Intel kills three projects, hits 90% EMIB

Intel is trying to fix two problems at once. One is strategic — too many bets, too much overlap, and not enough clarity about where Intel actually wins in AI systems. The other is physical — if you want to sell giant AI packages, your advanced packaging has to work at scale. This week, those two threads snapped into focus: Intel has reportedly killed three internal projects in a fresh reorganization, and at the same time its EMIB packaging line is being framed as finally good enough to matter. (digitimes.com) ### Who’s actually reshaping Intel? The big picture still starts with Lip-Bu Tan, who took over as Intel CEO in March 2025 and has spent the last year talking about rebuilding product leadership and tightening execution. But the operating story now looks more specific. DigiTimes says the latest reset is being driven inside the product and networking stack by Mi(digitimes.com)d what gets cut. (newsroom.intel.com) ### What got cut? The cleanest reported fact is the count: three internal projects were killed in roughly two months. The paywalled reporting points to a multi-year reset rather than a one-off cleanup, which matters because Intel has spent years spreading effort across CPUs, GPUs, foundry, networking, software, and custom silicon without always making the (newsroom.intel.com)equally sacred. (digitimes.com) ### Why lean back toward CPUs? Because Intel still has a natural control point there. In AI infrastructure, the CPU increasingly acts like the traffic cop — orchestrating memory, networking, storage, and the accelerators around it. That does not mean GPUs are suddenly unimportant. It means Intel seems to be recentering the system around the thing it already own(digitimes.com)s. That is a much more defensible posture than trying to out-Nvidia Nvidia everywhere. (digitimes.com) ### What is EMIB, in plain English? EMIB is Intel’s way of stitching multiple chiplets together inside one package without building one giant monolithic die. Think of it like using short, dense bridges between neighborhoods instead of paving one enormous slab of city. For AI chips, that matters because the package now carries a huge part of the system challenge(digitimes.com)in one place. (wccftech.com) ### Why does 90% yield matter so much? Because advanced packaging only helps if it is manufacturable. Wccftech says EMIB yield has reached about 90%, which, if the figure holds up, is the kind of threshold that changes the conversation from “interesting lab capability” to “maybe this can ship in volume.” The sa(wccftech.com) is not a small tweak — it is Intel saying it wants to compete on package scale, not just transistor slogans. (wccftech.com) ### Why is Wall Street suddenly excited? Because the market is starting to believe Intel might have an actual turnaround shape. CNBC says Intel stock jumped 114% in April, the best month in the company’s 55-year Nasdaq history, after a strong earnings beat and growing enthusiasm around AI demand. Intel reported Q1 2026 adjusted EPS of 29 cents on $13.58 billion in revenue, both ahead of expectations. (cnbc.com) ### So what changes for customers? The sales pitch gets more neutral. Instead of forcing every workload toward an Intel GPU story, Intel can argue for a system-level choice: use CPUs where orchestration and general-purpose compute matter, GPUs where parallel throughput wins, and ASICs where a fixed workload justifies custom sil(cnbc.com) help build the right one.” (digitimes.com) ### Bottom line? Intel’s cuts and packaging progress are the same story from two angles. The company is trying to become narrower in strategy and bigger in system ambition. If that works, Intel stops looking like a chip company chasing every lane and starts looking like a platform company choosing its shots.