# data center revenue — X 热门讨论 (2026-09-26 22:50 UTC)
## @SemiconductorsX (Semiconductor Insider) · 09-26 07:59 · ♥35 ↻3 💬1 Intel sold NAND to SK Hynix for $9 billion in 2021 so it could stop funding a cyclical flash line and put the money on logic. Solidigm later caught the AI SSD wave. That sale still matches the diagnosis: the last decade broke on process, not on NAND.
The Lip-Bu Tan print is the test of that bet. Q2 2026 revenue was $16.1 billion, up 25% year on year.
Data Center and AI was $6.3 billion, up 59%. Intel 18A is in high-volume manufacturing at about 80–85% yield and 30,000 wafers a month.
18A-P is in risk production: 9% more performance at the same power, or 18% less power, plus about 30% frequency at 0.5 V on production silicon from RibbonFET GAA and PowerVia backside power.
Yields moved from the mid-60s to the mid-80s in one quarter. 18A-P stays design-rule compatible and adds Power Boost plus thermal work.
Foundry interest from Apple, NVIDIA, AMD and others is building. CPU demand is above supply.
Panther Lake and Diamond Rapids Xeons ride the new node. EMIB yield is 90–95%. 18A foundry output has run about 25% above internal targets in recent quarters.
14A high-volume is still aimed at 2028. It is the only U.S. name doing both leading-edge R&D and volume wafers.
Empty fabs were the old problem. A full book is a better one. The NAND exit funded the process work. The process work is what the stock is pricing.
$INTC > 引用 @lucaspinker: @SemiconductorsX The old Intel failed for a reason https://x.com/SemiconductorsX/status/2103756377269408005