Prediction: 2 AI Semiconductor Stocks That Will Join Micron and Broadcom in the $1 Trillion Club by 2028

AMD and ASML are presented as potential $1 trillion companies by 2028, supported by sustained AI-driven semiconductor demand. AMD's data-center revenue rose 107% year over year last quarter, with AI data-center revenue expected to grow well above 100% in the current quarter; consensus EPS is projected to increase from $2.65 last year to $15.61 next year. ASML plans to expand manufacturing capacity by 30% next year and potentially another 30% in 2028, a cumulative 69% increase, which could support roughly €74 billion of 2028 revenue under the article's assumptions.

Analysis

AMD’s rack-level strategy matters less as a near-term revenue event than as an attempt to remove the integration bottleneck that has protected Nvidia’s systems franchise. If Helios deployments are genuinely repeatable rather than bespoke customer engagements, AMD can raise GPU attach rates for EPYC CPUs and networking, improving blended gross margin and reducing the discount required to win accelerator share. The principal second-order loser is merchant server integration; the relevant proof point over the next 1-3 months is disclosed rack shipments, software qualification progress, and gross-margin trajectory—not customer MOUs.

The earnings assumptions embedded in AMD’s valuation leave little room for supply-chain or execution slippage. Hyperscaler AI budgets are increasingly constrained by power availability, networking, and utilization economics rather than accelerator demand alone; delayed data-center energization could push orders out without changing long-run demand. A failure to show sustained data-center mix improvement at the next two earnings prints, or a material price response from NVDA, would likely drive sharper multiple compression than the underlying revenue miss.

ASML is the cleaner structural exposure because incremental leading-edge capacity requires its installed base, while its service revenue and long equipment life cushion a semiconductor downcycle. However, capacity plans are not equivalent to recognized revenue: customer prepayments, export-license outcomes, fab construction timing, and foundry utilization determine conversion. The 6-18 month upside is strongest if memory makers sustain leading-edge DRAM/HBM investment alongside logic demand; the near-term risk is that a broad WFE digestion cycle delays tool acceptance and revenue recognition.

Consensus appears too focused on trillion-dollar arithmetic and too little on value capture. Rack-scale systems could expand AMD’s content per deployment but may also shift more working-capital, warranty, and support burden onto AMD, limiting the expected margin benefit. Conversely, ASML’s scarcity value is likely underappreciated if advanced packaging and HBM bottlenecks ultimately force more leading-edge wafer starts rather than simply reallocating existing capacity.

Key Decisions for Investors

• Prefer long ASML over AMD on a 6-18 month horizon: initiate on WFE-sector pullbacks, targeting relative outperformance through 2027 capacity conversion. Thesis is falsified by material order-book deterioration, customer acceptance delays, or export restrictions that impair the advanced-tool mix; use a 12-15% relative stop versus SOXX.
• Run a tactical AMD/NVDA relative-value short (long NVDA, short AMD in equal beta-adjusted dollars) into the first Helios shipment and subsequent earnings update. Expected payoff is 10-15% relative if AMD cannot demonstrate rack-scale gross-margin accretion; cover if AMD reports meaningful production deployments plus data-center gross-margin expansion while NVDA guides a material systems slowdown.
• Do not chase AMD outright before independently verifiable shipment data. Set an alert for disclosed Helios backlog, GPU availability, and customer deployment scale; only upgrade to long exposure if management shows repeatable revenue conversion and no incremental working-capital drag across two reporting periods.
• For a bullish semiconductor-capex expression with defined risk, consider ASML 6-9 month call spreads financed by selling a higher strike, sized modestly. The trade requires confirmation that backlog conversion and 2027 capacity additions remain on schedule; exit after any export-control revision or guidance that pushes customer tool acceptance beyond the planned ramp.