I think the author summarized his hypothesis quite well in the title: The Transaction Costs of Tokenizing Everything [... are too high compared to the costs of bandwidth, etc.]. If you introduce a bidding market for every computational resource then that itself creates new costs (keeping all the systems running that create bids, check if they have been filled, cancelling old ones, drafting bidding protocols, etc.) that might largely offset the gains in more efficient resource allocations.
That's like going up to Eisenhower after he proposed the interstate system and saying, "Wait, the cost of paving over that much land and building that many overpasses might largely offset any gains of this supposed more efficient transportation network."
All new technologies introduce costs, but ideally they also act as multipliers that bring far more benefits.
The multiplier effect might not be enough, but then again, it might be. You're not really saying anything.
Better to talk about under what circumstances the benefits-multiplier might not be enough to make up for the costs. For example, someone elsewhere in the thread said that yes, if humans were bidding on such tokenization, it might create too many barriers. But look at all the ad networks we've built where billions of tiny auctions happen instantaneously -- that's all by machines. So if the system is built for machines, the multiplier effect could be potentially huge.
I think the underlying problem is that being in charge of the transaction system confers so much power. The promise of that power creates a huge incentive to build such systems (and actively undermine any technical or political threat to their influence) even if it's worse for everyone else.
