CLSA raises AMD stock price target to $710 on MI-455 GPU outlook

CLSA raised AMD's price target to $710 from $575 and maintained an Outperform rating, driven by expectations for higher-volume, higher-priced MI-455 GPU sales. The firm lifted fiscal 2027-2028 EPS forecasts by 25%-29% and now expects AMD to reach its $20 non-GAAP EPS target in 2027, forecasting $20.25—about 30% above consensus. CLSA estimates MI-455 racks could carry a $5 million ASP, with AMD recognizing roughly 70% as revenue and generating gross margins in the high-40% to low-50% range; however, independent GPU performance testing remains unavailable.

Analysis

The upside case is less about unit volume than whether AMD can monetize a rack-level system rather than a standalone accelerator. If system revenue scales with gross margin near 50%, operating leverage can be substantial because incremental AI revenue absorbs a largely fixed software, sales, and R&D base; however, this also makes the thesis unusually sensitive to customer acceptance of the full platform, networking interoperability, and production yields. The absence of independent performance and availability validation means the market is currently capitalizing an estimate stack rather than a de-risked product cycle.

AMD’s likely second-order beneficiaries are AI infrastructure suppliers with content per deployed rack: MU for high-bandwidth-memory exposure, AVGO/MRVL for connectivity, and ANET/COHR for scale-out networking and optical interconnect. COHR has the cleaner near-term read-through if hyperscaler cluster build-outs remain capacity constrained, but its benefit depends on optical content rising faster than transceiver price erosion. INTC is not a direct beneficiary: a credible alternative GPU platform raises the competitive intensity for foundry capacity, advanced packaging, and enterprise server wallet share rather than improving Intel’s AI positioning.

The valuation framework embeds both a large earnings ramp and sustained premium multiple through 2027. Over the next 1-3 months, the key catalyst is customer qualification evidence—named design wins, production timing, and supply commitments—not further target-price revisions. Over 6-18 months, the critical falsifier is any indication that AMD must discount hardware or fund software/customer engineering to displace incumbent platforms; that would impair the assumed margin mix before reported revenue misses become visible.

Contrarian view: the market may be underestimating the strategic value of a second qualified AI compute supplier to hyperscalers, which can support adoption even if headline benchmark leadership is mixed. But it may also be overestimating how quickly qualification converts into recognized revenue; hyperscaler procurement can reserve supply well before broad deployment, creating a booking-to-revenue gap and a vulnerable period for an elevated multiple.

Key Decisions for Investors

• Do not chase AMD on the analyst revision alone. Establish a 1-3 month watch item for a long only after independent MI-455 benchmarks and at least one production-scale customer deployment are disclosed; require evidence that AI gross margin is holding above the high-40% range. A failure to provide this evidence by the next major product/customer update is thesis-negative.
• Express the infrastructure spillover with a 3-6 month long basket of MU, ANET, and COHR rather than incremental AMD beta. Favor ANET/MU over COHR on risk-adjusted basis: their demand exposure is broader, while COHR carries greater execution and pricing risk. Exit or reduce if hyperscaler capex guidance weakens or optical lead-time commentary normalizes.
• For AMD holders, use a valuation hedge rather than an outright short: long AMD / short SMH or SOXX in a modest ratio over 3-6 months. This isolates company-specific share gains while protecting against AI-capex, rates, or semiconductor-multiple compression; unwind if AMD’s AI revenue guidance fails to outgrow the semiconductor index’s implied earnings trajectory.
• Avoid treating INTC as a sympathy long. Revisit only if Intel demonstrates external foundry wins or material AI-server share gains; a stronger AMD platform is more likely to pressure Intel’s enterprise CPU economics than create investable sector-wide upside.