NVIDIA’s stock is up just 10% this year. AMD has surged 142%. Micron has gained 213%. The Philadelphia Semiconductor Index is up 71%. And NVIDIA, the company that defined the AI era, is trading at its lowest valuation in nearly five years.
At roughly $212 per share, the market is pricing NVIDIA as if it will have near-zero growth after 2027, according to Morningstar analyst Brian Colello. He sees fair value closer to $280 — about 16x projected FY2029 sales. NVIDIA’s current EV/EBITDA of 17x forward earnings is less than half its five-year average of 36x.
The bear case is straightforward: competitors are emerging from every direction. SambaNova, Cerebras, and startups are building custom chips. Google now sells access to its in-house TPU. Amazon, Meta, Microsoft, and OpenAI are all developing their own silicon. AMD’s Helios server rack system, launching this year, directly competes with NVIDIA’s Grace Blackwell and Vera Rubin lines. Revenue scale itself creates a growth ceiling — a $560 billion company can’t grow at startup rates.
The bull case is equally coherent. Analysts expect NVIDIA’s revenue to grow 42% to $560 billion next fiscal year, and another 23% the year after. By contrast, AMD — trading at 53x forward earnings — is expected to generate only $78 billion in revenue by 2027. The valuation gap between the two companies defies arithmetic.
More subtly, NVIDIA’s moat in inference chips (the market for running AI models, not training them) has actually strengthened according to recent reports. And if a future AI investment correction hits, it may hurt NVIDIA less than its challengers — many of whom would abandon their own chip efforts and return to the CUDA ecosystem rather than ride out the downturn.
As Colello puts it: “NVIDIA looks quite cheap on a multi-year view.” The question is whether hyperscaler capex holds up and whether NVIDIA can defend its market share. His answer to both is yes.
Translated and adapted from Sohu IT (it.sohu.com).