Intel also developed a track record of only being able to do a good job on one x86 design at a time. Doing Pentium M and Pentium 4 at the same time didn't work out so well, and neither did Core and Atom. And then for a several years the number actually dropped to zero as they had to keep recycling their existing designs and couldn't get a successful post-Skylake architecture out the door. Now that they're shipping both Core and Atom lineages on the same silicon, they might do a bit better at not leaving one to fall into irrelevance.

That sounds a lot like Google, but Intel doesn't throw as many things at the wall to see what sticks. NUCs have been around for 10+ years; that's much longer than Stadia or Reader.

My greatest fear is when they decide they don't know what to do with FPGAs anymore and get rid of Altera (please do this, Intel, your website doc format is awful compared to good old Altera's)

Funnily enough, Intel did used to make PLDs in the 90s. And back then, Altera actually rebranded and sold them!

The whole hardware industry has boom/bust cycles because the margins are thinner.

Intel can easily launch a new small form factor line, with a new name, next year if the market is looking good.