Alchip Technologies Ltd: Strong Revenue Ramp from 3nm Project Starting in June

Core Conclusion

Alchip’s 3nm AI chip project (Trainium4) is set to begin its revenue ramp in late May or early June 2026, driving a pronounced back-end loaded 2026 revenue profile. Management confirmed the company remains the sole physical design partner for both 3nm and 2nm generations of this chip, creating a multi-year revenue visibility that the market underappreciates. While 2026 upside is constrained by wafer capacity, 2027 volume expansion offers material upside. Combined with a 1Q26 gross margin of 50.2% that beat both Morgan Stanley’s 45.0% and consensus’s 39.8% estimates—driven by a higher NRE mix—the fundamentals support the Overweight rating and NT$5,088 price target.

What the Market May Be Underpricing

The market appears to assign insufficient probability to two key elements:

1. The exclusivity of Alchip’s role in Trainium4 → The company is the only physical design partner for both 3nm and 2nm nodes of the same AI chip. This entrenchment reduces competitive risk and extends the revenue runway well beyond the initial 3nm ramp.

2. 2027 volume upside → Management explicitly stated that 2026 wafer capacity limits near-term growth, but volume increases are expected in 2027. The market may be pricing in only the current year’s constraints without fully discounting the step-function revenue jump once capacity opens.

3. Gross margin resilience → 1Q26’s 50.2% gross margin is not a one-off; it reflects a structurally higher NRE mix that management indicated remains very strong, particularly from North American AI customers. The market may view the beat as transient rather than indicative of a sustained margin profile.

Evidence Chain

1. 3nm AI chip ramp timeline → Management stated the chip will begin ramping in late May or early June 2026, with stronger momentum in 3Q26. 2026 revenue will be back-end loaded, with auto customers dominating 1H26 and the AI chip driving 2H26.

2. Exclusive design partner status → During the earnings call, management highlighted that Alchip is the only physical design partner for both the 3nm and 2nm versions of Trainium4. This is consistent with the company’s strategy of not exiting custom CPU or COT businesses as long as it can add value beyond pure production.

3. 1Q26 earnings outperformance → Revenue of NT$4,192 million was in line with MS estimates but gross margin of 50.2% exceeded both MS (45.0%) and consensus (39.8%). Operating margin of 32.7% also beat MS’s 22.6%. The NRE pipeline remains very strong, especially from North America AI markets.

4. Revenue composition → Auto customers are the largest contributor in 1H26, while the 3nm AI chip will drive the 2H26 surge. This supports the back-end loaded view and implies the current quarter’s lower revenue is not indicative of full-year weakness.

Key Divergences and Risks

• Chip demand trajectory → If AI chip demand grows slower than expected, the 3nm ramp could be delayed or volumes reduced. This would compress 2H26 upside.

• Wafer capacity constraints → Management explicitly noted 2026 upside is limited by wafer capacity. Any extension of this bottleneck into 2027 would cap the volume expansion that we see as the next major catalyst.

• China HPC localization slowdown → A deceleration in China’s high-performance computing localization could reduce NRE revenue, a key driver of the margin beat.

• Pricing competition → Increased competition in the design service space could pressure gross margins, particularly if Alchip’s value-add is commoditized.

Valuation or Trading Implications

Using a residual income model with a 10.4% cost of equity (2% risk-free rate, 6% equity risk premium, 1.4 beta), 16% intermediate growth, and 5% terminal growth, the base-case fair value is NT$5,088. At the May 8 closing price of NT$4,890, the implied upside is 4%. The key catalysts for re-rating are:

• Winning a new project from a major CSP other than AWS (e.g., Meta).
• Clear guidance on 2027 Trainium4 volumes once wafer capacity constraints ease.

Maintain Overweight. The risk/reward is asymmetric, with the downside largely priced in after the stock’s correction from its 52-week high.

Appendix Data Summary