According to the "high growth rate" curve in the graph above, the blockchain will be 10,000x as large in 2035. Here's a quick estimate of some of the costs to run a node twenty years from now, assuming no change in the cost of disk space or Internet bandwidth for 20 years.
Nice work Peter R. I often find your presentations compelling and this is no exception.
One quibble. You gloss over the distinction between miners and nodes:
Even under these (a) optimistic growth rates, and (b) extremely pessimistic cost improvements for disk space and bandwidth, the cost is still not so prohibitive that there would only be "5 big miners left." For example, I'd image that all major research universities, bitcoin corporations, various branches of governments, and power users (whales) would still run nodes (1000+ nodes). The network could actually be more decentralized than it is now.