It's worse for trusted hardware than most people think. The framework I came up with predicted a number of attacks including analog and material swapping at fabs. So, that's an initial test. Here the basic risk analysis:

https://news.ycombinator.com/item?id=10468624

The smartphone analysis I did also has general stuff in it:

https://news.ycombinator.com/item?id=10906999

I predicted the A2 paper on analog compromise happening, at a high-level rather than specific attack, largely due to our hardware guru on Schneier's blog bragging about mixed-signal attacks years ago. He taught us they resisted attempts to counterfeit or patent sue them by disguising key functions in analog or RF components the digital tools couldn't even see. He said competitors did, too, with him regularly having to carefully inspect lowest-level representation of 3rd-party components. I have a feeling they were cloning them, too. ;) Anyway, the actual products were already subverted years ago just for competitive advantage, counterfeiting, etc. He said some counterfeiters were so good they cloned his company's products down to the transistors. I said, "Holy shit!"

One more thing for you while I'm still on this: cost reduction via merged designs. The mask and fab runs for prototyping cost tons of money. A well-known way to reduce that is many companies sharing one run (eg shuttle run or MPW) to get their test chips cheaper. A less-known trick, at least outside embedded, is them putting multiple products on one ASIC to do same thing for production runs with a factory-setting telling it what chip to look like. My hardware guy gave example of 3G or WiFi circuitry embedded in microcontroller used in your input devices... perfect for keylogging... that was only incidentally there since supplier offered both feature-phone SoC's and peripheral chips. And simply didn't want to manufacture two lines. Such extras might be re-enabled, even remotely, depending on how they control access to them. So, gotta watch out for them.