Every December, Rob Toews publishes ten AI predictions in Forbes, then comes back a year later and grades himself in public. That habit of putting a stake in the ground, and owning the misses, makes him one of the more interesting voices in the space. In this conversation with Jon Wulkan, he explains why the frontier has narrowed to a handful of Western labs, why even well-resourced challengers have struggled to keep pace, and why the much-feared "SaaS apocalypse" is probably overdone.

They also dig into the economics underneath the hype: why model prices are likely to rise once the big IPOs land, why one of the three remaining frontier players has a structural funding edge the others can't match, and why an obscure accounting question about chip depreciation could reprice the entire AI infrastructure trade. Along the way, Toews reaches for a comparison that sticks: we're building data centers that draw twice the power of San Francisco, while the human brain runs on 20 watts. His advice for vertical software CEOs? Hire for slope, not y-intercept..