VLSI & fabrication

tape-out

Tape-out is the moment a chip design crosses from drawing to building — when the finished layout is signed off and handed to the foundry to be turned into silicon. Think of an architect who has fussed over every wall for years finally rolling up the blueprints and shipping them to the construction crew. From here the design is frozen: the bricks are about to be laid, and once the diggers arrive there is no nudging a single line.

Concretely, the layout (a GDSII or OASIS file) goes to the fab, where it is used to make a mask set — the quartz-and-chrome stencils that print each layer onto the wafer. At a leading node a mask set can cost millions of dollars, so tape-out is the point of no return: any bug found now means a respin. Weeks of processing follow, and then the anxious wait for 'first silicon' — the very first wafers back, to see if the chip powers up and runs.

The name is a fossil: layouts were once written out onto magnetic tape and physically mailed to the fab, so finishing the design meant 'taping out.' The tapes are long gone — today it's an encrypted GDSII or OASIS file — but the word stuck. Note that tape-out is not the same as first silicon (the working wafers that come back weeks later), nor as tape-in (an internal checkpoint before tape-out), and a full dedicated mask set is distinct from a cheaper MPW shuttle, where many designs share one reticle to split the cost.

Also called
tapeout流片投片下线下線