ASICs and on-premise AI infrastructure boom
The next phase of AI infrastructure spend will shift toward on-premise ASICs rather than cloud-based GPU clusters, benefiting specialized hardware providers.
The argument
The speaker argued that while massive GPU clusters in the cloud are used for training AI models, enterprises will build out localized, on-premise 'AI factories' using ASICs to protect sensitive data from cyber threats and cloud vulnerabilities.
The thesis, stress-tested
✓ What validates it
- ✓Accelerating revenue growth in Marvell's ASIC division
- ✓Enterprise capital expenditure shifting from cloud subscriptions to on-premise hardware
▸ Risks discussed
- ▸Supply chain bottlenecks in advanced semiconductor manufacturing
- ▸Rapid obsolescence of specific ASIC designs
Hear it yourself
"The cloud, which is where these massive GPU clusters are training the IQ of the brain. So those continually need to be built out because we need to solve fusion. We need to solve longevity."
00:00 / 00:14
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE