3 AI Chipmakers With Far More Upside Than AMD

The article argues AMD is losing momentum in AI chips versus Nvidia, Broadcom, and Amazon. AMD’s data center revenue grew 39% YoY in Q4, below Nvidia’s 75% and Broadcom’s 106%, while Amazon’s chip business is said to be running at a $20 billion annualized rate and growing triple digits. The core message is bearish on AMD’s relative upside because its premium valuation already discounts growth that may be difficult to deliver.

Analysis

The market is implicitly pricing AI silicon as a winner-take-most GPU race, but the bigger second-order shift is moving from generalized accelerators to workload-specific silicon. That structurally favors vendors with tighter hyperscaler co-design relationships and broader system-level control, because they capture more of the value chain and can lock in roadmap dependency before customers standardize around a single platform. The implication is that the competitive moat is migrating away from raw FLOPS and toward software integration, packaging, networking, and supply allocation.

AMD is the clearest valuation mismatch in this setup: it is still being awarded optionality on share gains even though the burden of proof has shifted. If hyperscalers increasingly diversify across custom ASICs and in-house designs, AMD’s upside becomes more cyclical and execution-dependent, while downside is amplified by a premium multiple that assumes faster data-center share gains than the operating data supports. That makes AMD vulnerable not to one rival, but to a fragmented competitive field that can cap its TAM expansion and compress its negotiating leverage.

The beneficiaries are more interesting than a simple Nvidia-versus-everyone framing. Broadcom likely captures the highest-quality economics because custom silicon tends to deepen customer stickiness and expand attach rates into networking and design services, while Amazon monetizes AI chips indirectly by improving cloud margin and retention rather than through chip ASPs alone. A less appreciated risk is that accelerated in-house chip adoption by hyperscalers may pressure merchant GPU pricing over the next 12-24 months, but the larger longer-term effect is that smaller AI infrastructure vendors could be squeezed as capex concentrates around the few firms able to fund custom roadmaps.

Key Decisions for Investors

• Avoid initiating new long AMD exposure for the next 3-6 months; if already long, reduce on strength and treat any post-earnings pop as an opportunity to trim into a premium multiple with weakening relative growth.
• Long AVGO vs short AMD as a 6-12 month pair trade: Broadcom has the cleaner monetization path from AI capex secularization, while AMD is more exposed to narrative compression if hyperscaler custom silicon continues to take share.
• Add NVDA on pullbacks rather than momentum-chasing: the risk/reward is still favorable over 12 months because the near-term downside is mostly multiple-related, while the upside remains tied to ecosystem control and supply chain scarcity.
• Use AMZN as a hidden AI capex beneficiary over 12-18 months: long AMZN into weakness if the market overfocuses on retail, since chip adoption should support AWS margin durability and customer lock-in.
• For tactical expression, buy 6-9 month AMD put spreads funded by selling out-of-the-money calls; this captures downside from multiple compression while limiting carry if the stock continues to drift on AI enthusiasm.