00:00We are looking at chipmakers, memory chipmakers in particular, and joining us now for more
00:03on what we've learned here from Google is Mandeep Singh, global head of technology research with
00:09Bloomberg Intelligence. So memory chipmakers were some of the biggest winners in tech and in the
00:14stock market overall in recent weeks because of AI and all the optimism it's generated. Now Google
00:19is saying it might have something to disrupt all of that. Is what Google is saying new? Is it
00:24unexpected? No it's not new. I think it references a paper that was released one year back. But I
00:31think what it goes to show is because of the constraints around memory right now everyone is
00:38talking about you know releasing new products and obviously there is a big mismatch when it comes to
00:43supply demand especially for HBM. And they're saying we can do it better than the rest of the peers when
00:51it comes to the compression algorithm that they have and what it means in terms of their overall
00:57infrastructure versus you know everyone else who's spending you know over 100 billion dollars in
01:02capex on almost 30 percent could be going to procuring memory. Is this theoretical right now or is this
01:08actually happening what Google has said its algorithm can do. I think that's where Google has that
01:13advantage where they have their own TPUs accelerators. They're not relying on Nvidia to really
01:20implement a lot of these novel techniques. And it could very well be the case. You know you look
01:27at the performance of Gemini. I mean it sort of it shows you that how well they have been able
01:33to
01:34improve the latency and also be able to deploy that scale. I mean remember Google is doing it across
01:41their family of apps. They have transformed search. So from that perspective I think the memory shortage
01:47probably isn't hitting them as hard as it is you know at some of the other peers. Do you think
01:54that
01:54other companies in addition to Google had can come up with a similar kind of compression algorithm. I mean I
02:00would be surprised if others don't copy what Google is doing like typically when something like this happens
02:07everyone would implement those sort of efficiencies. Remember DeepSeek? Everyone pivoted to a reasoning model within the next six months.
02:15And guess what? The demand actually took off because you know everyone implemented that and and it
02:22helped you know drive more usage. So if anything I think it should drive more usage and the effect of
02:30that will be more memory demand. All right so Mandeep Singh clearly a proponent of Jevin's paradox where
02:36the more efficient a technology becomes the more demand will actually increase. Thank you so much
02:40Mandeep Singh of Bloomberg Intelligence.
Comments