I think this is a crucial point. Based on Ben's premises, his conclusions make sense. But what if you alter the premises, for example, by assuming that compute will happen on another substrate? If you choose a biological substrate, then you can move compute from inside one's pocket, to inside the body. And for many functions, you wouldn't need the cloud. I doubt that the dominant companies in silicon-based tech today have the expertise to make that shift.

A lot of work is being done to make bio-silicon fusion real, with use cases like creating olfactory sensors.

And our increasing control over both brain and genes may be the pathway to more general biological computation.

https://www.ucsf.edu/magazine/control-brains-genes