No Priors · Friday, October 2, 2026
Walter Goodwin observes that while many AI ASICs currently appear similar, relying on common components like HBM and tensor cores, there's a growing need for differentiated accelerators. Companies like Fractile aim to offer new capabilities, particularly in memory bandwidth, to address evolving workload demands.
“One of the things that I think is very striking is there's a lot of relatively identikit chips out there. And this is kind of a structural, it's a kind of structural industry property where if you look especially across kind of hyperscaler efforts and, you know, Google kind of kicked this off with the TPU more than 10 years ago, you see this paradigm where although there are a number of kind of proprietary internal chips, Microsoft with Maya, OpenAI now with Jalapeno. These chips that are ultimately developed and delivered in partnership with a relatively small number of what we might call ASIC houses, these kind of back-end delivery houses, Broadcom is the largest, a $2 trillion company, that help people kind of realize these designs.”
“But I think what's exciting about building chips that enable fundamentally new capabilities, is you then also get to explore whether there are kind of gravitational pulls that you yourself get to pull the new model landscape in because of those properties. And so for Fractile, you know, this idea of having like 25 times more bandwidth per chip than an HBM-based chip, we get to explore, for instance, internally, these kind of ideas of scaling laws for bandwidth.”
“I think as you then see this move into chips that actually give new capabilities, you know, that's a place where, you know, I suddenly see for what we're doing, real need for everybody that wants to deploy AI at the frontier, they have to have some solution to running these models orders of magnitude faster.”