Samsung Foundry Competition Threatens GUC’s AI and CPU Dominance; Downgrade to Equal-weight

Core Conclusion

After a 135% rally, GUC’s valuation has stretched to 50x 2027 P/E, embedding assumptions that it keeps a near-monopoly on Tesla’s custom AI chips and Google’s CPUs. New evidence points to a structural erosion: Samsung’s foundry is set to capture a share of these marquee programs, diluting GUC’s wafer volume and medium-term earnings power. Our estimates stay unchanged, but upside has evaporated. We downgrade to Equal-weight with a target price of NT$4,888 and recommend taking profits pending fresh design wins.

What the Market May Be Underestimating

Consensus still prizes GUC as the sole beneficiary of next-gen AI and CPU outsourcing. In reality, two trends are undermining this thesis: hyperscalers actively diversify their foundry supply, and Samsung’s advanced nodes are approaching commercial viability. The packaging reconfiguration for Tesla’s AI5 and the likely placement of Google’s Axion 3 at Samsung’s SF2 process challenge the assumption that GUC’s revenue trajectory will steepen. The downgrade is not about current numbers breaking; it is about a ceiling that has become lower and less certain.

Evidence Chain

Tesla AI5 – partial displacement. Industry checks indicate that the AI5 packaging configuration could be altered from two TSMC-based AI5 chips to one AI5 plus a co-packaged backup AI4 built on Samsung’s 5nm. This instantly cuts the GUC-linked wafer content per vehicle. Looking ahead, next-gen 2nm AI6 is likely to be produced by Samsung and Intel foundries, not just TSMC. Although our model includes zero AI6 revenue, this undercuts the bull case that GUC would dominate Tesla’s future AI silicon. The stock’s 10% single-day drop reflected this reality check.

Google CPU – capacity diversification. Google’s Axion 3 CPU is expected to leverage Samsung’s SF2 2nm process, where yields are improving. The customer’s desire for dual-sourcing makes this almost inevitable. Our estimates already incorporate conservative volume assumptions for Google CPU and Tesla AI5, so no mechanical cuts are required. But the room for upside—which justified the multiple expansion—has contracted materially.

Key Risks

• Upside. TSMC could allocate more leading-edge wafers to GUC at preferential pricing to claw back share. A new large-scale project win from a major cloud service provider would reset the growth narrative and warrant a return to an Overweight stance.
• Downside. If Samsung’s 2nm yields mature faster than expected, or if Intel’s foundry gains traction, GUC’s existing book could suffer more than our base case implies. TSMC’s own capacity tightness and high wafer prices may continue to push hyperscalers toward alternative suppliers, accelerating share loss. The valuation at 50x 2027 P/E provides no buffer; any downward estimate revision would compress the multiple sharply.

Valuation or Trade Implication

Target price NT$4,888 stands, derived from unchanged earnings forecasts. The risk/reward has turned negative after the price surge. Downgrading to Equal-weight is a call to lock in gains. We would need concrete signs—either aggressive wafer support from TSMC or transformative new ASIC engagements—to upgrade again.