Rosenblatt names top semiconductor picks amid data-center strength

Rosenblatt Securities named AMD its top semiconductor pick heading into earnings season, citing strengthening data-center demand despite an approximately 25% average share-price decline over the past month among data-center-exposed semiconductor companies. AMD CEO Lisa Su raised the projected 2030 server CPU market opportunity to $220 billion from $120 billion, while TSMC’s 2nm capacity is expected to double in 2027 and support AMD’s Venice CPU ramp. Rosenblatt sees supply-chain capacity—not weak demand—as the main longer-term growth constraint, describing current constraints as addressable tactical challenges.

Analysis

The more useful signal is a possible change in AI infrastructure mix, not a blanket “semis are cheap” call: agentic workloads could add CPU-intensive orchestration alongside accelerator demand. If that workload shift appears in customer deployments, AMD may gain incremental server-CPU share without requiring a broad reversal in AI spending. But the cited demand evidence is indirect; Rosenblatt’s expanded TAM is a management estimate, not proof of realized orders.

Near term, earnings and guidance are the test. A sector drawdown can create room for a relief move, but capacity constraints may defer shipments and revenue rather than eliminate demand. Over 1–3 months, watch AMD server-CPU commentary, customer adoption, and supplier lead times. Over 6–18 months, execution and foundry/back-end capacity determine whether the opportunity converts into volume. TSMC could benefit from higher utilization and advanced-node demand, while constrained equipment supply could limit the pace of AMD’s ramp; Intel is a competitive risk if it improves server execution. Memory suppliers such as Micron may gain from higher system memory content, but not necessarily in proportion to CPU demand.

Contrarian risk: the market may be discounting a real bottleneck and treating a long-dated TAM as near-term earnings. The bullish thesis weakens if AMD’s server-CPU guidance or shipment outlook fails to improve, or if customer commentary indicates agent workloads are not materially increasing CPU demand. The article supplies no valuation, positioning, or estimate-revision data, so avoid assuming the 25% sector decline makes AMD inexpensive.

Key Decisions for Investors

• Prefer a staged AMD long over an immediate full-size entry: establish only a starter ahead of earnings, then add if server-CPU guidance and shipment commentary confirm demand rather than relying on the TAM claim. Keep sizing modest given execution and supply-chain uncertainty.
• Consider a relative-value position—long AMD versus a diversified semiconductor basket such as SOXX—only if AMD’s earnings revisions hold up better than the group. This expresses company-specific execution without betting that all semiconductor multiples rebound.
• Track AMD server-CPU revenue/guidance, customer adoption signals, and foundry/back-end lead times over the next 1–3 months. Falsify the thesis on a material downward revision to AMD’s server outlook or evidence that supply delays are pushing deployments out rather than merely shifting timing.
• Do not chase the broader supplier read-through yet: TSMC and Micron may benefit if CPU/server builds translate into wafer and memory demand, but confirm capacity allocation and order conversion before adding exposure. Treat the 2027 capacity outlook as a watch item, not a near-term earnings catalyst.