The chip industry has a big problem at the moment. The world’s leading foundry, Taiwan’s TSMC, has effectively no more wafers it can sell to make more artificial intelligence chips on its three nanometer technology. Nvidia, AMD, and Broadcom – which helps Google and others design in-house AI processors – have dried up capacity.
“TSMC is the real bottleneck,” Semianalysis President Doug O’Loughlin told me in an interview.
That’s why anyone interested in having a shot at producing a new piece of silicon is casting around for alternatives. Samsung and Intel are both in the “danger zone.”
The tentative agreement between Apple and Intel for chip manufacturing, first reported by the Wall Street Journal, confirmed what the industry had awaited for months.
A deal between the two companies sounds good on paper, but to understand what it means for Intel, investors need to know three key things: What chip? What manufacturing technology will it involve? How much volume will it drive? None of that is clear at the moment.
Even with Apple as a customer, the company has failed to keep up with Moore’s law, the prediction that chips will get faster and cheaper every two years.
“No company in history has ever fallen off the Moore’s law curve and made it back on,” Seaport Research analyst Jay Goldberg said.
