Micron, AMD and Broadcom jumped on the latest demand read. At 41x forward earnings, the valuation finally respects the data, and it is still short of the ceiling.
AMD is a simple thesis now. The headline read is that the stock is at 108x trailing earnings and 41x forward. For two years bears have pointed to those numbers as proof that AMD is priced for perfection. They are wrong. The forward earnings number is still too low. The multiple is rich but defensible, and the upside comes from the denominator rising faster than consensus.
This is the rare setup where a stock that has rallied 3x can still be a buy. The arithmetic is visible in the 2025 numbers and the 2026-2027 roadmap. Our view is constructive at current levels and aggressive below $270.
AMD revenue for calendar 2025 came in at $34.6 billion, up 34 percent from 2024. Net income scaled to $4.3 billion from $1.6 billion. Operating margin expanded from 7.4 percent to 17 percent in one year. Those numbers already reflect a genuine mix shift toward higher-margin data centre GPU revenue. What they do not fully reflect is the MI350 and MI400 roadmap that ramps through 2026 and 2027.
The latest demand read pushed Micron, AMD and Broadcom higher. That was not a sentiment move. That was the market pricing in a sharper back half of 2026 on the back of hyperscaler capex comments. AMD's specific exposure to that incremental demand is through the MI instinct family, where gross margin sits materially above the corporate average. Every incremental billion dollars of data centre GPU revenue adds roughly 55 to 60 percent of gross profit, versus the ~48 percent corporate gross margin average.
The second point is share. Nvidia dominates training. AMD has been gaining share in inference, where the price-performance calculus is more competitive and where the market is larger in unit terms. Inference grows faster than training through 2027 according to every major hyperscaler's internal capex allocation. AMD does not need to beat Nvidia. It needs to hold or grow share in inference while the market grows 40 to 60 percent per year. That is the base case.
Third point. AMD's server CPU business keeps compounding. The EPYC Genoa and Turin families have been gaining share from Intel steadily. That is a separate revenue line with its own cyclical tail that provides earnings stability even through GPU competition noise.
TickerXray Report
Get the complete AMD report with all 12 quantitative models, AI-generated investment thesis, and real-time data.
At 41x forward earnings the market is pricing roughly $7 per share in fiscal 2027 earnings. Our model has fiscal 2027 earnings in the $9 to $10 range if data centre GPU revenue scales as guided. That is a 25 to 40 percent upside on the earnings number alone. A 35x forward multiple on $9 per share is $315. At 40x it is $360.
Free cash flow compounded from $3.2 billion in 2021 to $6.7 billion in 2025. Capex ran at $974 million in 2025, which is low relative to the $25 to $30 billion range Nvidia is spending. That capex efficiency is a structural advantage. AMD does not own its own fabs. It rents capacity from TSMC. The capital intensity of the business model is materially below Intel's and below Nvidia's on a normalised basis. That keeps FCF conversion high.
Operating margin at 17 percent is still well below the Nvidia peer comp at 60 percent plus. Nobody expects AMD to reach Nvidia margins. But 25 percent by 2028 is entirely plausible as data centre mix grows. That would take operating income from $3.7 billion in 2025 to $15 billion plus by 2028.
The bear case is simple. Nvidia dominates, and AMD is always just behind. In the training market that is true. In inference the gap is narrower. Any broad AI capex cut by hyperscalers would hurt AMD. Intel's eventual return to competitiveness would pressure the CPU business. All of this is priced at some level.
The bearish scenario where AMD underperforms Nvidia on MI400 execution and simultaneously loses CPU share to Intel is the only scenario in which the $7 consensus earnings number proves accurate. That is a low probability outcome. It is not zero, but it is not the base case.
AMD is a buy. The forward earnings number is conservative, the roadmap is delivering, and the multiple premium is justified by the structural shift into inference. Our fair value range is $320 to $360 based on 36x forward against 2027 estimates. That is 12 to 24 percent upside.
We're buyers below $270. The catalyst is the MI350 launch cadence and the hyperscaler capex print in the next two quarters. Historically, the second year of an AI compute cycle belongs to the fast follower, not the leader. Nvidia owns the current cycle. AMD is positioned to capture the back half.
Full forensic analysis of AMD
+ 6 more models included
150,000+ stocks covered
Global coverage across 60+ exchanges. Every report includes all 12 quantitative models and AI analysis.
View plansEvery report runs 12 quantitative models and generates an AI investment thesis. From Piotroski scores to manipulation detection -- get the full picture in seconds.
12 forensic models
Piotroski, Altman, Beneish, DuPont & more
AI investment thesis
Synthesized outlook on every stock
Manipulation detection
Spot red flags before they hit the news
150,000+ tickers
Global coverage across 60+ exchanges
Expected return
Forward return projections for every stock
Real-time data
Live prices, insider trades, news sentiment
Free accounts get 1 report per month. Pro gets unlimited.
Our earlier piece 'AMD's Longest Winning Streak Since 2005 Tells a Bigger Story' flagged pattern momentum. The Insider Tracking Desk updates the view with fresh insider-filing commentary following recent disclosures.
Four dimensions of comparison. Three favour Nvidia. The one tied dimension does not compensate. The Valuation Desk declares a winner.
Revenue surged 34% to $34.6 billion, free cash flow nearly tripled to $6.7 billion, and the stock just broke to all-time highs on institutional accumulation.