You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
On CNBC this morning, Google, Amazon and Microsoft all want to build their own AI silicon. Any thoughts or comments on this appreciated. It looks like TSM may become and stay the leader in this area with their advanced processes. I feel this competition and amount of applied resources may make the AI/ML revolution one of the fastest in history and have a dramatic impact not only on business, but everyday living in a time frame that at one time would be considered unimaginable. It seems everyone is piling on the AI bandwagon in everything from chips to ecosystems to applications in almost every field and endeavor. Any thoughts on more advanced processes in making AI chips would be appreciated.
The Challenge is the hype vs reality. AI is growing very vary fast (See nvidia). But the hype is growing infinitely faster. As with all things, the financials will decide what really happens. Does it make sense for Google to develop their own chip if it only needs 200K units? Do they want to differentiate? Can nvidia, Intel, AMD continue to own the market by lowering the margins to reasonable numbers? Do AI applications pay for the billions spent to Nvidia and TSMC. We will see but I would expect a correction that follows the Gartner hype cycle (which I love).
On the technologies, I would expect it to be fastest processors, with tons of chiplet integration for power. Think of it as a supercomputer on a PCB.
