Nvidia remains the defining AI infrastructure stock, but two other semiconductor companies are generating stronger returns in 2026: Micron Technology is up 59% year-to-date, and TSMC reported Q1 revenue of $35.9 billion — a 41% year-over-year increase. Nvidia, by comparison, is up roughly 8% over the same period despite expectations for 77% revenue growth in its upcoming quarter.
The divergence is worth examining. It reflects where investors see the best combination of near-term earnings delivery and forward valuation — and it suggests that the AI trade in 2026 is more nuanced than simply buying the company that makes the chips that run large language models.
Why Micron Is Outperforming in 2026
Micron’s 59% year-to-date gain is tied primarily to one theme: AI memory demand. The high-bandwidth memory (HBM) chips that power Nvidia’s H100 and upcoming Blackwell GPUs require enormous volumes of fast DRAM, and Micron is one of the two companies (alongside SK Hynix) that can manufacture them at scale.
Key factors behind Micron’s run:
- HBM3E supply constraints: Demand for HBM3E — the memory standard required for frontier AI training — exceeded industry supply throughout late 2025 and into early 2026. Micron, which ramped HBM production ahead of schedule, has been able to charge premium prices.
- Data center DRAM pricing recovery: The broader DRAM pricing cycle turned upward in mid-2025 after a prolonged oversupply period. Micron benefits across its entire product line, not just HBM.
- NAND pricing stabilization: Micron’s storage products (NAND) also saw pricing stabilize, contributing to margin improvement across the company.
The straightforward version: every H100 cluster needs Micron’s memory. As AI training deployments scale, so does Micron’s revenue.
TSMC’s Q1 2026 Numbers
TSMC reported Q1 2026 results that showed the scale of the AI capex cycle in manufacturing terms. Revenue of $35.9 billion was up 41% year-over-year, driven by:
- Continued N3 (3nm) ramp: Apple, Nvidia, AMD, and other major customers have been increasing orders for chips manufactured on TSMC’s 3nm process nodes.
- CoWoS packaging demand: The advanced packaging technology that Nvidia uses to combine multiple chips into a single package (used in H100 and Blackwell) is manufactured by TSMC. Demand for CoWoS capacity has been running ahead of TSMC’s ability to build it.
- N2 (2nm) preparation: TSMC is preparing its N2 process for volume production later in 2026, and early customer design wins for N2 are already contributing to revenue through engineering development fees.
TSMC’s gross margin has also been improving as the mix of advanced nodes (which carry higher prices) grows as a share of total revenue.
Where Nvidia Stands
Nvidia is not underperforming in any absolute sense — an 8% YTD gain is solidly positive, and the company is widely expected to report approximately 77% revenue growth in Q1 FY2027 (the quarter ending in April 2026). For a company of Nvidia’s scale, that growth rate is extraordinary.
The relative underperformance compared to Micron and TSMC reflects valuation math. Entering 2026, Nvidia was trading at a significantly higher earnings multiple than either Micron or TSMC. When earnings growth is priced in at a high multiple, even strong results don’t necessarily drive large additional stock gains — the market has to keep revising upward to generate momentum.
Both Micron and TSMC were trading at more modest valuations entering 2026, which gave them more room to generate returns as the AI revenue came through in actual results.
The Contrast: Palantir Down 18%
Not all AI-adjacent companies are participating in the 2026 rally. Palantir is down approximately 18% year-to-date, despite reporting continued government and commercial AI contract wins.
Palantir’s situation illustrates the valuation problem in reverse. The company entered 2026 at an extremely high price-to-sales ratio, and any slowdown in growth narrative — even a modest one — is disproportionately punished when the multiple is that high. Palantir’s actual business continues to grow, but investors buying at 2025’s elevated prices have seen losses.
What This Means for Investors
The 2026 AI trade has several layers:
| Company | YTD Performance | Why |
|---|---|---|
| Micron | +59% | HBM supply constraints, DRAM pricing recovery |
| TSMC | Revenue +41% YoY, stock solid | N3 ramp, CoWoS demand |
| Nvidia | +8% | Strong absolute growth, but high entry valuation |
| Palantir | -18% | High valuation entering 2026, narrative slowed |
The pattern suggests that AI-related outperformance in 2026 has been concentrated in the companies with the most direct and least-priced exposure to the physical infrastructure buildout: memory chips and chip manufacturing capacity. These are businesses where demand is largely non-discretionary (you cannot run a frontier AI model without sufficient memory and manufacturing capacity), supply is genuinely constrained, and pricing power is real.
Whether that pattern holds through the rest of 2026 depends primarily on whether AI infrastructure investment continues to accelerate. Amazon’s $330 billion Anthropic commitment, announced this week, is one data point suggesting the answer is yes — at least at the hyperscaler level.
Related: Amazon Commits Up to $330 Billion to Anthropic
Related: Tesla Q1 2026 Earnings: Gross Margin Hits 2-Year High
Sources: Motley Fool, Bloomberg, TSMC Investor Relations, Micron Technology