"No one can explain to me what fab-lite means", complained Penn. Does it mean having a small in-house fab with a foundry for volume? It can't be cost-effective with a TSMC 150,000 wafer a month Gigafab. So you get bean-counters saying: Why are you doing this if it's not cost-effective?
"Or does fab-lite mean that you're going fabless at a given technology node?" asked Penn, "but dumping your fabs does not make you fabless." Fabless companies are a lean, keen, mean business model. You don't get there just by dumping your fabs."
"Or does fab-lite mean you continue to build fabs but use foundries to smooth the peaks and troughs of demand?" asked Penn, "that's a classic adversarial, Them & Us, business model which is clearly not in the foundries' best interests."
"Fab-lite relies on an endless supply of cheap foundry wafers," added Penn, "TSMC is the only foundry company making money. TSMC is one of the best companies in the world, but its revenues per square centimetre has been going down and down. We expect this. While we want our revenues per square centimetre to go up and up we expect foundries revenues per square centimetre to go down."
That is unsustainable. And it's particularly unsustainable when the world looks like moving towards several years of shortage of foundry capacity.
"Fab-lite is an illusion - you are neither IDM not fabless", concluded Penn, "potentially it's the worst of both worlds."