tapeout
Tapeout is the finish line of chip design: the moment you hand the completed, fully verified layout to the fab to be manufactured. Picture an architect who has spent months tweaking every wall, pipe, and outlet, then finally rolling up the master blueprint and couriering it to the builder. Once it leaves your hands the design is frozen and the digging starts — there is no nudging a wire after this. That is why tapeout sits at the very end of the back-end flow, only after floorplan, placement, clock-tree synthesis (CTS), routing, then extraction plus timing signoff (STA), and finally DRC, LVS, and the power checks have all passed clean. You do not tape out hopeful work; you tape out work that has survived every signoff gate.
What actually ships is a layout database — historically GDSII, increasingly OASIS — describing every polygon on every layer. It goes to the mask shop, where each layer becomes a photomask: a quartz-and-chrome stencil that lithography uses to print that layer onto the wafer. Because committing a full mask set at a leading node can cost millions of dollars and weeks of fab time, tapeout is the true point of no return — any bug caught afterward means a costly respin. Then you wait. Weeks later the first wafers come back as 'first silicon,' and the team moves to bring-up: powering the chip on a board, checking it boots, and measuring whether it actually meets the timing and power you signed off in simulation.
Tapeout is the last step of the physical-design flow: every prior gate must pass before the GDSII leaves for the fab.
The name is a fossil — finished layouts were once written to magnetic tape and physically mailed to the fab, so completing the design was literally 'taping out'; the tape is long gone but the word stuck, and a clean tapeout still does not guarantee working silicon, only that the design met every check you knew to run.