Quick thought: Should Intel spin off it's fabs?
Their current business model isn't likely to work for fab customers
I read SemiAnalysis occasionally and they just released a great run-down on how making fabs a separate business unit is opening up some cost savings. The problem is whether these cost savings will be enough to make fabs a successful business for Intel.
I believe the biggest challenge for Intel isn’t whether it can cut costs or catch up in process size to TSMC and others, but whether it can land enough external customers to make fabs a successful business model. Intel’s achilles heel in fabs is it directly competes with the largest potential customers, Apple, AMD, nVidia, Samsung, etc. How resistant to giving their cutting edge designs to Intel employees for fabbing will those companies be? I suspect the answer is “quite”.
Not only is there the risk of directly tipping their future roadmaps to a competitor, but by giving Intel Fabs business they are directly subsidizing that competitor. It’s not an impossible task, as nVidia has publicly said they are open to Intel making their GPUs, but said nothing about whether they would provide their highest volume or cutting edge designs. Again, I’m skeptical that if nVidia does provide Intel Fabs with business that it will ever be in large volumes or for their most important products.
If Intel isn’t able to land a large amount of external business for their fabs, they are at risk of lacking the volume to lower their cost curves as rapidly as TSMC. And if they have to offer lower pricing than TSMC to land clients, that’s a double whammy of higher costs and lower revenues. TSMC may have great margins, but fabs are incredibly capital intensive, without TSMC’s massive volumes it’s margins would be small or even negative.
These are all reasons why I’ve been skeptical of Intels pivot from their core business of building x86 compatible proprietary processors that still has a significant moat, to pour massive amounts of capital into the moatless world fabs. Even if they succeed there is a risk they just split the business with TSMC in a price war where both earn terrible returns on capital. But its too late to turn back now, so what is the best way to ensure Intel Fabs has the greatest chance of success?
This is where a spinoff comes into play. An independent fabs company does not have the same baggage of directly competing with its customers that Intel has, which should make landing customer orders much easier. It also removes the massive ongoing capex requirements from Intel’s balance sheet. Obviously this isn’t part of Pat Gelsinger’s plan, or it would already be in motion. But when their fabs are truly ready to ramp high volumes of production and are experiencing how difficult it is to land customers against TSMC, Samsung and others, the board and the CEO at that time will have few choices other than a spinoff to improve their fabs competitive position significantly.