This was not just informative but so well written. You're an inspiration.
Question: Nvidia is generations ahead of the startups. You commented on the path to AI on the edge- what will be the evolution of compute intensive, parallel processing training?
Thanks Apoorva! I think we're seeing the evolution of that training from Nvidia right now: systems-level designs at the data center level, building larger and denser data centers, and then looking for ways to optimize costs and power consumption. Energy continues to be a bottleneck, and there's no one solution to solve it (see previous article on energy). Finally, customers will continue to pay for leading-edge performance for the foreseeable future.
This was not just informative but so well written. You're an inspiration.
Question: Nvidia is generations ahead of the startups. You commented on the path to AI on the edge- what will be the evolution of compute intensive, parallel processing training?
Thanks Apoorva! I think we're seeing the evolution of that training from Nvidia right now: systems-level designs at the data center level, building larger and denser data centers, and then looking for ways to optimize costs and power consumption. Energy continues to be a bottleneck, and there's no one solution to solve it (see previous article on energy). Finally, customers will continue to pay for leading-edge performance for the foreseeable future.
Seems like the big winner in lots of different scenarios is Taiwan Semiconductor. The risk, of course, is military action by China.
Well said, incredible company. Likely a future deep dive on here.
Superb article, thank you...
Thanks Leslie!
Interesting article, i liked the deep dive into the details. It relates to an article I released few weeks ago about an overview of the semiconductor industry: https://reassembler.substack.com/p/inside-the-chip-wars-power-profit-fight-over-the-new-oil