Tape-out
The tape-out is the moment when the finished design of a computer chip is sent to the factory that manufactures it. After that, nothing in the design can be changed anymore, which is why this step is regarded as a milestone in the chip industry.
A computer chip first comes into being as a plan on a computer. Engineers determine where billions of tiny switches sit and how they are wired together. This plan is checked, simulated, and refined over months or years. At some point it is considered finished and is handed over to the factory that physically produces the chip. This very handover is called the tape-out. The name dates back to the 1970s: back then, the design data was shipped on magnetic tapes that were carried out of the data center.
The point of no return
Software can be fixed after release with an update. That is not possible with a chip. As soon as the factory starts, the pattern is burned into silicon. A flaw in the design is then present in every single unit produced. That is why the tape-out is a high-risk decision.
The risk is also financial. For a modern high-performance chip, merely preparing production often costs several dozen million euros. The biggest expense is the masks: these are stencils used to expose the circuit pattern onto the silicon wafer. A chip made at a very fine process node needs 60 to 100 of these. If a second run has to be started because of an error, the money is gone and the schedule slips by months.
For companies and investors, the tape-out is therefore a signal. It shows that a project has moved out of the design phase and is becoming real. For young chip companies, the first tape-out often decides the next funding round.
From design to mask
A long chain of verification precedes the tape-out. The circuit is first described as behavior, then translated into concrete components, and finally arranged as a layout. Simulations follow: Does the chip get too hot? Do the signals arrive in time? Do all components comply with the factory’s requirements? These requirements are called design rules and specify, for example, minimum distances between wires.
Once everything has been checked, a file is created containing the complete geometry of all layers. Common formats are called GDSII or OASIS. This file is what gets handed over today — magnetic tapes have long since disappeared, but the name stuck. The factory converts the data into mask sets and begins manufacturing.
From tape-out to the first testable chips typically takes three to four months. These first specimens are called silicon or first silicon. They are measured in the lab and put through the wringer with test programs. If minor defects are found, a respin follows: a corrected version for which only part of the masks need to be rebuilt.
Tape-outs in news from the semiconductor industry
The term regularly appears in business news, usually in sentences like “successful tape-out in 3-nanometer production.” What is meant is: a customer has submitted their design to a contract manufacturer such as TSMC or Samsung. Such reports are an early indicator. They signal roughly a year in advance which products are coming.
Practically all chips found in everyday life are affected: processors in phones, graphics cards, AI accelerators for data centers, but also the control chips in cars and washing machines. Every one of these components had a tape-out at some point. For simple chips on older process nodes, it is a routine event; for the largest AI chips, it is an event with its own press release.
A common misconception: tape-out does not mean the chip is finished. It only means the design is finished. Between tape-out and market launch there are still tests, usually one or two rounds of corrections, packaging the chips into their housings, and ramping up mass production. Anyone reading a tape-out announcement should therefore expect to wait about a year with patience.