I thought the bigger problem with the claim was that it didn't mention that Intel did that with:
1. Ultra-expensive, full-custom design.
2. A massive budget from a massive amount of sales to fund No. 1.
Most ASIC's being made use standard cell wherever possible because the companies, even very profitable, either couldn't afford full-custom or didn't think the cost was worth it. The few that do it are kind of like elite class of chip makers with piles of money. So, it seems improper to compare what a smaller effort on low-tech nodes can do with an elite, chip maker on what was then cutting-edge node. Maybe smaller players can do it today with lessons learned, current tech, lower cost of labor, and so on. Still be skeptical.
Note: I am keeping up with old techniques and processes at 350nm and up for subversion-resistant chipmaking. We stopped being able to visually look for backdoors at around 250nm or so. That means chips that will be verifiable by large numbers of people will have to be above that. I think I'd get 100-400MHz like those in this article, though, given my counter-arguments above.
