Billionaire Stanley Druckenmiller Has Sold Micron, Broadcom, and Intel. Here's the Biggest AI Chip Designer Left in His Portfolio.

Stanley Druckenmiller opened an AMD position while exiting Broadcom, Intel and Micron, leaving AMD as his sole major AI-chip holding after previously selling Nvidia. AMD expects server-CPU revenue to rise 80% in 2H 2026 and 70% in 2027, while GPU growth above 100% could more than double total data-center revenue next year, supported by strong Helios demand and deals with OpenAI, Meta and Anthropic. The opportunity is tempered by valuation: AMD trades at 63x forward earnings, over twice Nvidia's and Broadcom's multiple, although analysts forecast earnings to double next year and rise another 50% afterward.

Analysis

The investable question is not whether AMD can grow rapidly, but whether it can convert component wins into system-level revenue without sacrificing gross margin. Rack-scale adoption can raise content per customer and reduce qualification friction, but it also shifts execution risk toward HBM availability, platform validation, networking interoperability, and on-time deployment—areas where NVDA's installed software ecosystem remains the principal competitive barrier. Announced hyperscaler relationships should be treated as pipeline indicators rather than revenue evidence until management discloses production volumes, backlog conversion, and data-center gross-margin trajectory.

AMD's valuation leaves limited tolerance for a merely solid launch. With the stock priced for a sharp earnings inflection, upside requires evidence over the next 1-3 quarters that CPU share gains and accelerator attach rates are occurring simultaneously; isolated CPU strength would likely be insufficient to sustain multiple expansion. Conversely, a modest reduction in next-year data-center growth expectations could create disproportionate downside because the market is underwriting a successful transition from chip vendor to integrated AI-system supplier.

The less obvious beneficiary of broader rack-scale deployments is AVGO: higher-density AI clusters increase switching, optical/connectivity, and custom-silicon content even if AMD takes accelerator share. The contrarian view is that AMD's opportunity may be underappreciated at customers seeking a second source, but the market may be overestimating how quickly second-source qualification becomes normalized production purchasing; AI infrastructure buyers prioritize utilization and software maturity over nominal hardware savings.

Key Decisions for Investors

• Keep AMD on a conditional long watch rather than chase spot: initiate only after the next earnings release shows raised 12-month data-center revenue guidance and confirms production deployment, not just customer commitments. Use a 3-6 month 5-10% out-of-the-money call spread to cap premium exposure; thesis is falsified by a data-center guide cut or gross-margin compression despite revenue growth.
• For a relative-value expression, consider long AMD / short INTC over 6-12 months in equal beta-adjusted dollars. AMD has a clearer path to monetizing heterogeneous CPU-plus-accelerator demand, while INTC faces foundry capital-intensity and execution constraints; exit if Intel demonstrates sustained server-share stabilization plus improving foundry losses, or if AMD fails to deliver sequential data-center acceleration.
• Maintain AVGO as the higher-quality ancillary AI exposure over a directional AMD chase. Add on evidence that hyperscaler cluster buildouts are expanding, as networking/custom-silicon content can capture rack-scale spend regardless of the winning GPU supplier; key risk is a hyperscaler capex pause or a material shift toward vertically integrated networking.
• Avoid using MU as a direct AMD read-through. Memory pricing and inventory cycles can dominate AI-demand benefits; revisit only if HBM supply tightness materially constrains accelerator shipments, which would turn MU into a bottleneck beneficiary but simultaneously weaken the AMD deployment thesis.