Finn Stockinger
Equity Analyst & Investor, Technology Research
Reported talks between SK hynix and Intel to make memory chips in the US are another strong data point for Intel Foundry, though nothing is signed. After multiple leaks of SK hynix looking at EMIB, big players are clearly treating Intel as a real alternative for advanced packaging. A deal would move EMIB from a backup option to core AI infrastructure, offloading packaging bottlenecks to US soil while undercutting TSMC on cost. Volume is not expected before 2027 or 2028. x.comÂ
Intel has a supply deficit, not a demand problem, and is shifting capacity to high-margin data center CPUs. It is undershipping demand by over $1 billion a quarter while Data Center & AI operating margin hit 39.5%. Q2 revenue was $16.1 billion, up 25%, with 41.8% gross margin helped by about a 50% cut in 18A wafer costs. Custom ASIC run-rate reached about $2 billion. The test is whether 18A yields keep pace as 2027 capex for 14A fabs adds depreciation that could pull margins back down. x.com x.com x.com x.com x.com 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
Intel's EMIB packaging is ramping as advanced packaging becomes the next chokepoint after HBM. EMIB substrate yields reportedly rose from ~30% in 2Q26 to ~45%, with targets of 50% by 4Q26 and 60% by 1Q27. Google is expected to adopt EMIB-T in 2027 and AWS is testing it. CEO Lip-Bu Tan says Intel has to prepay substrate suppliers to secure capacity. CoWoS capacity is set to grow about 20x from 2023 to 370k wpm by 2028, more than TSMC can carry alone. x.comÂ
SK hynix reportedly considering Intel Foundry for HBM4E base dies is a negotiating tactic against TSMC, not a real win. Companies, Nvidia included, keep 'considering' Intel and then go quiet, because they know Intel still falls short. x.comÂ
Advanced packaging is where the new Moore's law lies, and Intel's EMIB is becoming a major business. Per Digitimes, cited by @pequityresearch, the first TPU using EMIB has met yield standards, making it significantly more likely that future TPU generations keep using it. A quoted assessment calls EMIB a multi-billion-dollar-per-customer business with 30% operating margins and phenomenal ROIC, now headed by SK hynix's former CEO. x.com x.comÂ
US sovereign AI and defense demand for domestically made compute will bring Intel most, if not all, US semiconductor designers as customers. PC weakness also lets Intel shift wafers to higher-margin server CPUs, but that takes a few quarters; capacity to ship more server CPUs is the limiter. The $240 billion 2030 server CPU market forecast is a stretch case, not the base case, versus roughly $200 billion consensus. x.com x.com x.com x.com x.comÂ
Intel's advanced packaging story is already priced in; the stock got there a little too fast. EMIB and EMIB-T are legit, and the foundry pivot gives Intel a real narrative, but the market has moved ahead of it. x.comÂ
Intel's EMIB-T packaging could reaccelerate now that the root causes of substrate yields around 45-50% have been identified. Momentum for EMIB-T is rising for Google's TPU v10, supporting the mid- to long-term positive view despite the v9 timing slip. x.comÂ
Intel's near-term growth is constrained by a Kiryat Gat factory blast and memory-driven PC weakness. The weak Q4 outlook calls for answers: the blast looks set to cut roughly $300 million of revenue in Q4 and could have a similar impact in Q1, limiting data-center growth. Fab ownership gives Intel pricing power and enterprise demand is strong; external foundry customers would add another tailwind if won. CEO Lip-Bu Tan said Nova Lake will allow Intel to take back share. x.com x.com x.com x.com … Show more
Intel looks a tad overvalued near $100, so it is not a buy right now, even though its fundamentals are strong. It is state-backed, demand is so high that lower-grade chips are sold rather than discarded, and its downstream stack, including EMIB packaging, is the strongest in about a decade, possibly outpacing its foundry business per Morgan Stanley reporting. x.comÂ
SK hynix weighing Intel Foundry for HBM4E base dies would be the validation Intel Foundry has been waiting for: a customer judging it on cost and capability for the most in-demand product in semiconductors, not a subsidy or national-security case. Per @jukan05, SK hynix wants to diversify beyond TSMC, whose base die reportedly costs 3-4x the core memory die, a cost it cannot pass through under long-term HBM deals. Its packaging VP is also evaluating Intel's EMIB against TSMC's CoWoS. x.comÂ
