HP probably sold tens of billions of US$ of Spectrum (Precision Architecture).
Itanium was a technology and Intel-partnership success, but AMD was able to steer the market to stay with the x86 instruction set architecture. In the end, the better technology isn’t what wins, it’s what the customers choose.
The Machine isn’t hard to understand, but it upends 60 years of “how computing is done”. My own view is that using persistent memory as shared storage, which requires a major application rewrite, and which upends the way Intel makes money, will be an important 2020s technology. Martin just ran out of runway.
Oh, and SOS (silicon on sapphire)? That was already in trouble when I started at HP in 1980, upstairs in the same building where the SOS fab was. Turned out the defect density could not be brought down enough to get chip yields above (say) 1%, so wasn’t useful for building commercial products. SOS was radiation-hard and found its niche in chips for satellites.