Intel Foundry’s 14A Process Gains Attention as Analysts Point to Major AI and Chipmaking Opportunities
Intel Foundry may be approaching a major turning point as fresh analyst commentary points to growing customer interest in its next-generation chip manufacturing technologies. After a recent trip to Asia, KeyBanc analysts reportedly returned with a highly positive view of Intel’s foundry roadmap, especially around its 18A-P and 14A process nodes.
The biggest takeaway is that Intel appears to be gaining traction with some of the most important names in artificial intelligence, cloud computing, semiconductors, and high-performance computing. While Intel’s 18A-P process has received plenty of attention, KeyBanc believes many large customers may ultimately lean toward Intel’s more advanced 14A process for future products.
According to the analysts, Intel has secured manufacturing-related deals for networking switches from Broadcom and server CPUs from Google. The company is also said to be involved in producing server CPUs for Microsoft and gaming GPUs for NVIDIA. Beyond those reported wins, other major players, including Anthropic, Amazon, and SpaceX/Tesla Terafab, are said to be evaluating Intel’s 18A-P and 14A technologies.
The most eye-catching claim involves Apple. KeyBanc believes Intel has won a multi-generation, multi-product contract tied to Apple’s M-series and A-series chips. Notably, the analysts suggest this work would be based on Intel’s 14A process rather than 18A-P. If accurate, that would represent one of the most significant foundry wins in Intel’s modern history.
Intel 14A is expected to be a 1.4nm-class fabrication process and one of the company’s most ambitious technologies. It is designed to combine several major advancements aimed at improving performance, power efficiency, and chip density.
One of the key features is PowerDirect, Intel’s backside power delivery system. This technology moves power wiring to the back of the silicon wafer, freeing up the front side for data signal routing. The goal is to reduce electrical resistance, improve efficiency, and allow more compact chip designs.
Intel 14A is also expected to use RibbonFET 2 transistors with a second-generation Gate-All-Around architecture. This approach gives Intel tighter control over extremely small transistors, helping improve drive current while reducing power leakage. For customers building AI accelerators, mobile processors, server CPUs, and GPUs, those improvements could be highly valuable.
KeyBanc estimates that Apple alone could represent around 600,000 wafers on Intel’s 14A process. That would translate into a potential revenue opportunity of roughly $15 billion to $20 billion for Intel. Meanwhile, Anthropic’s possible custom ASIC orders could reach 5 million to 10 million units, representing an estimated opportunity of $40 billion to $80 billion when advanced packaging is included.
Advanced packaging is becoming just as important as leading-edge chip production, and Intel appears to be positioning itself aggressively in this area. KeyBanc estimates that Intel’s EMIB-T packaging technology could generate between $12 billion and $15 billion from Google-related AI chip programs, including millions of units tied to the Humufish TPUv9 and Icefish TPUv10.
Intel could also earn between $2 billion and $3 billion by packaging around 1 million units of Qualcomm’s AI300 ASIC, according to the estimates. In addition, Intel’s EMIB-M technology is expected to gain momentum through Amazon Trainium 3 and Microsoft Cobalt 400 orders.
Intel’s EMIB technology allows multiple chiplets to be connected inside a single processor package using tiny silicon bridges embedded in the package substrate. Instead of requiring a full silicon interposer, EMIB uses localized high-density connections only where chiplets need to communicate. This can reduce cost, improve efficiency, and enable more flexible chip designs.
EMIB-T takes the concept further by adding Through-Silicon Vias, or TSVs, directly through the embedded silicon bridges. These vertical pathways allow power and high-speed signals to move from the bottom of the package through the bridge and into processors or memory stacked above it. This enables more advanced 3D chip packaging, which is increasingly important for AI processors that require high bandwidth and low latency.
The EMIB-T substrate relies on several important components. It starts with a base organic panel containing precision-drilled cavities where silicon bridges are placed. These panels are supplied by companies such as Ibiden, Shinko, Unimicron, and AT&S. Insulating dielectric build-up layers are then laminated over the embedded bridges using industry-standard ABF materials. The silicon bridges themselves contain TSVs that help route power vertically through the package.
However, there are still manufacturing challenges. KeyBanc notes that Intel’s EMIB-T substrate yields are currently estimated at around 50% to 60% at Ibiden and only 20% to 30% at UMTC. Those yield levels could limit the initial production ramp for Google’s Humufish TPU to about 5 million to 6 million units.
Even with those constraints, other analysts believe Intel’s advanced packaging capacity could scale meaningfully over the next several years. One estimate suggests EMIB-T could support more than 2 million substrates in 2028 to meet demand related to MediaTek’s work on Google’s next-generation TPU products. MediaTek is believed to be a key design and development partner for Google’s future AI chip series, including the 2nm-class Humufish and its upgraded Triggerfish variant.
Intel may also receive another boost from its Foveros packaging technology. Foveros enables true 3D active-die stacking, allowing chips to be stacked directly over a base die or interposer. By using extremely dense vertical interconnects and copper-to-copper hybrid bonding, Foveros can deliver very low signal latency and high bandwidth, making it attractive for future AI accelerators and high-performance processors.
If Intel can successfully execute on both its 14A process and advanced packaging roadmap, the company could become a much stronger competitor in the global foundry market. The combination of leading-edge manufacturing, backside power delivery, Gate-All-Around transistors, EMIB-T, EMIB-M, and Foveros gives Intel a broad technology platform at a time when AI chip demand is surging.
The bigger picture is clear: Intel Foundry is no longer just a long-term turnaround story. If the reported customer interest turns into high-volume production, Intel could secure a major role in the next wave of AI infrastructure, custom silicon, server processors, and advanced chip packaging. For investors and technology watchers, Intel’s 14A and packaging roadmap may be among the most important developments to track over the next few years.
