The Great Custom AI Chip Trend (with Karl Freund)
Manage episode 405922768 series 3556820
In this installment of the neXt Curve reThink Podcast Silicon Futures series, Leonard Lee of neXt Curve and Karl Freund of Cambrian-AI Research talk about the emerging trend of custom AI chips that is starting to get a lot of attention in the semiconductor industry. Why is this happening and what is driving the interest and the economics of going with ASICs for AI and accelerated computing?
The duo talk and debate about:
(1:37) The custom AI chip trend
(2:34) Chiplets are changing the game
(4:37) The AI silicon premium and hyperscalers
(6:30) NVIDIA's play and motivations for custom chips
(7:51) NVIDIA's general-purpose accelerated computing
(8:45) The ASIC inflection point for Generative AI
(10:45) The expanding chip design ecosystem and options
(14:05) Is NVIDIA becoming a heterogeneous computing company?
(16:44) NVIDIA, the systems company
(17:34) Custom designs & IP to support growing need for application specific features
(19:39) The thinking behind that $30 billion TAM for custom chips
(21:24) CUDA - the software-defined FPGA based on GPU
(23:40) Follow the custom chip money!
Remember to like, subscribe to the neXt Curve reThink YouTube channel. Also, follow neXt Curve by subscribing at www.next-curve.com.
Check out Cambrian-AI Research at www.cambrian-ai.com and get in touch with Karl Freund on LinkedIn. You can also read his research on Forbes where he has a column.
118 episodes