TD Cowen reiterates AMD stock buy rating on AI workload strength

TD Cowen reiterated a Buy rating on AMD with a $675 price target, citing agentic AI and production inference workloads supporting CPU demand; AMD is currently quoted at $546.68, down 11% over the past week but up 147% year to date. AMD previously expanded its CPU total addressable market estimate to $220 billion and guided to 70% year-over-year server CPU growth in 2027; InvestingPro data cited 40% trailing-12-month revenue growth and analysts’ 47% fiscal 2026 growth forecast. Valuation is a counterpoint: the stock trades at a 139 P/E and InvestingPro’s Fair Value analysis suggests it may be overvalued. Other analysts cited in the article set AMD targets of $700–$800, and CEO Lisa Su confirmed plans to boost chip supply in 2027 for rising AI demand.

Analysis

The key underwriting question is not whether agentic workloads use more CPU coordination, but how much of that incremental workload AMD can convert into share, pricing, and earnings soon enough to support a premium valuation. More orchestration can lift server-CPU demand, yet it does not guarantee AMD wins the socket: buyers may prioritize availability, platform qualification, and total-system cost, leaving Intel room to capture incremental volume. Nor does higher CPU activity necessarily increase CPU content per AI system proportionally; software optimization and workload consolidation could limit the revenue conversion.

The 2027 supply response creates a timing mismatch: investors may capitalize expected capacity and demand before shipments and margins are verifiable. A delay in customer qualification or a slower deployment curve would expose AMD to estimate compression, while successful execution could extend the CPU cycle beyond GPU-led expectations. Broadcom’s stated positioning in high-speed DSPs is a separate 1.6T networking exposure, not evidence of AMD’s CPU economics; it offers a less direct way to participate in AI infrastructure buildout.

Near term, the sharp pullback may invite dip-buying, but elevated expectations make stabilization and estimate follow-through more important than analyst targets. Over 1–3 months, watch server CPU guidance, cloud/customer deployment evidence, and relative estimate revisions for AMD versus Intel. Over 6–18 months, verify whether expanded supply translates into realized shipments and attractive economics. The contrarian risk to the bullish narrative is that CPU demand grows while AMD’s share or revenue per workload fails to; that would make the growth story real but the equity payoff disappointing.

Key Decisions for Investors

• Avoid chasing AMD on the workload narrative alone. Consider a staged long only after price action stabilizes and subsequent guidance or customer evidence confirms server CPU demand; use the next earnings updates as the near-term catalyst check.
• Watch AMD versus Intel server-CPU estimate revisions and evidence of customer wins. If AMD’s growth guidance holds but Intel captures the incremental deployments, reassess the AMD share-capture thesis rather than treating sector demand as an AMD-specific catalyst.
• Treat Broadcom as a separate AI-networking watchlist exposure, not an AMD hedge. Any relative position should wait for confirmation of 1.6T design wins and should account for the fact that the article provides no valuation or relative-risk data.
• Falsify the bullish AMD thesis if server CPU guidance is cut, shipment timing slips, or reported growth fails to convert into earnings; the 2027 supply ramp is not itself proof of demand or profitable utilization.