Why I Continue to Prioritize AMD Stock Over Nvidia

The article argues AMD’s upcoming MI450 chip could narrow the gap with Nvidia’s Vera Rubin, citing a 2nm process versus 3nm and 432 GB of HBM4 memory versus 288 GB. It also highlights Nvidia’s much larger $4.9 trillion market cap and lower P/E of 31 versus AMD’s 171, suggesting Nvidia has less room for tenfold upside while AMD may appeal more to growth investors. Overall, the piece is opinionated rather than event-driven and is unlikely to materially move the broader market.

Analysis

The market is likely to treat MI450 less as a standalone product event and more as a litmus test for whether AMD can compress the performance gap without conceding the software moat. If AMD proves credible on throughput, memory, and total cost of ownership, the next-order effect is not just share gain versus NVDA; it is a widening of the buyer universe for AI infrastructure, especially hyperscalers that want bargaining leverage against a single dominant vendor. That dynamic favors TSM as the real upstream toll collector because incremental competitive intensity usually translates into more wafer demand, more packaging complexity, and tighter advanced-node allocation.

The more interesting risk for NVDA is not an absolute slowdown in AI demand, but a valuation-duration problem. When a company becomes the benchmark compounder, the market starts discounting future dominance years earlier, which caps multiple expansion even if fundamentals remain excellent. In that setup, NVDA can still outperform on earnings while underperforming on stock price versus a smaller challenger with more asymmetrical perceived upside; that is the second-order effect the market often misses.

AMD’s setup is also fragile: the valuation already prices in a cleaner execution path than the fundamentals currently justify. If MI450 slips, is power-inefficient in practice, or software adoption lags, the stock can de-rate quickly because the premium leaves little room for disappointment. The relevant horizon is 1-2 quarters around launch validation and early customer wins; the longer-term thesis is 12-24 months if AMD can turn product parity into durable platform share.

Contrarian read: consensus may be underestimating how much NVDA can continue to monetize ecosystem lock-in even if relative hardware advantage narrows. The bigger dislocation may be in expectations, not market share; AMD can win headlines without yet winning enough wallets to sustain its multiple. If that happens, the better trade may be owning the supplier stack and fading the highest-expectation chip name rather than making a pure directional bet on AMD.

Key Decisions for Investors

• Long TSM on a 6-12 month horizon: the cleanest way to express continued AI capex intensity regardless of whether AMD or NVDA wins incremental share; risk/reward is better than taking binary product risk in either chip designer.
• Buy AMD on a staged basis only into launch confirmation or post-launch pullbacks over the next 1-2 quarters; upside is meaningful if MI450 validates, but size should be capped because the valuation already embeds success.
• Hedge or reduce NVDA exposure into any post-launch relative-strength gap in AMD: use NVDA as a quality compounder, but recognize that its multiple may remain range-bound even with strong earnings; better as a cash-flow hold than a momentum chase.
• Pair trade: long AMD / short NVDA only if MI450 benchmark and early OEM design-win data show a real performance-per-watt or TCO edge; otherwise the pair is too close to a beta trade and NVDA’s software moat likely dominates.
• For higher conviction, consider AMD call spreads 3-6 months out rather than outright stock: defined risk captures launch optionality while limiting damage if execution or sentiment disappoints.