AMD’s 106 Times Multiple Meets a 14 Gigawatt Ramp

AMD last closed at $608.10, up 183.95% in 2026, which values the company at $993 billion after a September print at $1 trillion. On 2026 earnings the stock trades at 106 times, against 22.93 times for Nvidia and 39.09 times for Broadcom.

The gap is a 2026 snapshot. The same $608.10 price is 45.1 times 2027 earnings estimates of $13.49 a share, if OpenAI, Meta, and Anthropic actually take the racks those estimates assume.

A 106 Times Multiple on This Year’s Profit

The case against holding AMD into 2027 starts with the tape, not the product roadmap. Shares hit $658.52 on October 6, then slipped as investors marked the year as three-quarters done and asked what a second encore would even look like. At $608.10 the company is already a near-trillion-dollar stock. Another triple would put it near $3 trillion, a size that only a handful of firms occupy, and only if every other giant stood still.

Peers in the same AI buildout are cheaper on this year’s profit. Nvidia closed at $229.28, down 0.52%, with a current-year multiple of 22.93 times and a market value of $5,536 billion. Broadcom sits at 39.09 times with a value of $1,726 billion. AMD’s 106 times multiple is the outlier in that set, and it is the number bears keep putting on the table.

The other knock is sequencing. AMD announced large hyperscaler supply pacts in 2025 and early 2026, then watched the stock re-rate as those chips showed up in live inference fleets. Paying twice for the same purchase order is a bad habit. If 2027 is only more headlines about the same OpenAI and Meta metal, the multiple has little room left to expand, and holders need the business to grow into the price.

Wall Street’s 2027 Estimates Put AMD at 45 Times

Chip multiples look absurd at the top of a ramp and ordinary once the revenue lands. MarketScreener’s compiled estimates, taken at the $608.10 close, show that path in three columns. Fiscal 2026 sales sit at $50.948 billion and earnings at $5.728 a share, which is the 106 times print. Fiscal 2027 sales rise to $88.942 billion and earnings to $13.49, which is 45.1 times 2027 earnings. Fiscal 2028 sales go to $123.765 billion and earnings to $20.39, or 29.8 times.

AMD EARNINGS PATH AT $608.10

Those 2027 earnings are 136% above the 2026 figure, more than double, while sales rise 75%. If the stock goes nowhere and the estimates hold, the multiple compresses because profit showed up, not because the share price collapsed. That is the version of “history” the rich-multiple argument skips. It is also an estimate, not a shipment. Miss the $13.49 and 45.1 times becomes a memory, and 106 times is the number that sticks.

Nvidia can still grow faster from a much larger base and still look cheaper on this year’s books. Nothing in the table erases that. It does change what “expensive” means if 2027 is the year Instinct racks stop being a press release and start being a line on the income statement.

When Do the OpenAI and Meta Racks Hit Revenue?

AMD has now named three gigawatt-scale GPU customers, for 14 gigawatts in total. The dates matter more than the round number, because the first metal is only now scheduled to move.

THE HYPERSCALER CLOCK

1. October 6, 2025: AMD and OpenAI announce a 6 gigawatt, multi-year, multi-generation GPU pact, with the first 1 gigawatt of Instinct MI450 chips set to begin in the second half of 2026.
2. February 24, 2026: AMD and Meta announce a matching 6 gigawatt Instinct deal on Helios racks, with first-gigawatt shipments also set for the second half of 2026 on a custom MI450-based GPU and 6th Gen EPYC “Venice” CPUs.
3. July 22, 2026: AMD and Anthropic announce up to 2 gigawatts of MI450-series GPUs in Helios, with the first gigawatt beginning in the first half of 2027, plus an AMD equity commitment of up to $5 billion.
4. Second half of 2026: First OpenAI and Meta gigawatt shipments are scheduled to start; AMD later said OpenAI’s Helios phase would come online in the fourth quarter and accelerate through 2027.

Jean Hu, AMD’s chief financial officer, put a dollar range on the OpenAI pact that the stock had already begun to treat as certain. He did not call it a second helping of 2025 news. He called it future sales.

Our partnership with OpenAI is expected to deliver tens of billions of dollars in revenue for AMD while accelerating OpenAI’s AI infrastructure buildout. This agreement creates significant strategic alignment and shareholder value for both AMD and OpenAI and is expected to be highly accretive to AMD’s non-GAAP earnings-per-share.

Jean Hu, EVP, CFO and treasurer, AMD, on the 6 gigawatt OpenAI GPU deal

Meta’s release was thinner on dollars and clearer on use. Mark Zuckerberg, Meta’s founder and chief executive, said the company was forming a long-term partnership “to deploy efficient inference compute and deliver personal superintelligence,” and that he expects AMD “to be an important partner for many years to come.” Lisa Su, AMD’s chair and chief executive, said the Meta work spans Instinct GPUs, EPYC CPUs, and rack-scale systems built for Meta’s loads. Meta already runs millions of EPYC CPUs and large MI300 and MI350 fleets, so this is an expansion of a live account, not a cold start.

Patrick Moorhead, a former AMD corporate vice president who runs Moor Insights & Strategy, wrote after the Meta announcement that the first gigawatt is binding and the remaining five gigawatts are contingent on milestones. That split is the execution risk inside the 14 gigawatt headline. The first racks have to ship on time, or the later tranches never vest and the 2027 estimates above them have to come down.

Penny Warrants That Vest Only on Shipments

OpenAI and Meta each received a warrant for up to 160 million AMD shares at $0.01, 320 million in all, against 1.63 billion shares outstanding. If every tranche is exercised, the share count rises about 19.6%. That dilution is the part of the deals the tape hates on quiet days. The vesting rules are the part that ties the customers to the loading dock.

HOW THE PENNY WARRANTS WORK

• Size and price: Each of OpenAI and Meta can buy up to 160 million shares at one cent, with five-year terms.
• Shipment gate: The first tranche vests with the first 1 gigawatt of Instinct shipments; later tranches vest as purchases scale toward 6 gigawatts.
• Price gate: Vesting also depends on AMD share-price targets, including a $600 level that the stock has already traded above, so deliveries are now the main switch.
• What does not vest: If a customer never takes the later gigawatts, those warrant slices never become shares, and the matching revenue never shows up either.

Hu said the OpenAI structure should add to non-GAAP earnings per share even after the extra shares. That only works if the GPU gross profit on each vested gigawatt outweighs the dilution. Anthropic’s pact is a different instrument, an AMD investment of up to $5 billion rather than a 160 million share coupon, with first silicon in the first half of 2027. The three customers are not one trade. They are three clocks.

OpenAI’s Private Revenue Now Moves the Stock

On October 8 a report put OpenAI’s annualized revenue near $50 billion, about $20 billion below figures that had circulated near $70 billion. AMD fell 3.90% that session and closed at $620.68. Nvidia, Oracle, and Microsoft sold off with it. By the October 9 close the stock was at $608.10, down 2.03%, or $12.58, from the prior session. No AMD order was cancelled in public. A private lab’s run-rate was enough.

That is the concentration the 14 gigawatt stack creates. AMD now trades, in part, as a listed proxy for customers that do not file 10-Qs. A booking scare at OpenAI hits the chip supplier the way a missed cloud quarter hits a landlord. The inverse is also true: if those labs keep building, AMD’s 2027 sales path is less of a wish and more of a delivery schedule. The October tape showed which way the lever still points.

Helios, the rack AMD is using to fill those orders, packs 72 Instinct MI455X GPUs and 18 6th Gen EPYC “Venice” CPUs. In the second-quarter release AMD said the rack is being deployed by Anthropic, Meta, Microsoft, OpenAI, Oracle, HUMAIN, and several smaller clouds. Named logos are not recognized revenue. They are a customer list waiting on watts, power feeds, and HBM.

Lisa Su Is Booking Wafers for 2027 Demand

The June quarter already showed what a data-center ramp looks like before the first OpenAI gigawatt is fully in. AMD reported a record $11.5 billion quarter, up 50% from $7.685 billion a year earlier and up 13% from $10.253 billion in the first quarter. Data-center sales were $6.7 billion, up 107%, and 58% of the company. Non-GAAP diluted earnings were $1.66 a share; GAAP earnings were $1.38, on net income of $2.3 billion. Client processors rose 23% to $3.1 billion, while gaming fell 31% to $779 million, so the mix is already an AI-and-server story with a shrinking console stub.

Gross margin is the dull risk inside those growth rates. Non-GAAP gross margin was 56% in the June quarter, after 55% in the first, and AMD guided to about 56% again for the third quarter. Units and average selling prices in servers rose at a double-digit rate, yet the margin barely moved, which means memory, packaging, and other inputs are eating the price increases. The year-ago comparison also had help: Q2 2025 included $800 million of inventory charges tied to U.S. export rules on Instinct MI308 GPUs, which had crushed that period’s GAAP margin to 40%.

We delivered an excellent quarter, with record revenue and profitability as Data Center revenue more than doubled year-over-year. We enter the second half with strong momentum as EPYC demand accelerates, Instinct deployments scale and Helios begins to ramp.

Dr. Lisa Su, chair and CEO, AMD second-quarter 2026 results

For the third quarter AMD guided to about $13 billion of sales, plus or minus $300 million, roughly 41% higher than a year earlier and 13% higher than the June quarter, with non-GAAP gross margin still near 56%. On the August call Su said data-center revenue should more than double in 2027 and that the server CPU market AMD is chasing could reach about $220 billion by 2030. In Taipei on October 6 she told reporters demand is still higher than supply, that AMD will raise supply a lot in 2027, and that capacity planning with partners now runs three to five years out, including high-bandwidth memory and advanced packaging.

None of that settles 2027 for holders sitting on a 106 times 2026 multiple. It does define the test. The stock already paid for the announcement cycle. The next payment is supposed to arrive as Instinct silicon, Helios racks, and the first vested gigawatts, which is when $13.49 of earnings either shows up or does not. Su can book wafers. OpenAI and Meta still have to take them.

Disclaimer: This article is news reporting and analysis of AMD’s share price, published estimates, and announced customer contracts, and it is for information only. It is not investment advice, a recommendation to buy or sell AMD or any other security, or a forecast of 2027 returns. Readers should consult a licensed financial adviser or investment professional who can consider their own objectives and risk tolerance before acting on any figure in this piece. Prices, multiples, earnings estimates, and contract statuses reflect the sources cited as of the dates in the article and can change with the next close, filing, or customer update.