AMD's business selling chips for AI data centers has taken off, with that part of the company more than doubling in size over the past year.
Big customers like Meta and OpenAI have signed multi-year deals worth tens of billions of dollars for AMD's AI chips, giving real visibility into future growth.
But the stock price has already climbed so high that it is betting on nearly flawless execution for years to come, and it has run past what most Wall Street analysts think the shares are worth over the next year.
Overvalued.
A quick look at how a few well known investing styles might weigh this stock.
Investors who follow Cathie Wood's approach would point to data center revenue more than doubling year over year as proof AMD's AI chips are becoming a real second engine alongside its server processors, not just a side project. They would see the multi-year supply deals signed with major AI companies as evidence AMD is positioning itself at the center of the AI buildout, even if the payoff from those deals is still mostly ahead of it.
A Fisher-style investor would focus on AMD's long record of plowing cash into research and its string of real wins in server processors, where independent market trackers now put AMD above 46% of server chip revenue, up sharply from a few years ago. They would want to see that same execution carry through on the newer AI chip and embedded chip lines before fully trusting management's broader growth story.
Investors who follow Peter Lynch's approach would appreciate that AMD's story is easy to follow: it keeps taking processor business away from its longtime rival Intel and is now fighting for a slice of the AI chip market. They would also flag that the stock's price has run far ahead of current profits, so a Lynch-style investor would want the growth to keep showing up in actual earnings, not just announcements, before calling it cheap.
A Druckenmiller-style investor would view AMD as a direct way to ride the global buildout of AI computing capacity, with data center sales now making up more than half of company revenue and growing at triple digit rates. They would also watch closely for signs the AI spending cycle is slowing, since this kind of thematic position can reverse quickly if the broader buildout cools.
A Buffett-style investor would be cautious here: chipmaking is a capital intensive, boom and bust business, and AMD's profits have swung wildly over the past decade as demand rose and fell. They would also note that AMD still earns a much lower return on the money invested in the business than its larger rivals, which is not the kind of durable, high return moat this style of investing looks for.
These takes are illustrations of well known investing styles applied to this stock. They are informed guesses, not the actual views, statements, or holdings of the people named.
AMD is clearly delivering on the parts of its story tied to processors it already ships at scale, server chips and overall PC chips, where independent industry trackers confirm real, growing market share taken from Intel. The newer AI chip story is real but much earlier stage: AMD's AI chips are shipping and growing fast, but they remain a small fraction of the AI chip market, and its largest AI supply agreements are future commitments that have not yet turned into delivered revenue. Meanwhile the embedded chip business bought in 2022 has shrunk for two straight years, falling well short of the growth it was expected to add.
Independent chip market tracker Mercury Research reported AMD reached a record 46.2% of server processor revenue and 33.2% of server chip unit shipments in the first quarter of 2026, up from 27.2% of units a year earlier, confirming the gains from outside sources rather than AMD's own claims.
AMD's data center revenue, which includes these AI chips, more than doubled year over year to $6.7 billion in the second quarter of 2026, and major cloud providers have committed to offering AMD's newest AI chips. Independent market estimates still put AMD's AI chip line at only roughly 5% to 7% of the AI chip market versus Nvidia's 75% to 81% share, so the business is growing quickly but remains a minority challenger in scale.
The companies' own October 2025 announcement confirms the agreement, but the first phase, only 1 gigawatt, is not scheduled to begin shipping until the second half of 2026, and AMD's stock warrants tied to the deal vest only as OpenAI hits specific technical and purchase milestones over several years. No independent evidence yet shows revenue from this agreement actually landing in AMD's results.
AMD's own reported embedded segment revenue fell from $5.3 billion in 2023 to $3.6 billion in 2024 and to $3.5 billion in 2025, a two year decline driven by customers working down excess inventory in industrial and communications markets, the opposite of the growth this acquisition was expected to deliver.
Based on public evidence such as filings, customer announcements, and delivery records.
Most of AMD's revenue now comes from its data center business, selling server processors and AI accelerator chips to cloud companies for the computers that run websites and AI services. A second segment sells processors for everyday laptops and desktops plus the custom chips found inside game consoles like the PlayStation and Xbox. A third, smaller segment sells specialized chips, built largely from its 2022 purchase of Xilinx, used in factory equipment, cars, telecom gear, and aerospace systems.
We'd revisit this call if the price falls below $250.00 or rises above $1.1K.
These are review triggers for us to take another look, not price targets or a promise the stock will reach them.
| Metric | AMD (subject) | Peer median |
|---|---|---|
| EV/Sales | 29.76x | 26.32x |
| EV/EBIT | 279.09x | 68.11x |
| EV/EBITDA | 279.09x | 66.87x |
| P/B | 16.40x | 18.53x |
| Gross Margin | 49.5% | 57.0% |
| ROIC | 6.1% | 14.7% |
| ROE | 6.9% | 18.7% |
| Rule of 40 | N.M. | 21.3% |
AMD trades at $633.91 (Oct 2, 2026 close), market cap near $1.03 trillion, against trailing twelve month revenue of $41.3 billion (Q3 2025 through Q2 2026) and trailing twelve month free cash flow of about $7.83 billion. Trailing GAAP price to earnings is roughly 162x, price to sales roughly 25x, price to trailing free cash flow roughly 131x. These trailing multiples are extreme in isolation but reflect a mix that shifted hard toward AI data center hardware inside the trailing window; they compress quickly if the 2027 guidance below is met.
Q2 2026 revenue was $11.5 billion, up 50 percent year over year, GAAP gross margin 54 percent (non-GAAP 56 percent), GAAP operating income $2.0 billion, GAAP net income $2.3 billion, GAAP diluted EPS $1.38 (non-GAAP $1.66). Data Center revenue of $6.7 billion (58 percent of total) grew 107 percent year over year and 16 percent sequentially; Client and Gaming was $3.8 billion (up 6 percent, with Client up 23 percent but Gaming down 31 percent on weak semi custom console demand); Embedded was $977 million (up 19 percent). Q3 2026 guidance is about $13.0 billion, give or take $300 million, 41 percent year over year growth, non-GAAP gross margin about 56 percent, with continued double digit growth expected in Data Center and Embedded alongside a weaker Gaming segment.
Trailing twelve month free cash flow of about $7.83 billion is built from Q3 2025's $1.53 billion, Q4 2025's $2.1 billion, Q1 2026's $2.6 billion and Q2 2026's $1.6 billion. The sequential dip in Q2 2026 reflects inventory and capacity build ahead of the Instinct MI400 and Helios ramp, not weaker demand. Full year 2025 operating cash flow from continuing operations was $6.49 billion.
Cash and short term investments were $13.1 billion against total debt of $3.2 billion at the end of Q2 2026, a comfortably net cash balance sheet that can fund a GPU ramp without external financing.
Data center year over year growth: 107 percent (Q2 2026) versus 57 percent (Q1 2026) versus 39 percent (Q4 2025) versus 22 percent (Q3 2025), an accelerating trend as Instinct MI350 and MI400 shipments scale. Non-GAAP gross margin: 56 percent (Q2 2026) versus 57 percent (Q4 2025), broadly stable through the product transition. Free cash flow margin: 14 percent (Q2 2026) versus 25 percent (Q1 2026), lumpy around inventory builds. AMD's share of AI accelerator revenue was roughly 5 to 7 percent in 2025 against Nvidia's estimated 75 to 81 percent in 2026 (UNAVAILABLE: a precise, directly comparable 2026 AMD specific share figure). Nvidia's data center revenue was $193.7 billion for its fiscal 2026, roughly 13 to 15 times AMD's quarterly data center run rate.
Nvidia remains the dominant AI accelerator supplier by a wide margin and trades at a far lower forward multiple, about 24x forward earnings, than AMD's own forward multiple of about 40x on 2027 non-GAAP consensus EPS near $15.6, even though Nvidia's data center business is roughly an order of magnitude larger. That gap signals the market is pricing AMD's growth rate rather than its current scale, leaving little room for AMD's growth to merely match, rather than beat, Nvidia's trajectory.
The AI infrastructure buildout, hyperscaler capital spending, sovereign AI deals, and now direct compute for equity structures like the Meta and OpenAI warrant deals, is the dominant demand driver for both AMD and Nvidia. A slowdown in hyperscaler capex growth or a funding event at a major AI lab customer is the single largest macro swing factor for this stock.
CEO Lisa Su has sold stock regularly through 2026 under a Rule 10b5-1(c) plan adopted September 9, 2025 (for example 125,000 shares on June 10, 2026 for $57.6 million, plus further sales in August and September 2026 at prices between about $475 and $510 per share). These are pre scheduled, plan based sales rather than a fresh bearish signal, but the cumulative dollar amount is large and worth monitoring for plan amendments.
Catalysts: Q3 2026 earnings expected around November 3, 2026; initial MI400 and Helios volume shipments ramping through Q4 2026 and into 2027; potential new hyperscaler deals beyond Meta and OpenAI. Risks: execution risk on a brand new rack scale platform and chip family ramping at volume for the first time; revenue concentration in two mega deals; renewed or expanded China export licensing uncertainty on MI308 products, including a possible U.S. government revenue share claim that has not yet been formalized in regulation; the ROCm software ecosystem still trails CUDA.
Average sell side 12 month price targets tracked in early October 2026 (about $567.63 to $618.51 depending on source) sit below the current $633.91 price, meaning the market has already run past Wall Street's own targets. The 320 million performance warrants issued to Meta and OpenAI (roughly 20 percent dilution at full vest) carry a $600 stock price vesting hurdle that the stock has now cleared, making dilution a live rather than hypothetical factor. Falsification: if Q4 2026 or Q1 2027 data center revenue growth decelerates meaningfully below the 100 percent plus run rate implied by management's 2027 doubling target, or 2027 non-GAAP EPS consensus stalls well under $15 instead of climbing toward management's own above $20 framing, the growth priced into the current level would be falsified.
Bear case (ramp delays, Nvidia retains share): 2027 non-GAAP EPS near $10, 25x multiple, about $250 a share. Base case (guided 2027 targets met, no major slip): 2027 EPS near $16, 38x multiple, about $610 a share. Bull case (AMD executes ahead of plan, takes real share): 2027 EPS near $23, 46x multiple, about $1,050 a share. The current price sits almost exactly at the base case.
Mechanical model output, a screen, not a fair value.
Fair value is triangulated from scenario-weighted 2027 non-GAAP EPS outcomes ($10 bear, $16 base, $23 bull, versus a Street consensus near $15.6 and management's own framing of EPS materially above $20) paired with growth-confidence price to earnings multiples (25x, 38x, 46x), then adjusted for the roughly 20 percent share dilution from the Meta and OpenAI warrants that are already economically in the money above $600 a share.
The research read is more favorable than the mechanical screen's output.
Confidence = 0.7 × research sub-score + 0.3 × mechanical sub-score. Mechanical sub-score rewards the Layer 1 screen for agreeing with Layer 2's direction (corroboration), not just for being conviction=NORMAL in isolation.