In short
First-silicon success is at a twenty-year low and roughly three quarters of chip projects run late, according to the industry's own surveys. Every response I see is an engineering response: more verification, better tools, bigger teams. The numbers keep getting worse because the cause is not engineering.
The numbers
The Siemens and Wilson Research Group functional verification study for 2024 put first-silicon success at 14 percent, the lowest in the twenty years the survey has run, and found that around three quarters of chip projects finish behind schedule. These are not new trends. They have been moving the same direction for a decade, through enormous improvements in EDA tools and verification methodology.
Why better engineering has not fixed it
If the cause were engineering, two decades of better engineering would have moved the number. It has not, which tells you the cause lives somewhere the tools do not reach. Read the post-mortems and it is always there: a coverage curve that flattened and was reported as on track, a capacity request deferred to a quarterly review, an IP drop that moved and was absorbed by cutting a regression, firmware planned to a date that changed. Each one is a small decision made without seeing what it would cost downstream.
Chips got bigger and chains got longer while the way programs are run stayed the same: status collected by hand, reviewed on a cadence, decided on information that is already old. The complexity grew past what a review can see.
What would move the number
Not another tool for the engineers; they have good ones. A layer above the tools that holds every commitment to the date, sees drift the hour it starts, and puts the fix in front of the person who owns the date while it is still cheap. That is a change to how the program is run, not to how the chip is designed, and it is the one change the industry has not yet made at scale.
I do not think the number is a verdict on the industry's engineers. I think it is a measure of how far programs have outgrown the way they are seen. That is a problem I know how to fix, and it is why Tymeline exists.