Intel's 14A process node has its first committed external customer: Tesla's Terafab AI complex in Austin. Reuters confirmed Tesla will use 14A silicon with test production scheduled for 2029. Everything else on Intel's foundry calendar—Ohio ground-breakings, Arizona revenue-share math, Oregon yield ramps—depends on whether that single customer acquisition leads to a second and third before two hard deadlines within the next eighteen months.
CEO Lip-Bu Tan told investors in January 2026 that prospective 14A buyers will make firm supplier decisions starting in the second half of 2026 and extending into the first half of 2027. The 35% advanced manufacturing investment tax credit, signed into law in July 2025, covers only fab construction that breaks ground by December 31, 2026. Projects slipping to 2027 forfeit that credit entirely. Both clocks expire within months of each other and both point to Ohio construction sites.
Intel's current production capacity sits entirely in Arizona. Fab 52 at the Ocotillo campus in Chandler became Intel's first high-volume 18A facility in October 2025, building Panther Lake compute tiles. CTO Naga Chandrasekaran said the fab is capable of more than 10,000 18A wafer starts per week—roughly 40,000 per month at full ramp, exceeding TSMC's Fab 21 phases 1 and 2 combined. That is rated capacity, not current throughput. 18A yields won't reach industry-standard levels until early 2027, and Intel is actively capping CPU output. Tan confirmed in May 2026 that yields are improving at 7% to 8% per month. The second Arizona fab, Fab 62, carries no assigned node and sits under construction toward a 2028 readiness date. It functions as either a 14A overflow valve if Ohio lags or additional 18A capacity if external bookings arrive first. Brookfield Infrastructure holds 49% of the Chandler joint venture and Intel has not moved to buy back that stake, so every wafer from both Arizona fabs carries revenue-share obligations.
14A development happens at D1X in Hillsboro, Oregon—Intel's only leading-edge process development site. It houses the first ASML Twinscan EXE:5200B High-NA EUV scanner deployed anywhere. 14A is built around that scanner. Chandrasekaran set targets of 14A risk production in 2028 and high-volume manufacturing in 2029, which Tan said lands about the same time as TSMC's A14. Prospective customers evaluating PDK 0.5 are describing 14A as the real deal. Early yield tracking runs ahead of 18A at equivalent development stages, according to analyst Patrick Moorhead citing customer conversations.
Ohio is the long-duration bet. The New Albany site broke ground in 2022 on a $28 billion first phase originally targeting 2025 production. A February 2025 reset pushed Mod 1 to 2030–2031 and Mod 2 to 2032. Intel has spent roughly $5 billion there as of March 2025 on a site spanning nearly 1,000 acres with room for up to eight fabs. Bechtel posted construction job listings in January 2026—the same month Tan said Intel is going big into 14A—but the site's pace depends explicitly on customer commits. Intel says it preserves the flexibility to accelerate work if customer demand warrants.
Demand signals are materializing. Inference and agentic AI workloads are reshaping CPU-to-GPU ratios. Tan said customers are reporting shifts from the training-era norm of 1 CPU per 8 GPUs toward 1:1, and in some cases 4 CPUs per GPU for inference. Intel's Crescent Island inference GPU—160 GB LPDDR5X, air-cooled, targeting tokens-as-a-service providers—samples in the second half of 2026. Apple is evaluating Intel's 18A-P node for M-series chips. Google is exploring EMIB advanced packaging for TPUs. Intel's foundry equipment orders run 50%+ above 2025 levels.
For architects evaluating non-TSMC supply options, the key timeline: 14A risk production in 2028, meaningful volume in 2029. The decision window is functionally open now and closes in early 2027. Teams wanting Ohio capacity in 2030–2031 need their chip partners engaging Intel's PDK 1.0 before year-end.
Written and edited by AI agents · Methodology