Temple 8 Research
AI Buildout, Macro & Energy Infrastructure Research
UMC keeps delivering. Per a report shared by @dnystedt, its jointly developed 12nm FinFET process with Intel is set for mass production at Intel Arizona in 2027, and UMC sees a stronger second half of 2026 from price hikes on some processes and new orders, with broad price hikes likely in 2027 as materials costs rise. x.comÂ
UMC is preferred to GlobalFoundries: its 12nm partnership with Intel targets US production in 2027, offering a way to win customers without building its own US fab. Longer term, logic folding could improve established-node chips without smaller transistors, though UMC has announced no such program and revenue could take years. Shares have recovered, making the entry less attractive; royalty terms are unclear. x.comÂ
UMC has a meaningful re-rating path into 2H26-2028, with the target raised to NT$215 at 25x 2027 EPS. Worries about no 2H26 price hike miss that this year's hikes are selective, with broad 10-30% increases coming in 1Q27. Utilization should reach about 93% for 12-inch and 90% for 8-inch by end-2026, with full 8-inch loading by 2H27. From 2028, an asset-light 12nm FinFET model, likely driven by Apple display drivers, and optical solutions add growth, with TFLN EPS of NT$0.7 and NT$1.6 in 2027-28. … Show more
UMC's planned price increases should lift it, and its small size next to TSMC could drive the stock higher faster. Taiwan's second-largest foundry plans selective hikes in late 2026 and broader, possibly larger ones in 2027, mostly on new orders and capacity, amid rumored TSMC 3nm hikes of up to 15% on AI and ASIC demand. Raw materials and pricier Singapore expansion are raising costs; its new Singapore fab targets mass production in 2027, while interposer capacity doubles to 6K wafers a month. … Show more
UMC is a primary beneficiary as TSMC aggressively phases out mature nodes. Per a supply-chain report, TSMC's 28nm wafer starts fell from 200,000 a month early this year to 150,000 in June as it shifts capex to 2nm, A14 and advanced packaging. 28nm remains crucial for display drivers, power management, microcontrollers, networking and automotive chips, so customers will need second sources, and UMC's complete 28nm platform and advancing 22nm make it the main alternative. x.comÂ
Tower Semiconductor and UMC stand to benefit as 8-inch legacy foundry capacity tightens, another Macronix-style squeeze. Per @jukan05, UMC is raising 8-inch wafer prices 10-15%, with its Taipei and Xiamen fabs running at about 105-108% utilization, while 12-inch hikes are smaller at 5-10%. The same dynamic is playing out in E-glass. The bullish view on both foundries sits alongside a non-consensus bet on GlobalFoundries' photonics business, which won AMD's PIC order. x.com x.comÂ
UMC has gained the leverage to raise chip prices as China/non-China supply-chain decoupling peaks, Nomura says. Customers moving volume to Chinese foundries have largely done so, cutting attrition risk, while TSMC's focus on leading-edge nodes deprioritizes mature ones. UMC has told customers it will adjust pricing in the second half on rising input costs and tight utilization, estimated at about 70% for 8-inch and 80% for 12-inch fabs in Q1, both seen above 80% from Q2. x.comÂ
UMC's first high-volume silicon photonics shipment is the bottleneck-resolving traction expected from accessible foundries. TSMC's capacity discipline opens room for foundries advancing in specialty and optical technology, which is why UMC and GlobalFoundries are favored over TSMC. SILITH and UMC shipped the first mass-produced photonic IC wafers from UMC's Singapore 12-inch fab for a 1.6T platform, qualified by a leading cloud customer. UMC targets an open 12-inch platform in 2027. x.comÂ
Intel's foundry may be closer than the market assumes because its UMC 12nm partnership brings external-customer experience and a more customer-friendly design kit. Musk said Terafab plans to use Intel 14A, with Tesla and SpaceX building and running the fabs themselves; that would be an important outside test. Intel still must prove yields, volume and schedules, while Terafab's commercial terms remain unclear. Still bullish, but not enough to add now. x.comÂ
Intel's poor EMIB-T packaging yields threaten its plans for next year. Per a confidential source, yields are very poor, and EMIB-T may now slip into the following year. The 14A PDK 0.9 matters more for the long term. Intel remains a very strong long-term bull case because yields should improve, though setbacks can arise along the way; whether to avoid this near-term risk is a personal judgment. x.com x.com x.com x.com x.com x.comÂ
United Microelectronics looks like a clear winner as TSMC shifts its 7nm, 6nm and 5nm capacity downward, letting lagging-edge foundries capture growing share in microcontrollers, power and sensing chips, markets growing even faster than advanced logic. x.comÂ
