00:00We talked about the retail interest was quite massive.
00:03Tell us a little bit more about the company.
00:05Yeah, it's a very interesting company.
00:07It's founded by an ex-AMD employee.
00:10And it really does seem to be the playbook for many of these Chinese startups these days
00:14that they work in a big international company.
00:16So for this particular company, MetaX, the CEO worked for 14 years at AMD.
00:21Five years ago, he left to focus on his own startup.
00:25And he has been hiring aggressively a crop of top-tier talents.
00:29One of his first hires was the CTO, who also was one of the first Chinese female scientists for AMD as well.
00:38So very similar to MoreThreads, which is also started by an ex-NVIDIA employee.
00:43And we saw the retail portion of the IPO oversubscribed by nearly 3,000 times
00:49and really eye-popping sort of gains that we saw at the trading debut today.
00:54It really makes me question as well what happens in the underwriting process.
00:57You know, how do they value a company that is still unprofitable at this point?
01:01Yeah, it's interesting too, Bel.
01:02I mean, we talk about, you know, China's really trying to create these national champions
01:06and these sort of alternatives to NVIDIA.
01:10Bel, how does really MetaX stack up among some of the national names that you've been covering?
01:16Yeah, so I think MetaX is a name that's actually being consistently highlighted to me by investors.
01:20I know people are very bullish on this company and the prospects for it.
01:24But as Min Min said there, it still is one that's loss-making,
01:27even though it's actually projecting a faster path to break even than you see from MoreThreads.
01:32But in terms of how much of the market it's capturing right now,
01:35I think it's looking at around 1% of China's chip needs.
01:38So it really is just a minnow.
01:40But at the same time, the technology that it's pursuing is certainly very targeted
01:44because unlike MoreThreads that's sort of focused on consumer gaming GPUs
01:49and fully functional GPUs, MetaX is really pursuing high-performance AI training
01:56and data centre workloads.
01:57So they've got chips that are very much specialised for this purpose.
02:00So again, I think that just speaks to why there is so much interest in this stock,
02:04particularly when you had the reporting earlier this week or last week from us at Bloomberg,
02:09that scoop that China's looking at directing as much as $70 billion into the ecosystem
02:14to more fully build it out and focus on that self-reliance.
02:19Yeah, and a key question is whether, of course, China is approving
02:23or at least allowing these Chinese companies to buy up NVIDIA H200 chips, right,
02:27which President Trump gave the green light to just a few weeks ago as well.
02:30How is that going to all fare, you think, for MetaX?
02:33Yeah, of course, it will have short-term impact on the company
02:36because if Chinese companies get the green light to buy from NVIDIA,
02:39then they will not need as much of these domestic homemade chips.
02:43But I think longer term, the theme of domestic substitution is going to stay, right,
02:47because I think sources have been saying that the Chinese government's thinking,
02:51even if they were to allow people to buy the H200 chips,
02:54they see that as just a temporary solution to buy time for local chip makers
02:58to catch up with foreign technology.
03:01Because one of the big bottlenecks right now that people are telling us in the industry
03:05is that Chinese chips are able to do inferencing work,
03:08but they still can't do training work,
03:10which requires, you know, ingesting large amounts of data
03:15and then finding the connection between them.
03:17So that's where companies like MetaX are still far from catching up with the likes of NVIDIA.
03:22The company did say that their latest chips are close to catching up with NVIDIA's H100 chips,
03:28but that is, you know, the second latest generation that NVIDIA has.
03:33It's one generation away from the latest Blackwell chips.
03:36Yeah, Bell, you know, talk a little bit more about that, right?
03:38There's so much AI euphoria.
03:40You're seeing that very, like, you know, today with MetaX,
03:42but does it in some way mask some of the long-term technological challenges
03:46that China AI still faces right now in terms of the manufacturing side?
03:51Yeah, well, manufacturing has been one of the key bottlenecks,
03:53and you've got two main foundries, there's SMIC, there's Hua Huang,
03:57but it's only SMIC that can produce those advanced nodes,
04:00and that is the big challenge, I guess, or one of the key bottlenecks,
04:03is just getting that manufacturing capability.
04:06I do think it's interesting, though,
04:07because we've heard from a range of different chip designers,
04:10the likes of Canbercon, for instance, Huawei,
04:13that are all talking about scaling up their production capabilities next year
04:17or having greater production output.
04:19So exactly how SMIC and Hua Huang plan to handle that,
04:23I think, is going to be really interesting to track.
04:25But, yeah, in terms of where MetaX sits in this,
04:29I mean, yes, they're catching up to Mimin's point.
04:32The H100, again, it's from that Honeywell series.
04:35You've got the Blackwell right now.
04:36You're going into Rubin architecture next year for NVIDIA,
04:39so clearly a bit of a ways behind still.
04:42But we have seen chip makers as well really kind of getting quite creative
04:46with the way that they look at their chip making technology.
04:48For instance, Huawei as a company has pursued supercomputing clusters,
04:53for instance, that, yes, might have a high energy need,
04:56but still can get high compute power as well
04:59to try and compete with what you get from the U.S.
05:04And Mimin, we're not quite done, right?
05:06I mean, we're talking about there's still a flurry of companies
05:09in the AI space that are looking to list for next year.
05:12What are we tracking so far for next year?
05:13So we have companies like Yangzi Memory Technology, Changsin Memory Technologies.
05:18These are two that could be seeking an IPO potentially next year
05:22on the star market as well.
05:23And it really points to the success of the Chinese government policy so far
05:27because they have loosened the listing rule
05:29to allow unprofitable companies to list on the star market.
05:33And that's why more threats and MetaX are able to list.
05:37And as Bell mentioned, the high electricity need for companies like Huawei as well,
05:43many provinces are now providing subsidies,
05:46lowering electricity costs to support the build-out of data centers in China.
05:49So, yes, many industry watchers are saying that this policy
05:54from the Chinese government really is supporting the sector.
05:57And that's why we're seeing so much interest on an index like the Chinex, right?
06:01There's been a lot of concern around AI valuation being
06:05or overexpended sort of CapEx spending among the top-tier companies.
06:10But many people are now looking at some of the second-tier companies
06:12that are listed on the Chinex or the star market.
06:15Bell, next year is going to be quite busy for you too, I'm guessing.
06:18Tell us what's on your radar in terms of some of these AI plays next year.
06:23Yeah, well, exactly to Min Min's point.
06:24I mean, you've got a host of different listings
06:27that we're looking at in the China chip space.
06:29I think also one that we really should mention as well
06:31and one that I've already just been in touch with even as late as today is Baidu
06:35because there are reports that they're looking at listing off their Kunlun Shin division,
06:39which is the maker of the Kunlun chip.
06:41So this is, again, you know, chips that are being used to power data centers.
06:46This is a company or business unit that's been valued at as much as $3 billion.
06:50So just another name we're going to be tracking very closely next year.
Comments