AMD (NASDAQ: AMD): The Best Growth Story in Chips Trades at Three Times Nvidia's Multiple
- Hawkmont Research
- 6 days ago
- 15 min read
Updated: 4 days ago
The Best Data Center Growth Story on the Market Bought Its Two Biggest Customers With Its Own Stock
Published August 3, 2026, ahead of AMD's Q2 2026 earnings report, scheduled for August 4.
Company Snapshot
Metric | Value |
Ticker / Exchange | NASDAQ: AMD |
Recent price (Aug 2, 2026) | $472.15 |
52-week range | $149.22 - $584.73 |
All-time high (close) | $580.91 (June 30, 2026) |
Approx. market capitalization | ~$770 billion |
Shares outstanding | ~1.63 billion |
Trailing P/E | ~156-174x (source-dependent) |
Forward P/E | ~58.5x |
PEG ratio | 1.08 |
EV/EBITDA | ~105x |
Debt/equity | 0.06, effectively low leverage |
ROE / ROIC | 8.06% / 7.83% |
Beta | 2.47 |
Next earnings report | August 4, 2026, after market close |
52-week price change | Approximately +196% |
Change from all-time high | Approximately -19% |
Analyst consensus | "Strong Buy," average price target ~$575 (~16% implied upside) |
Table of Contents
Executive Summary
Business Overview
Deep Dive: The Vendor Financing Web, When Your Two Biggest Customers Become Your Two Biggest Shareholders
Deep Dive: China, Export Controls, and the TSMC Dependency
Financial Snapshot
Competitive Positioning
Valuation
Bull and Bear Scenarios
Risks and Sensitivities
Conviction and Portfolio Framework
Appendix: Sources and Comparables
Disclosures and Important Notices
1. Executive Summary
AMD has spent the past year converting itself from a credible but distant number two in AI silicon into the only company on earth that has signed multi-gigawatt, multi-year AI accelerator commitments with two of the industry's most important buyers, OpenAI and Meta, each covering up to 6 gigawatts of compute. Data Center revenue grew 57% year over year in the first quarter of 2026 to $5.8 billion, and management guided second-quarter revenue to $11.2 billion, comfortably ahead of Street expectations. On the numbers alone, this is one of the strongest growth stories in large-cap technology.
It is also a growth story AMD partly financed with its own equity. Both the OpenAI and the Meta agreements include performance-based warrants for up to 160 million AMD shares apiece, vesting as each customer hits deployment milestones and as AMD's stock price clears specific targets. Combined, that is warrants for up to roughly 320 million shares, against approximately 1.63 billion shares currently outstanding, tied directly to the same two customers whose demand is driving the growth numbers the market is now paying up for. This is not a hidden fact; AMD has disclosed it clearly on earnings calls. But it is a structural feature of the current AI infrastructure buildout, vendor financing dressed up as strategic partnership, that deserves the same scrutiny investors have started applying to similar arrangements elsewhere in the AI supply chain.
The stock has priced in a great deal of this optimism already. AMD trades at roughly 58 times forward earnings and roughly 105 times EV/EBITDA, multiples that sit well above Nvidia's own forward P/E of roughly 20-24 times and EV/EBITDA near 28 times, despite Nvidia holding an estimated 80%-plus share of the AI accelerator market against AMD's mid-single-digit share. Return on equity and return on invested capital at AMD, 8.06% and 7.83% respectively, remain a fraction of Nvidia's 114% and 105%. The market is not pricing AMD as a proven, high-return incumbent. It is pricing AMD as the most credible challenger to that incumbent, on the assumption that today's single-digit market share becomes something much larger over the next several years.
Layered on top of the growth and valuation story is a real, recurring geopolitical risk that has already cost the company real money once. US export restrictions on AI accelerators to China forced AMD to take an $800 million inventory charge and absorb a $1.5-1.8 billion revenue hit in 2025, roughly 10% of that year's data center business, before a partial license reversal allowed limited MI308 shipments to resume. That policy environment has proven genuinely volatile, reversing direction multiple times within a single year, and AMD's China exposure remains a live, unresolved variable rather than a settled one.
None of this diminishes the underlying operating story: Data Center is now unambiguously the center of AMD's business, EPYC server CPUs continue taking share from Intel, and the Instinct GPU roadmap has, for the first time, produced deals large enough that analysts are genuinely reconsidering AMD's long-run AI accelerator share. This note treats that operating strength and the structural risks around vendor financing, valuation, and China exposure as equally real, and asks readers to hold both at once rather than defaulting to either the bull or the bear framing alone.
2. Business Overview
AMD operates across three reporting segments:
Data Center: server CPUs (EPYC), data center GPUs and AI accelerators (Instinct), DPUs, and FPGAs from the Xilinx and Pensando acquisitions. This is now the company's largest and fastest-growing segment, and the one driving the current valuation debate.
Client and Gaming: Ryzen client CPUs, gaming GPUs (Radeon), and semi-custom console chips. A more mature, cyclical business that still grew 23% year over year in Q1 2026.
Embedded: FPGA and embedded processor products for industrial, automotive, aerospace, and communications customers, which returned to growth in Q1 2026 after a period of inventory-driven softness.
The Data Center segment's growth engine has two components: continued EPYC server CPU share gains against Intel (fifth-generation "Turin" EPYC accounted for more than half of total server revenue by late 2025), and the ramp of the Instinct GPU line, specifically the MI350 series currently shipping and the next-generation MI450 (built on CDNA 5 architecture at TSMC's 2-nanometer node with HBM4 memory) beginning volume shipments in the second half of 2026 as part of the Helios rack-scale AI system. AMD has stated a target of $20 billion in annual data center GPU revenue within roughly two years of the MI350 ramp beginning, a target that now looks more credible given the OpenAI and Meta commitments described below, though it remains a target rather than a delivered result.
3. Deep Dive: The Vendor Financing Web, When Your Two Biggest Customers Become Your Two Biggest Shareholders
In October 2025, AMD and OpenAI signed a multi-year, multi-generation agreement for OpenAI to deploy up to 6 gigawatts of AMD Instinct GPUs, beginning with a 1-gigawatt MI450 rollout in the second half of 2026. AMD expects the relationship to generate tens of billions of dollars in revenue over its life. To align incentives, AMD issued OpenAI warrants for up to 160 million AMD shares, vesting in tranches tied to deployment milestones (1 GW, then progressively toward 6 GW) and to AMD's own stock price clearing specific targets.
In February 2026, AMD struck a structurally similar agreement with Meta: another 6-gigawatt, multi-year, multi-generation commitment, again beginning with a 1-gigawatt MI450 deployment in the second half of 2026, again accompanied by a performance-based warrant for up to 160 million AMD shares. Management described the Meta deployment as expected to generate "double-digit billions of dollars per gigawatt" in data center AI revenue, and confirmed the Meta and OpenAI agreements are structured independently with no overlap.
Why this matters beyond the headline demand numbers. Combined, the two warrant grants cover up to roughly 320 million AMD shares, a meaningful figure against approximately 1.63 billion shares currently outstanding, all tied to the purchasing commitments of exactly two customers. This is economically similar to a pattern that has drawn increasing scrutiny elsewhere in the AI infrastructure buildout: a chip supplier extending equity-linked incentives to its largest buyers, which has the effect of making customer demand and supplier valuation partially self-reinforcing. If AMD's stock rises because the OpenAI and Meta deals are perceived as validating its AI roadmap, and a portion of the warrant vesting is explicitly tied to that same stock price, the demand signal and the valuation signal are not fully independent of one another.
What it is not. These are disclosed, negotiated commercial agreements with real, contracted deployment volumes and real revenue attached, not the more opaque circular financing structures (vendor loans funding customer purchases of the vendor's own product) that have drawn the sharpest criticism in the broader AI infrastructure sector. AMD's warrants compensate customers for demand commitment and technical collaboration; they do not, on current public disclosure, fund the customers' purchases directly.
What investors should track. The pace and size of warrant vesting as deployment milestones are hit, the potential dilution impact if AMD's stock clears the price targets embedded in the warrant structure, and whether either OpenAI or Meta's own capital spending plans show any signs of strain that would affect their ability to honor multi-year, multi-gigawatt commitments. Concentration in two customers for a growth narrative this large is, on its own, worth pricing as a distinct risk factor, independent of how the warrant mechanics specifically work.
4. Deep Dive: China, Export Controls, and the TSMC Dependency
A volatile, reversing policy environment. US export restrictions on advanced AI accelerators to China have changed direction multiple times within the past 18 months. An April 2025 ban on all AI GPU exports to China was followed by a July 2025 reinstatement of licenses for Nvidia's H20 and AMD's MI308, then by informal guidance from Chinese authorities discouraging domestic firms from ordering the newly re-licensed US chips, then by a December 2025 policy shift allowing further, though still restricted, shipments. AMD took an $800 million inventory charge and absorbed a $1.5-1.8 billion revenue hit in 2025, roughly 10% of that year's projected data center business, when the initial ban hit. The company subsequently received "some" export licenses for MI308 shipments to China and booked approximately $390 million of MI308 revenue in the fourth quarter of 2025, revenue it had explicitly excluded from its own guidance until the licenses were confirmed.
AMD's actual China position is small and contested. Independent unit-share tracking put AMD at roughly 4% of China's domestic AI chip market by early 2026 (about 160,000 units), well behind Nvidia's declining but still larger share and dramatically behind Huawei's roughly 20% share and rapidly scaling Ascend accelerator line. Beijing has separately directed newly built, state-funded data centers to use only domestically produced chips, a policy that structurally caps the addressable China opportunity for both AMD and Nvidia regardless of what export licenses Washington grants.
The TSMC dependency runs the other direction too. AMD is fabless and depends entirely on TSMC for advanced-node manufacturing (the MI450 is built on TSMC's 2-nanometer process) and for CoWoS advanced packaging capacity, which has been a binding industry-wide constraint. AMD reportedly secured roughly 11% of TSMC's 2026 CoWoS allocation, a meaningful but finite share that AMD does not control and that TSMC also allocates to Nvidia and other customers. Any tightening of that allocation, independent of AMD's own execution, is a direct constraint on how fast the MI450/Helios ramp can actually happen.
Net effect. China is a real, if currently small, swing factor in AMD's data center revenue, one that has already moved in both directions within a single year for reasons entirely outside AMD's control, and the company's ability to fully capture its own roadmap depends on advanced packaging capacity it does not own.
5. Financial Snapshot
Full-year 2025 (reported) and recent quarters:
Metric | FY2025 | Q4 2025 | Q1 2026 |
Total revenue | $34.6B (+34% YoY) | $10.3B (+34% YoY) | $10.3B (+38% YoY) |
Data Center revenue | n/a | $5.4B (+39% YoY, record) | $5.8B (+57% YoY) |
Data Center operating income | n/a | $1.8B (33% margin) | $1.6B (vs. $932M PY) |
Client and Gaming revenue | n/a | n/a | $3.6B (+23% YoY) |
Non-GAAP gross margin | 52% | ~55%* | 55% |
Non-GAAP EPS | $4.17 | n/a | $1.37 (vs. consensus, a beat) |
GAAP gross margin | n/a | n/a | 53% |
GAAP net income | n/a | $2.5B (+42% YoY) | $1.4B |
GAAP diluted EPS | n/a | n/a | $0.84 |
Free cash flow | n/a | $2.1B | n/a |
*Q4 2025 non-GAAP gross margin benefited from a roughly $360 million reserve release and roughly $390 million of MI308 China revenue; excluding both, management characterized underlying gross margin as approximately 55%.
Q2 2026 guidance (reporting August 4, 2026): Revenue of approximately $11.2 billion, plus or minus $300 million, notably above the roughly $10.5 billion Street consensus modeled ahead of the print, with non-GAAP gross margin guided at approximately 55%.
Balance sheet and capital return: AMD ended Q1 2026 with $9.2 billion remaining under its share repurchase authorization, low leverage (debt/equity of 0.06), and trailing-twelve-month operating cash flow of $9.73 billion against capital expenditures of just $1.15 billion, yielding $8.57 billion of free cash flow.
Reading the numbers: The gap between AMD's Data Center segment operating margin (33% in Q4 2025, roughly 28% in Q1 2026) and its consolidated GAAP operating margin (11.78% trailing twelve months) is large, and reflects corporate overhead, amortization of intangibles from the Xilinx and Pensando acquisitions, and stock-based compensation charged at the consolidated level. Investors comparing AMD's segment-level profitability (genuinely strong) to its consolidated GAAP profitability (comparatively modest, and the basis for its very high trailing P/E) should be explicit about which number they are using, since the two tell meaningfully different stories about the same quarter.
6. Competitive Positioning
AI accelerators. Nvidia remains the dominant player in AI accelerators by a wide margin. Independent estimates put Nvidia's AI accelerator market share at a peak of roughly 87% in 2024, slowly giving ground since, while AMD's Instinct line generated an estimated $7-8 billion in 2025, roughly 5-7% of total AI accelerator revenue. Critically, AMD's average selling price (roughly $29,000 per GPU by one estimate) runs below Nvidia's (roughly $33,000), meaning Nvidia's revenue share exceeds even its already-dominant unit share. The OpenAI and Meta commitments described above are the strongest evidence yet that AMD's share could move meaningfully higher, with one industry estimate suggesting AMD's CoWoS packaging allocation alone could support a "mid-teens" AI accelerator market share target, but that remains a forward-looking projection, not a delivered outcome.
Server CPUs. AMD's EPYC line continues taking share from Intel in the data center CPU market, with fifth-generation "Turin" EPYC processors accounting for more than half of AMD's total server revenue by late 2025, a genuinely durable competitive position built over several product generations.
The custom silicon threat. A structurally important risk to both AMD and Nvidia: hyperscalers are increasingly designing their own AI accelerators to reduce dependence on merchant GPU vendors entirely. Google's TPU v6, Amazon's Trainium2, and Microsoft's Maia 100 all compete for internal workloads that would otherwise be addressable Instinct or Nvidia GPU demand, and this trend does not require AMD or Nvidia to lose a single competitive bid to erode the total addressable market both are pricing against.
China's domestic alternatives. Huawei's Ascend line, alongside Cambricon and Alibaba's T-Head, are scaling rapidly in China specifically because both AMD and Nvidia face export restrictions there, giving domestic Chinese vendors a protected, policy-driven runway that is unlikely to reverse regardless of future US export-license decisions.
7. Valuation
At a recent price of $472.15 and roughly 1.63 billion shares outstanding, AMD carries a market capitalization of approximately $770 billion.
Multiple | AMD | Nvidia (for context) |
Forward P/E | ~58.5x | ~20-24x |
Trailing P/E | ~156-174x | ~30x |
PEG ratio | 1.08 | 0.44 |
EV/EBITDA | ~105x | ~28x |
ROE | 8.06% | 114.29% |
ROIC | 7.83% | 104.67% |
Debt/equity | 0.06 | 0.07 |
The contrast in the table above is the central valuation tension in this stock. Nvidia, the dominant incumbent with an estimated 80%-plus share of AI accelerator revenue and returns on equity and invested capital above 100%, trades at a materially lower forward multiple than AMD, whose returns on equity and invested capital sit in the high single digits. The market is not pricing AMD as a proven, high-return business the way it prices Nvidia. It is pricing AMD as an option on a much larger future market share, one where the OpenAI and Meta commitments are treated as strong evidence that option is likely to pay off, discounted against AMD's currently modest actual returns on capital.
What appears priced in: successful, on-schedule MI450/Helios ramp in the second half of 2026, continued EPYC share gains against Intel, and a trajectory toward AMD's own stated target of $20 billion in annual data center GPU revenue within roughly two years of the MI350 ramp. What appears comparatively under-priced, relative to how the market treats similar structural features elsewhere: the customer-concentration and equity-alignment risk embedded in the OpenAI and Meta warrant structures, and the binary, policy-driven volatility of China export rules that has already produced an $800 million charge once in the past 18 months and could do so again in either direction with little warning.
Consensus sell-side sentiment remains bullish, with an average price target near $575 (roughly 16% above the recent price) and a "Strong Buy" rating, reflecting confidence that the Data Center ramp will continue to outrun even AMD's own upgraded guidance, as it has in each of the last several quarters.
8. Bull and Bear Scenarios
Bull case.
AMD has, for the first time, secured demand commitments large enough (6 gigawatts each from OpenAI and Meta, on top of continued hyperscaler EPYC adoption) that a genuine, durable shift in AI accelerator market share away from Nvidia's near-monopoly looks plausible rather than aspirational. Data Center revenue growth of 57% year over year and a Q2 guide well above Street consensus suggest the ramp is, if anything, ahead of where the market modeled it entering the year. A low-leverage balance sheet, $8.57 billion of trailing free cash flow, and a large remaining buyback authorization give AMD financial flexibility to keep investing in the MI450/Helios roadmap without external financing pressure. Under this case, today's elevated forward multiple is justified by a market-share trajectory that closes much of the return-on-capital gap with Nvidia over the next several years as Data Center GPU revenue scales toward AMD's own $20 billion target.
Bear case.
AMD's current returns on equity and invested capital, in the high single digits, do not yet support the multiple the market is paying, which sits well above Nvidia's despite Nvidia's vastly superior actual profitability. A meaningful share of the growth narrative rests on two customers, OpenAI and Meta, whose purchasing commitments are partially equity-linked through warrant structures that make AMD's own stock price and its two largest customers' continued goodwill more intertwined than a typical arm's-length commercial relationship. China policy has already cost AMD real money once in the past 18 months and remains genuinely unpredictable, capable of moving materially in either direction with little advance warning. The custom silicon threat from Google, Amazon, and Microsoft caps the addressable market both AMD and Nvidia are competing for, regardless of how the AMD-versus-Nvidia share battle plays out. And AMD's own advanced packaging supply, the binding constraint on how fast it can actually ship the MI450 ramp the market is pricing in, is allocated by TSMC and not fully within AMD's control.
Both scenarios describe the same ramp. The disagreement is over how much of it is already reflected in a stock trading at roughly three times Nvidia's forward multiple with a fraction of Nvidia's returns on capital.
9. Risks and Sensitivities
Prioritized, not exhaustive:
Valuation risk relative to demonstrated returns. Forward multiples several times Nvidia's, alongside return-on-equity and return-on-invested-capital figures far below Nvidia's, leave limited room for any deceleration in the Data Center ramp without significant multiple compression.
Customer concentration through equity-linked deals. Up to 320 million shares in combined OpenAI and Meta warrants, tied to two customers whose demand is central to the current growth narrative and whose vesting is partly tied to AMD's own stock price.
China export-policy volatility. A policy environment that has reversed direction multiple times within 18 months, already producing an $800 million inventory charge and a $1.5-1.8 billion revenue hit once.
TSMC and CoWoS packaging dependency. AMD's ability to execute the MI450/Helios ramp is bound by advanced-node and packaging capacity it does not control and shares with Nvidia and other customers.
Custom silicon substitution risk. Google, Amazon, and Microsoft's internal AI accelerator programs reduce the addressable merchant GPU market independent of the AMD-Nvidia competitive dynamic.
GAAP-to-non-GAAP margin gap. A wide spread between strong segment-level profitability and comparatively modest consolidated GAAP margins, worth tracking for signs of convergence or divergence over time.
Execution risk on the MI450/Helios timeline, given the scale of the OpenAI and Meta commitments now riding on an on-schedule second-half 2026 ramp.
Beta and volatility. A beta of 2.47 and a stock that has already moved roughly 19% off its all-time high, evidence that even strong operating results do not fully insulate the shares from sharp repricing.
10. Conviction and Portfolio Framework
The following is offered as an interpretive framework for readers doing their own position-sizing work, not a recommendation.
AMD sits in an unusual position within an AI/semiconductor sleeve: it is neither the dominant, proven incumbent (that remains Nvidia, at a lower multiple and dramatically higher returns on capital) nor an unproven, pre-revenue challenger. It is a company with real, accelerating Data Center revenue and, for the first time, demand commitments large enough to make a genuine market-share shift plausible, priced at a multiple that already assumes a meaningful part of that shift happens on schedule.
The central sizing question is less about whether AMD's Data Center business is real (the Q1 2026 numbers say clearly that it is) and more about how much of the next several years of share gain is already reflected in a stock trading at roughly three times Nvidia's forward earnings multiple. Readers with high conviction in AMD's ability to execute the MI450/Helios ramp and convert the OpenAI and Meta commitments into durable, expanding market share have a stronger case for a larger position than readers who are primarily attracted to the growth headline without having formed a view on the valuation gap relative to Nvidia's own multiple and returns on capital.
Key monitoring points going forward:
Q2 2026 earnings (Aug 4): watch specifically for any commentary on MI450/Helios shipment timing and whether the 1-gigawatt initial deployments with OpenAI and Meta remain on schedule for the second half of the year.
Warrant vesting activity and any disclosure on cumulative dilution as OpenAI and Meta deployment milestones are hit.
China export-license developments, given the policy's demonstrated volatility in both directions.
TSMC CoWoS capacity allocation commentary, as the binding constraint on execution speed.
Return on invested capital trend over the next several quarters, as the clearest signal of whether AMD's returns are converging toward or diverging from the multiple it currently commands.
Join 100,000+ readers who get Hawkmont Research delivered straight to their inbox. Independent, conflict-free equity research, no paywall, no compensation from the companies we cover. Hawkmont Research is free and will stay free.
11. Appendix: Sources and Comparables
Primary sources: AMD Form 10-Q for the quarter ended March 28, 2026 (SEC EDGAR); AMD Q1 2026 and Q4 2025 earnings press releases and conference call transcripts.
Secondary sources referenced: CNBC, CNN, and Yahoo Finance pricing and earnings-date coverage; StockAnalysis.com, Robinhood, and MacroTrends valuation and price-history data; TECHi, TradingKey, and AlphaSense Q1 2026 earnings analysis; Futurum Group, IntuitionLabs, HyperFRAME Research, and Enertuition coverage of the OpenAI and Meta accelerator agreements; MarketsDaily, DailyPolitical, and TickerReport coverage of the Meta warrant structure; Tom's Hardware, TechFlow, OC3D, WCCFTech, and Benzinga coverage of China export-control developments; CommandLinux and Silicon Analysts AI GPU market-share estimates; GuruFocus, FinanceCharts, MacroTrends, FullRatio, and Motley Fool coverage of Nvidia's comparative valuation.
Illustrative public comparables:
Company | Relevant comparison |
Nvidia (NVDA) | Dominant AI accelerator incumbent; primary valuation and market-share benchmark |
Intel (INTC) | Server and client CPU competitor, losing share to AMD's EPYC and Ryzen lines |
Broadcom (AVGO) | Custom silicon and networking competitor in the AI infrastructure buildout |
Marvell (MRVL) | Custom silicon competitor for hyperscaler AI accelerator programs |
Huawei (private) | Dominant domestic China AI accelerator competitor, insulated by export policy |
12. Disclosures and Important Notices
This report was independently produced by Hawkmont Research. Hawkmont Research has not received, and does not accept, compensation from Advanced Micro Devices, Inc., its affiliates, officers, directors, or any party with a financial interest in AMD securities in connection with this report. Hawkmont Research holds no position in AMD securities as of the date of publication.
This report is provided for informational and educational purposes only. It does not constitute, and should not be construed as, investment advice, a recommendation to buy or sell any security, or an offer or solicitation in any jurisdiction. Nothing in this report accounts for any individual reader's financial situation, risk tolerance, or objectives. All figures are drawn from public filings and secondary reporting as cited above and are subject to revision, restatement, or correction by the company in subsequent filings; readers should independently verify all data points against primary source filings before making any investment decision.
Forward-looking statements regarding guidance, product roadmaps, and upcoming earnings are based on company disclosures as of the date of this report and are subject to change. Past price performance is not indicative of future results.
Readers must form their own independent conclusions and are encouraged to consult a licensed financial or legal advisor before making any investment decision involving AMD or any other security discussed in this report.
