Charlie Kawwas head of $Broadcom (AVGO.US)$ semis business w...
Charlie Kawwas head of $Broadcom (AVGO.US)$ semis business was even more explicit and detailed on this today at GS, segmenting the AI accelerator mkt into three parts:
- Internet scale customers ( $Meta Platforms (META.US)$ $Alphabet-C (GOOG.US)$ Bytedance Tencent): significant logic and value for these companies spending $10s of Bns all on internal infra to customize ASICs for own environments. AVGO can bring new XPUs to mkt in a year. Sees most of this mkt going to custom silicon in 5yrs.
- Hyperscale Clouds: more work to do to get adoption of custom accelerators since you also need customers to be willing to work with them. But the leaders have enough scale to justify the hard work and a lot of this market can be custom silicon.
- Enterprise: no reason for these customers to take on the huge fixed cost of an ASIC and this will remain a merchant market.
- Internet scale customers ( $Meta Platforms (META.US)$ $Alphabet-C (GOOG.US)$ Bytedance Tencent): significant logic and value for these companies spending $10s of Bns all on internal infra to customize ASICs for own environments. AVGO can bring new XPUs to mkt in a year. Sees most of this mkt going to custom silicon in 5yrs.
- Hyperscale Clouds: more work to do to get adoption of custom accelerators since you also need customers to be willing to work with them. But the leaders have enough scale to justify the hard work and a lot of this market can be custom silicon.
- Enterprise: no reason for these customers to take on the huge fixed cost of an ASIC and this will remain a merchant market.
Overall he thinks the AI accelerator mkt (sized by others as potentially $400B) will be mostly semi custom within 5yrs. Talking his book obviously.
Disclaimer: Community is offered by Moomoo Technologies Inc. and is for educational purposes only.
Read more
Comment
Sign in to post a comment
