To what extent is the additional cost of a new bike worth it compared to purchasing a used frame and building a more customized machine?

In another post, a person said they had a $10k budget to buy a tri bike. Most folks suggested buying the latest superbike out there, and to an extent, for good reason. The newest bikes have the latest bits and pieces on them, and are theoretically the fastest bikes ever made. But, are they?

If we live in an "age of aero" in which gains with current frame design are marginal at best, are the newest crop of superbikes worth the price? How much has a Canyon, Cervelo, (insert brand here) improved in the last five years or so? To what extent is the additional cost of a new bike worth it compared to purchasing a used frame and building a more customized machine?

Are the newest crop of superbikes worth the price? ...

To what extent is the additional cost of a new bike worth it compared to purchasing a used frame and building a more customized machine?

The term superbike has just come to mean that it has dedicated bars - not a reflection on performance at all. But of course, the implication is that they are overall 'super' which makes it a terrible misnomer.
The number one thing that makes a tri/tt bike fast is allowing the rider to get into their best position - which means a good size range and bars with both fine adjustment and a good range of adjustment.
Number 2 and 3 are overall aero or fuel carrying - order depends on your priorities.

There are a lot of so called superbikes that are crippled by their lack of adjustability
Ventum - open mould bars with very limited range
Giant Trinity - very limited range
Cervelo P5D with EX11 - great Y with monopost but very limited Z and no tilt. Switch to EX10 and it is good.

Cervelo P5-Six, Scott Plasma 5, Shiv Disc.. basically the list includes nearly all top level bikes with proprietary bars until we get to the Speed Concept (which has a great range but it's complicated) and P3X, P5X. BMC TM01 if you switch to an Evo bracket.

On the Aero side - Cervelo data showed the P5D to be 9s faster over an IM than a P5. But the P5 could be optimised with better bars
Tour Mag data showed the P5X to be slower than the P5 when not loaded up.
Tour Mag also showed that the P5 doesn't really have many close competitors from other brands (BMC was next best).

In the end it depends on your priorities - if you want a bike that is clearly top of the line buy a 'superbike'
If you want to be the fastest you can be - figure out how to spend your $ to get the best possible position and a bike that supports that. Then optimise all the small factors. It's relatively rare for a 'superbike' to be the best option with this approach.

How much has a Canyon, Cervelo, (insert brand here) improved in the last five years or so? To what extent is the additional cost of a new bike worth it compared to purchasing a used frame and building a more customized machine?

Don't get a superbike unless you are fast. It would be a bummer to get passed by a lot of people on lesser bikes in a race.

The Cervelo P5, P3x, and P5X are probably the fastest out there still...

Don't get a superbike unless you are fast. It would be a bummer to get passed by a lot of people on lesser bikes in a race.

This is what keeps me from buying a PX, or even P5. I don't want to be that guy, so I just upgrade my bike and fly under the radar

bikes of the last decade are minor compared to fit geometry.

Cumulatively, that's about 20 watts faster for a 2019 P5 than an old P2C circa 2007.

If I have these numbers right and If the latest generation of "superbikes" is about the same as a 2019 P5, it's kind of amazing that maybe 20 watts is all the improvement that's been made in the last 12 years from 2007.