income of about $3.6bn, and applying the 89% operating-to-net conversion AMD posted in the June quarter across roughly 1.67bn diluted shares produces non-GAAP earnings of about $1.93 a share. Those are our estimates built on AMD’s own guidance, not company figures. Carried through a fourth quarter at a similar sequential slope, full-year 2026 non-GAAP earnings land near $7.15, and a 2027 at 45% growth lands near $10.40.
Base case, $615, probability 50%. That is 59x our 2027 estimate, a modest premium to the 54x the stock trades on today, and it implies a $1.004trn market capitalisation on the 10-Q share count. It is 9.9% above Friday’s close. The base case is simply that AMD converts the first gigawatt for each of OpenAI and Meta on schedule and margins hold at 56%.
Bull case, $725, probability 25%. Seventy times the 2027 estimate, a $1.18trn capitalisation basic and $1.42trn fully diluted for both warrants. This needs two things at once: the second gigawatt tranches to be pulled forward rather than pushed out, and non-GAAP gross margin to hold or improve while accelerator mix rises. The second condition is the harder one, because selling rack-scale systems at volume has historically dragged product margin down, not up.
Bear case, $385, probability 25%. Thirty-seven times the 2027 estimate and a $629bn capitalisation. AMD last closed below this level on 5 May 2026, so it requires no new low at all, only a retrace to where the stock sat four and a half months ago. The route there is a slipped MI450 qualification at one of the two anchor customers, or an export-control revision that strands inventory a second time.
Invalidation. A weekly close below $455 breaks the base case outright. That is the closing low of $456.16 set on 3 September 2026, and losing it would mean the market has stopped believing the second-half acceleration Jean Hu guided to. On the upside, a sustained move through $612.57 puts AMD in the trillion-dollar bracket on the reported share count and starts the final warrant tranches into view.
What would change our mind. A single line in the third-quarter release: non-GAAP gross margin. If it prints materially below the guided 56% while Data Center revenue grows, the merchant accelerator model is being bought at a discount and the bull case is wrong regardless of volume. If it holds at 56% or better on a higher accelerator mix, the base case is too conservative.
Frequently asked questions
What is the AMD stock price prediction for the next twelve months?
Our base case is $615 a share with a 50% probability, against Friday’s close of $559.82. The bull case is $725 at 25% and the bear case is $385 at 25%. Each level is derived from an estimated 2027 non-GAAP earnings figure of about $10.40 built on AMD’s own third-quarter guidance, at multiples of 59x, 70x and 37x respectively.
How many shares does AMD actually have outstanding?
The cover of the 10-Q for the quarter ended 27 June 2026 reports 1,632,475,042 shares outstanding as of 29 July 2026. Two outstanding warrants could add up to 320 million more at an exercise price of one cent, taking the fully diluted figure to 1,952,475,042, or 19.6% above the reported count.
Why does the $600 share price matter to AMD’s customers?
Both the OpenAI warrant of October 2025 and the Meta warrant of February 2026 tie vesting of the final tranche to AMD share-price thresholds that escalate to $600. Neither customer captures the full 160 million shares unless AMD trades there, which gives two of AMD’s largest buyers a direct financial interest in the equity as well as the product.
Is AMD taking share from Nvidia in AI accelerators?
On AMD’s reported figures it is growing quickly from a much smaller base: Data Center revenue of $6.7bn in the June quarter, up 107% year on year. That is a share gain in absolute terms. Whether it is a gain in percentage terms depends on Nvidia’s concurrent growth, which is disclosed separately and on a different fiscal calendar.
What is the biggest risk to the AMD bull case?
Gross margin. AMD guided third-quarter non-GAAP gross margin to about 56%, roughly flat on the June quarter. Rack-scale accelerator systems carry substantial third-party memory content, and if that mix dilutes margin as volumes scale, revenue growth stops translating into the earnings growth the valuation assumes.
Why did the Meta deal move AMD stock less than the OpenAI deal?
AMD gained 23.7% on 6 October 2025 when the OpenAI agreement was published, and 8.8% on 24 February 2026 when the structurally identical Meta agreement was announced. The most straightforward reading is that the market had already priced the template after the first deal and treated the second as confirmation rather than news.
This article is analysis, not investment advice. Figures are sourced from AMD’s filings with the US Securities and Exchange Commission and from market data pulled on 19 September 2026; price levels and earnings estimates are the author’s own. Capital is at risk and past performance does not indicate future results.