AMD vs Nvidia: The Valuation Winner Is Clearer Than Consensus Thinks

AMD has built a credible position as the second-place AI accelerator supplier. Its MI300 and MI350 platforms are achieving real commercial traction with hyperscaler customers seeking supply diversity from Nvidia. AMD's data centre revenue in FY2025 was approximately $14-15 billion, up roughly 95% year over year. That growth rate is genuinely impressive. Operating margin in the data centre segment is estimated at 20-25%, still well below Nvidia but improving.

Outside data centre, AMD has a legacy client PC business, an embedded business from the Xilinx acquisition, and a gaming GPU business. Each contributes $5-7 billion of revenue with variable margins. The diversification helps smooth revenue volatility but does not fundamentally change the AI-focused valuation thesis. Investors are buying AMD primarily for the data centre acceleration.

The stock trades at $280 per share against a $290 consensus price target. That is 3% upside to consensus. The multiple of 41x forward earnings is high for a company with 34% revenue growth and 17% operating margin. The premium to the semiconductor sector average of 23x is roughly 80%. The valuation requires AMD to continue compounding AI revenue at 60%-plus for several more years to support the current multiple.

Nvidia generated $215.9 billion of revenue in FY2026 (calendar 2025) with $130.4 billion of operating income. The company has become the most valuable chip company in history and the most valuable technology company globally. Data centre revenue represents approximately 88% of the total and grew at 65%-plus year over year. Operating margin of 65% is one of the highest sustained operating margins at this scale in modern corporate history.

At $200 per share, Nvidia trades at a 24.8x forward earnings multiple. The price target consensus is $269. That represents 34% upside. The multiple expansion from current levels has been modest over the past six months despite revenue growth acceleration, reflecting market concerns about the sustainability of the current AI capex cycle. Those concerns are legitimate but, in the Valuation Desk's view, overweighted at current prices.

Nvidia's moat is software more than hardware. CUDA has created a developer ecosystem that is structurally difficult to replicate. AMD's ROCm alternative is improving but has not yet reached parity in developer adoption, tooling maturity, or third-party library support. The software moat is what allows Nvidia to price at premium levels and earn the 65% operating margin. AMD cannot match that pricing power without building a comparable software franchise, which takes years.

Nvidia grew revenue at 65% year over year in FY2026. AMD grew at 34% in FY2025. Consensus expects Nvidia to grow 25-30% in FY2027 and AMD to grow 35-45%. The convergence argument relies on AMD's growth rate exceeding Nvidia's as Nvidia's base becomes too large to compound at the same rate. That convergence is mathematically inevitable at some point. The question is whether it arrives soon enough to justify AMD's premium multiple.

The Valuation Desk models both companies' revenue trajectories through FY2028. Under the current consensus, Nvidia's FY2028 revenue reaches $310-340 billion; AMD's reaches $75-90 billion. At those revenue levels, Nvidia is 3.6-4.5x the size of AMD. The gap narrows from the current 6.2x. That is convergence but slow convergence. AMD's multiple premium would need to close to match the historical narrow premium that second-place vendors typically earn.

Growth advantage: Nvidia on current trajectory, AMD on forward trajectory. Net: approximately neutral when forward-weighted. Winner on this dimension: draw.

Nvidia's 65% operating margin is the single most differentiated data point in this comparison. AMD's 17% operating margin is typical for a semiconductor fabless vendor. The gap reflects Nvidia's pricing power, software contribution, and mix of proprietary systems (DGX, Spectrum-X) in addition to chips alone. The margin gap is 48 percentage points. Closing even half of that gap over the next five years would require AMD to commercialise a software stack comparable to CUDA, which history suggests is a decade-long effort.

Margin advantage: Nvidia, unambiguously. The gap is structural rather than cyclical.

Nvidia generates $96.7 billion of free cash flow against a $4.9 trillion market cap, producing a 2.0% free cash flow yield. AMD generates $6.7 billion against a $454 billion market cap, producing a 1.5% free cash flow yield. Both are low yields by traditional value metrics; that is the nature of AI infrastructure investing. But Nvidia is the higher-yield option despite its premium market cap.

The deeper point is that Nvidia's free cash flow is still growing rapidly. FCF of $96.7 billion in FY2026 is 59% above FY2025 and 258% above FY2024. AMD's FCF of $6.7 billion is 179% above FY2024 but off a much smaller base. The absolute dollar growth in FCF favours Nvidia by a factor of roughly 10x. Compounding absolute cash flow at this magnitude creates optionality for capital return, M&A, and balance sheet flexibility that AMD cannot match.

FCF yield advantage: Nvidia.

The software moat favours Nvidia by a wide margin. CUDA adoption, developer community, and third-party tooling have been built over 17 years of focused investment. AMD's ROCm is improving but the catch-up work is substantial. By comparison, the narrower competitive dynamic in high-performance computing and AI model training has been Nvidia winning roughly 90%-plus of new deployments at the cutting edge. AMD's traction has been primarily at hyperscalers seeking supply diversity rather than technical differentiation.

The one area where AMD has genuine competitive advantage is in integration with CPU offerings for edge and inference workloads. The MI300A platform, which combines CPU and GPU on the same silicon, has meaningful advantages for specific workloads. That is real technical differentiation. But it is in a narrow segment of the broader AI opportunity.

Moat advantage: Nvidia, meaningfully.

Nvidia's top ten customers represent approximately 52% of data centre revenue. Concentration has increased rather than decreased as AI infrastructure spending has concentrated in a handful of hyperscalers. That is a structural risk. If any single hyperscaler materially slows its AI capex, Nvidia's revenue line feels it immediately.

AMD's customer concentration is lower in absolute terms but also in the data centre segment specifically. AMD's data centre revenue is spread across more customers but each customer's spend is smaller. The net concentration risk is approximately comparable on a revenue-weighted basis. This dimension approximately ties between the two companies.

The concentration risk is real for both. Investors should size positions accordingly. What separates the two on this dimension is that Nvidia's larger absolute revenue base absorbs customer-specific volatility better. A single customer cutting spend by 30% is a small fraction of Nvidia's total but could be meaningful to AMD's growth trajectory. Concentration dynamics: approximately tied with a slight edge to Nvidia for scale.

Across three prior technology cycles, the second-place semiconductor vendor has typically traded at a 40-60% discount to the incumbent leader. AMD vs Intel in 2012-2016 is the textbook precedent. Intel traded at 12-14x forward earnings; AMD traded at 8-10x forward earnings during that period despite AMD's higher growth rate. The second-place premium did not emerge.

The reason is that second-place vendors typically attract investor attention during the growth phase and lose it during the margin-compression phase. AMD's current 41x multiple is a growth-phase artefact. As the AI capex cycle matures into FY2027-28, the margin-compression phase is likely to arrive for AMD. At that point, the multiple will compress toward the second-place historical norm.

The implication: AMD today is priced for perfection on growth rate and margin expansion that is unlikely to materialise simultaneously. Nvidia today is priced at a more defensible level given its dominant position and established margin profile. History suggests the current multiple relationship is an inversion of the steady-state relationship. That inversion will correct.

Two specific developments would cause the Valuation Desk to revise this comparison in AMD's favour. First, material adoption of ROCm across a critical mass of AI developer workflows, combined with evidence of sustained pricing power for AMD in the AI accelerator market. Second, a regulatory constraint on Nvidia's market position that materially limits its ability to supply its full addressable market. Neither is currently probable on a twelve-month horizon.

The ROCm adoption thesis is the more credible of the two. AMD has been investing heavily in developer tools and third-party library support. The PyTorch 2.6 release included native ROCm support with performance within 15% of CUDA for most common workloads. That is progress. The question is whether the remaining gap closes in a timeframe that matters for the current multiple. Our estimate: probably not within the next two years, but possibly within four to five.

The regulatory constraint thesis is possible but politically complex. The most likely near-term constraint is export control extension that affects Nvidia's China business more than AMD's. That would create some relative advantage for AMD but would not fundamentally change the comparison in the US and European markets. Both companies face similar export restrictions on their most advanced products.

For investors building semiconductor exposure at current prices, the Valuation Desk recommends a position weighting that reflects the comparison outcome. A three-to-one preference for Nvidia over AMD on incremental capital deployment captures the fundamental differences. Investors already holding AMD at higher weights should consider trimming toward parity with Nvidia rather than accumulating further AMD exposure.

The comparison is not a statement that AMD is a bad business. It is a statement that AMD is priced at a premium to its fundamentals while Nvidia is priced at a discount to its fundamentals. The relative mispricing creates the comparison's clear answer. Both companies could deliver positive absolute returns from these levels; Nvidia is more likely to outperform AMD on a risk-adjusted basis over the coming eighteen to twenty-four months.

We remain constructive on the AI infrastructure theme overall. Within the theme, Nvidia is the cleaner expression of the thesis.