Why Advanced Micro Devices Stock Tanked on Tuesday

Reuters reported that Chinese AI developer DeepSeek plans to design its own AI chip for inference, which could reduce reliance on AMD and Nvidia. AMD shares fell nearly 7% on Tuesday on the risk that successful proprietary hardware would divert chip demand from third-party suppliers. The article notes export controls may limit sales, and chip development may fail, suggesting the selloff may be somewhat overblown.

Analysis

This is a classic narrative shock: the market is pricing a potential change in purchasing behavior as if it were already a revenue event. The real fundamental risk is not one Chinese lab building a chip, but the broader signal that large AI operators may increasingly internalize inference silicon to lower unit costs and reduce supplier dependence; if that takes hold, it compresses merchant accelerator TAM and weakens pricing power, but the effect is a 12-24 month story, not a next-quarter earnings hit.

The second-order dynamic is that a China-specific chip effort may actually be more of a domestic supply-chain bifurcation than a global disintermediation of AMD/NVDA. Export controls and software ecosystem friction make it hard for a local inference chip to become a universal platform, so the near-term spillover is likely sentiment damage to the AI hardware complex rather than immediate order cancellations. Relative to NVDA, AMD has less valuation support from the scarcity narrative, but the downside here is still mostly multiple compression, not a clear revenue cliff.

Contrarian view: the move looks overdone unless we get a second data point—tape-out confirmation, benchmark parity, or a customer list that extends beyond one developer. Falsifiers are straightforward: if AMD/NVDA data center demand, forward bookings, or guidance remain intact over the next 1-2 earnings cycles, this headline fades. If multiple large model builders announce custom inference silicon, then the thesis becomes structural and the whole AI semicap basket deserves a lower terminal multiple.

Key Decisions for Investors

• Tactically buy AMD only on continued post-news weakness and only if it reclaims the post-gap low; this is a 1-3 month mean-reversion trade, not a structural long. Risk/reward: upside to a mid-to-high single-digit rebound if the headline fades; stop if AMD makes a new relative low on volume.
• Express the broader view via a small long AMD / short SMH or SOXX pair for 4-8 weeks if sector sentiment stays fragile. The thesis is that the market is overpricing one rumor while broader AI demand remains intact; risk is a follow-on confirmation of custom-chip adoption.
• Do not short NVDA or AMD aggressively here without confirmation. If you want exposure to the downside, use limited-risk puts or put spreads into any implied-vol spike; the trade works only if the market starts treating inference insourcing as a repeatable industry pattern.
• Set a watch item for any Reuters follow-up on tape-out, foundry partner, or benchmark results. A credible commercial milestone would be the first real catalyst to extend the move from sentiment into fundamentals; absent that, expect the selloff to retrace within 2-6 weeks.
• If AMD/NVDA next earnings guide data center demand above consensus, treat this as a fade-the-headline opportunity and cover tactical shorts/hedges quickly. That would falsify the idea that merchant AI silicon is already being structurally displaced.