The document that decides whether Intel can sell manufacturing to the rest of the chip industry is not a press release. It is a process design kit: thousands of pages of electrical rules, device models and test structures that tell a designer what a factory can and cannot build.
Intel Foundry delivered version 0.9 of that kit for its 14A process to customers last week. The 0.9 release is the first to carry silicon-verified data, and the company says the node is close to mature on performance and yield. Version 1.0 is scheduled for the first quarter of 2027. Volume production is targeted for 2028.
The sequence matters more than any single date. A design kit at 0.9 is good enough for customers to begin serious work and not good enough to sign off on. Models still move, rules still change, and a block that simulates cleanly in September can fail in March. Version 1.0 is the release a customer can hand to a design team without scheduling a second pass.
For Intel, clearing that gate is a precondition rather than an achievement. The company’s foundry business has spent years arguing that it can manufacture for other companies at the leading edge, and the customers it needs — large design houses, cloud operators building their own accelerators — plan products two to three years ahead. A 2028 production date means the decision to use 14A is being made now, before the node has shipped a commercial wafer.
That is the arithmetic of the business. A leading-edge mask set costs tens of millions of dollars, and a design team is committed for years. No purchasing department takes that risk on a promise. What it wants is a kit with silicon behind it, a yield curve it can read, and a schedule that has not slipped twice.
The 14A name refers to the 1.4-nanometer generation in Intel’s angstrom-based nomenclature. The numbers are marketing as much as physics, and no two manufacturers count them the same way. What they describe is a recipe with a fixed set of capabilities: how small the features are, how dense the wiring, how much power the transistors leak, and what the factory can guarantee across a wafer.
The 0.9 kit tells customers where those numbers stand. It does not tell them where they will be at 1.0, and Intel has described its progress in relative terms. “Close to mature” is a phrase that invites a question about the distance left, and the company has not published the yield figures that would answer it.
The customers Intel is courting already have a supplier. Most advanced logic is made by TSMC, with Samsung as the second source, and both have design kits that have been in production for years. Moving a product to a new node at a new factory is not a matter of price. It is a matter of whether the design closes on time, whether yield holds at volume, and whether the second year of production resembles the first.
Intel’s argument is capacity and geography. Governments in the U.S. and Europe have been paying for advanced manufacturing on their own soil, and a domestic option has value to buyers who worry about concentration in one region. That argument carries weight with policymakers. It carries less weight with a chief executive choosing where to spend a design budget.
There is also the question of what 14A is for. Intel has tied the node’s build-out to demand from outside customers, which makes the kit a test of the sales pipeline as much as of the process. If the 0.9 release produces a queue of designs, the schedule holds. If it produces polite interest, the company has a factory plan without a customer list.
Analysts who follow the foundry business have been watching for this signal rather than for a customer name: evidence that designs are being drawn. A design kit in customers’ hands is the earliest form of that evidence, and it stays invisible from outside until a product ships.
The next checkpoints are easy to list. Version 1.0 in the first quarter of 2027. Tape-outs, which the company has not disclosed. And a production start in 2028 that will be judged against what TSMC and Samsung are shipping that year, not against where Intel stood in 2025.
For the design teams now opening the 0.9 kit, the calculation is narrower. They have a process measured on silicon, a schedule that held for this release, and a competitor with a much longer record. The kit answers the technical question. Whether it answers the commercial one is a decision each of them makes in the next few quarters, and Intel will learn the answer the way everyone else does, by counting the designs that come back.
