'Coming for the King' in AI. OpenAI, DeepSeek & Samsung. ARD post 197

0:01 Welcome back AI curious folk to AI ramblings daily podcast. I'm Michael Parrick your human AI guide. We're at episode 197. It's Monday. Hope you all had a terrific weekend. Today's theme coming for the king. It's a theme from a popular TV show, The Wire. You come at the king, you best not miss. In this AI tech wave, the number twos are keep keeping coming for the king. I want to talk about three events today with my takes. Let's get started. First, Open EI is continuing to try and finish its penultimate private round raise ahead of Enthropic, its arch rival, finishing its mega AI IPO at a valuation 1.5 trillion to $2 trillion or above before Thanksgiving. OpenAI has been trying to raise about $30 billion or so at a $1.4 $4 trillion valuation in the private markets ahead of their planned AI IPO mega AI IPO sometime early next year.

1:14 But the latest reports from the Wall Street Journal and the FD indicate that there may be some adjustments that the they may be looking at raising about 10 11 billion and maybe a valuation of one and a4 trillion dollars. So we'll see how that sorts out. It's still fluid. There's a group of investors from the UAE expected to lead. Big venture firms are in it. Soft bank is in it. And then also family offices around the world, individuals, etc. So, it's a little tougher. It's a tougher environment.

1:47 Obviously, we're on the eve of the midterms on Tuesday. Data centers are a big picture in in this in this whole midterms election on a bipartisan basis. I've written a lot about that and that's in the show notes. And both parties are basically taking a pause ahead of the midterms. We'll see how it all sorts out later in the week. But investors are concerned that even though the capital seems to be there for all these massive data commit data center commitments, especially by the two top companies that there may be delays because of regulations, permitting all the resistance in so many states, blue or red, against data centers. and we'll have more to say on that as well. In the meantime, everyone's going through Enthropics S1. There are going to be updates to that. So, OpenAI has a bit of a moving target in terms of how they finish their private raise ahead of their IPO. My take on all this is this will continue to be a bit of a fluid situation pretty much until this round is closed. So, it's not unexpected.

2:54 And look, the 1011 billion is still a huge amount of money. If you compare that to prior tech waves and other IPOs and private raises and so on, we're in extraordinary land in terms of raising huge amounts of funds is that's just that the mega AI IPOs, especially Elon, SpaceX earlier this June and now of course with Anthropic, we're kind of spoiled by these massive number of zeros. So we'll let's just take that a little bit with a lump of salt if we need to. Let's move to the second event today is Deepseek in China is taking a shot at anthropics leading position in AI coding and they're releasing a more autonomous coding agent. It's aimed squarely at Anthropics Claude Code which is the market leader. Of course the number two is OpenAI with Codeex.

3:49 There's more details in a Bloomberg piece which I'll have in the show notes. Deepseek's product is called Super Agent. It's open source for now and it's a prototype. It's watches it's it watches the developers group chats. It spots coding problems, suggest fixes without being asked. So there's a proactive element to this. And remember the anthropics and the opens of the world are not sitting still. They're continuing to improve their models as well. So this remains a competitive space and we're going to see more activity not just deepseek but others like Alibaba in China and of course OpenAI and other companies in the US. My take on this part, China is taking a solid punch. They're coming in against the king, going for the king as it were.

4:36 We'll see how it lands, but the moves are being made and we're going to see more of it, not less. And the it's not again, as I said, just going to be Deepseek, but it's going to be a whole bunch of other companies in China and in the United States. And Alibaba is on track for a potential supercale model of 5 to 10 trillion parameters. We'll see when that lands, but that's another piece to going to watch for in the coming weeks as Enthropic proceeds with its IPO before Thanksgiving. The third one I want to talk about today is Samsung. It's both companies Samsung and HSKNX. SKHX of Korea being the leading high bandwidth memory company and Samsung being the number two. Number three of course is Micron here in the United States. They they've been again continuing to the to be volatile.

5:29 Samsung shares slid as much as 7% on Thursday after it held back its new plan to reward shareholders. There were details which I talked about a couple of days ago on SKH Heinix. Their operating profit in their quarter had jumped some sevenfold and they committed over $45 billion more to shareholders. So about 75 billion total this year. SKH is sold out on its HBM and other memory through 2029. Samsung says they're sold out through 2027.

6:03 And my take here is that investors were not quite as impressed with Samsung's attempt to please shareholders with their very good results as they were with SK Heinik. So Samsung took a took a swing at the at the king and and missed a little bit. We'll see how that shakes out as well. My overall take coming for the king. You better not miss. And we have three examples. Again not as dramatic as going for the king for the jugular as well but we are seeing openi continuing to absolutely try their best against anthropic deepseek against all the US companies especially for anthropic and Samsung after the memory companies in particular skinex which remains the global re leader but remember there are also companies coming from behind CXM which I've written about in China and yang and and and and and lot more competition on mainland China as well and those are number four number five as well plus a whole range of of memory startups in AI that are emerging around the world. So it's there are a lot of lot more people that'll be coming for the king. Let's move to the gadget AI section today. Meta is apparently focused on humanoid robots with help from AI startups in China. Remember I've written a lot about China being the leading company in humano humanoid robots with their companies like Aubot and Unitry etc. The big one that gets the most visibility and the highest valuation in the United States of course is Elon Musk with Tesla and his Optimus humanoid robots where he's promising millions of robots in a few years which again I've written a lot about. I think those are relatively ambitious aspirations, but robots are a very big emerging area in the world of physical AI. I've talked a lot about Nvidia being in this space as well with their cluster of hardware and software. And so this is a going to be a rich area and Meta will likely be experimenting with their humanoid robots not only in terms of with technology from China but also potentially suppliers like Nvidia and others. And so here is Meta which is so aggressively focused on taking a swing in models against the anthropics and open AIs with the huge amount of commitments 150 billion or more in data centers but they're now also making sure they're focused on the physical AI side of the equation with humanoid robots. So we'll see this segment as it evolves but we do have a new player with extraordinary amount of resources. Let's move to the questions for the day today. Question one, who remains the global leader in humanoid robots for now. My my view, it's still the Chinese companies. They have the deepest broadest ecosystems in terms of all the components that are required for humanoid robots. And so the companies that I've talked about in earlier pieces, Adapot, Unitry, they're the ones that are front and center. The one that gets the most attention amongst institutional investors, especially here in the US and Europe, is of course Elon Musk with his flare for dramatic demonstrations, investor days, etc.

9:36 around his Optimus robots. And I've written a lot about that as well. But that notwithstanding, I think China remains still front and center because of their manufacturing capabilities. Question two, when do I view humanoid robots as being a viable, real, almost ready mainstream global market? from my perspective again I've written a lot about this but everything I see even in the latest updates I'm convinced that we're not going to see billions of humanoid robots and for another four or five years at least so 2031 or later and that's still me being a little bit on more on the optimistic side there's just a too much AI hardware software science data issues robotic problems and use cases to be figured out especially on the consumer side of the equation and we are nowhere in the the at even figuring out what the possible product market fit is for robots. So we'll see how it evolves but it's a long time coming but in the meantime all these companies now including Meta are going to be experimenting with this in this area. Today's AI reset to zero is number 1258. It's a part two on anthropics S1. I had a discussion with my friend Ram Alwalia at Lumida on the entropic S1 and I wanted to write up that account with some of my takes and also Ram's takes which is in part two of this discussion in 1258. So do take a look at that if you if you are interested and tomorrow we'll be back with ARD 198 and reset to zero 1259. I know tomorrow is election day so do go out and vote. Thanks for joining us today, AI curious folk.