China Goes All In on AI Chips, Anthropic Self-Corrects and AI Gets Real

Hey, happy Monday, everyone. Welcome to Techstrong gang. I'm Alan Schimmel, and I'm hosting today, sans our usual Chief Content Officer, Mike Vizard, who's out in lovely Minneapolis at the Linux Foundation's Open Source Summit, reporting on the state of open source.

So we should have a lot of articles and videos from Mike. While he's not here, though, let me tell you who is here. First of all, we've got some folks I haven't seen in a while, and it's good to see their faces.

First is my friend, Evgeni Karam. Evgeni, good to see you. I hope all is well.

Same here. Yep. Good.

Also, I haven't seen her, and she reminded me, in two whole weeks, Garima Bapav, who has been on a Tech Field Day field trip, which we're going to hear about. Speaking of Tech Field Day, Mr. Tech Field Day himself, Stephen Foskett.

Stephen, good to see you. And reporting to us from Amsterdam, where he's fighting through bandwidth issues, darkness, plagues, and everything else. But nevertheless, like the postman, he shows up, Jon Swartz.

Hey, Jon, good to see you. Good to see you. Hopefully, I'll make it through and not drive Taylor, our producer, crazy.

No, it's okay. They're already crazy. Anyway, as usual, it was a busy week last week.

It was a bit of a crazy weekend this weekend, as the news and the information just doesn't stop. I wanted to start off today, though, Jon, with an article that you had put up. Our glorious leader went to China with a bevy of billionaires to make deals, trying to turn a diplomatic trip into a business trip.

And that's all I'll say about that. But one of those leaders was Jensen Huang of NVIDIA, who, I imagine, was there trying to sell NVIDIA chips. And turns out, the Chinese, maybe he hit a great wall there.

What's the story, Jon? He stole my line. Yes.

So Trump went on this trip, diplomatic as well as business trip, with these tech leaders, and he did run into a great wall of sovereign economic nationalism. He picked up Jensen, by the way, in Alaska. Jensen was not going to go, but he went at the last minute, along with people like Elon Musk and Tim Cook, and nothing really much was accomplished- That's Tim Apple to you, Jon.

I know. His name's becoming dirt. The more he goes on these trips and shows up at these events, but that's another topic.

You talking about the dirt on his knee pads or his name? Probably. But nothing really much- Vizard's not here.

We could go wild, but go ahead Oh, wow. This is incredible. Oh my God.

Okay. So, the US went there, Trump went there to make deals. He wanted to become President Xi's friend.

None of this really happened. There were no semiconductor deals, global finance AI deals, progress in the war in Iran. NVIDIA really wanted something to happen because the Trump administration approved sales of the H200 to China in December, but since then, not one unit has been sold, according to several reports.

Not one unit has been sold to Chinese firms. So what's happening instead is that Beijing has actually been steering their domestic buyers toward local alternatives. So they have DeepSeek, for instance, is working, has announced its latest models are optimized for Huawei chips, and this kind of just shows that this US tech policy has inadvertently helped accelerate, I think, Beijing's drive for self-sufficiency and kind of locked out US customers.

The whole thing, to me, was kind of almost like a shambolic summit, and it really raised questions, I think we all do, about how NVIDIA's going to thrive in China and how any US company is, given the circumstances of what happened. And I don't think there's going to be any type of improvement in the coming months, if not a year. I got the perfect promotion.

They should include one of his golden tickets with each H200 CPU. I don't really ignite the- I thought you were going to say the Trump phone, which I believe is going to start shipping soon, evidently. Well, that's already made it.

Start shipping? They shut it down already. Oh, no, they announced that some units are going to start coming out, yeah.

It looks like a Walmart knockoff, actually, is what it is, but that's Openly, I announced the phone as well. Elon Musk announced the phone a couple of months ago. Yeah.

So I have some thoughts on this, but before I jump in, let me-- Stephen, I see you biting your tongue. Go ahead. Well, they weren't there very long, and it looks like this was a pretty, let's say, ad hoc trip, which is no surprise for a very ad hoc administration.

Mm-hmm. As far as I can tell, the only thing it accomplished is Trump didn't trip over the lintel when walking in the garden with Xi. Elon Musk made an awkward selfie with Lei Jun of Xiaomi, who he apparently didn't even recognize, even though that's his major global competitor in the automotive spaceAnd nothing else happened despite having all these people with them.

Well, nothing else we know of. Let's pray for Taiwan. Yeah, really.

Well, that was mentioned. It is interesting that the whole group went there with all these business people, all these major American business people. Maybe something's going to come of it.

Let's be optimistic. Maybe the meetings and dinners and so on they had with their Chinese counterparts would result in something. But like I said, things like the selfie moment with Elon Musk doesn't bode well because it looks like they weren't well-prepared, it looks like they didn't have much of an agenda, and that is unlikely to lead to much outcome.

So I'm going to try to take the politics of it off the table. But Steven, I agree with you 100%. I doubt they had an agenda, but this is who this president is most comfortable with surrounding himself with, the tech bro billionaire class and the oligarchs, if you will.

But here's the thing. There is a lesson here, and the lesson is taught to us by the Chinese. And this is a lesson they learned in the Opium Wars.

The best way to get off opium, right, or any drug for that matter, is to make your own market, have your own market for it. Rather than continuing to stay on Nvidia and having Nvidia as that choke point in their supply chain, they bit the bullet and said, "Okay, maybe we'll make do with less functionality for a period of time, but we'll be better for it in the long run because our AI will run on AI chips that we control, that cannot have the rug pulled out from under us, that can't go back and forth with tariffs or anything else. " At some level, I wish we would take that same medicine in making chips here in the US, and other technology.

Not just chips, but other technologies as well, where we're dependent on China or some other entity for our vital resource there, right? I think at some level you got to give the Chinese credit for doing that. They haven't bought one of those H200s.

Yeah. There's- An interesting aspect of that, to your point there, Alan, is that it's a long game. Trade is a long game.

International global relations is a long game. Business development is a long game, whether it's international or domestic or local. And you cannot do that in the midst of chaos.

It requires consistency and predictability. Yeah. I also have to add a few things here.

We have seen from the past, right? This looks like to me less of a interim sales hiccup kind of issue than a long-term strategy gap, right? So China is no longer a growth market.

And we need to understand that it's not also about only chips, because people have started to look at the AI stack, and that's how you can break the manufacturing bubble, right? So it's not about who and how and where the chips are manufactured. It's about how tightly integrated you are in the AI stack, right?

What kind of compelling solutions you will bring, which are taking the applications which are memory heavy, LLMs, for example. These applications will have bandwidth requirements, which the Nvidia chip, the H200 would fulfill, right? So it's more of an AI stack approach, which we need to follow through rather than just selling the chips.

Chips is one thing, and probably that will always be a question that where is it manufactured? One thing they talked about was this US-China board of trade that US is trying to push to include to actually reduce volatility for AI infrastructure vendors. That's kind of an interesting concept.

And to Steven's point, there is going to be another meeting of sorts, I think in a few months, in the fall. And then following that, there's going to be, I think the G7 summit is in Florida. I think it's in Mar-a-Lago of all places.

So there are going to be a couple more meetings after this. But again, as I think Steven and Alan pointed out, you've got two distinctly different approaches. You've got the chaos approach where you just hope something happens based on no research or any preparation, versus the Chinese approach, which is very thoughtful long-term and very strategic.

So there's no surprise what happens. Yeah. I guess how close we are to have a chip on our phone that can do offline LLM without going there, because Samsung claimed something already, and I think Apple will eventually release something as well, that you don't need to upload anything somewhere, and you can do your own LLM on the mobile phone, for example.

Privacy concern, where the data is going, is also interesting to see. And you very well might, and I do think that's Apple's long game. I think we know the answer, at least for the foreseeable future.

And we're talking about Taiwan, obviously, and whether that remains independent or not, I think you don't know what deals were made that weren't announced there. Yeah. The one issue that President Xi actually was most forthright about was Taiwan.

Yeah. And I don't think, again, the US diplomatic side, so to speak, had absolutely no idea how to respond to that. So- Well, no, I think unfortunately, we're in an administration where everything is for sale.

We're in a situation where everything is for sale. Or being bought. After that US approved sales of the H200 to China, you know that Trump reportedly bought more than a million shares of Nvidia stock, or someone representing him.

So, not a surprise, but that's kind of what we're dealing with. Let's put a happy face on this. Right.

Right? Competition brings out the best- Yeah ... usually.

And so maybe having a healthy alternative to an Nvidia, whether it be in China or here at AMD or other chip manufacturers, will spur more or faster, better solutions to market, whether we're talking about just the chip or as greener, as you say, about the entire stack. Right? Because I don't think China's going to be satisfied with just chips either.

Those chips are going to run on Huawei machines, and they'll be in Tencent or Baidu or whatever, data centers. And what's interesting is, China does their thing, the US does their thing, and a lot of this plays out in Europe, which will in some level pick the best solution, I think. It reminds me of the telco revolution, which happened, right?

Going back in time, you can see what happened with telcos, right? And this big Chinese vendor, which was there in existence with the kind of cost to the business, they were everywhere, right? And then what happened to them?

So it's the same kind of story repeats itself. I feel that a lot of these things will be triggered by how well we do software and hardware integration and the research into that space, because hardware alone will not dictate the market. No.

But an interesting thing, almost every hardware manufacturer I've ever encountered claimed to be a software company. Could be now. That's Nvidia.

Including Nvidia. Nvidia, yeah, as an exhibit A. Yeah.

But important word is claimed to be. Yeah. Well, show me the money, as they always say.

Anyway, we've got to move on. I think as you guys have rightly pointed out, this was a very short summit. It was short on agenda, short on what we're seeing come out of it, but we'll have to wait and see what the final outcome is there.

I wanted to move over, though, to an announcement Anthropic had about. " When I first saw the Dream On, I had visions of Aerosmith. But you got to give it to Anthropic, they seem to be capturing the imagination of the market.

They seem to be always just a beat or a half a beat ahead of the rest. But Evgeny, you want to jump on this, kick us off on this one about this new- Yeah, happy to do it. Sure.

It's an interesting feature. First of all, they're releasing feature like every day. I think you need to have a full-time job to understand what they're releasing.

And I decided that I'm going to just stop trying to read everything, because every time they release, people trying to teach you how to use it. And if you wait a week, they will do it by themself, or they fixed the problem before. But this particular feature, the dreaming feature, is very interesting from my perspective because everybody that you work a bit with Anthropic or even the other LLM, you see the problem after some time, because they forget the context, they forget what you talk, they kind of assume something, and at some point, you assume they are aware, different results.

So with the dreaming feature, the Anthropic is supposed to go back to the conversation during the day or even during the week, and remember and reconcile everything that happened. And this supposed to fix the hallucination more, the context of what's happening, and let you work without having so many problems. And I think one point that's very important, especially for the new people that working with Anthropic or other LLMs, we all got excited in the beginning, and then with time, we kind of phase away because we don't get the results we got in the beginning.

And even if you're trying to build your own app, the first version is amazing because you build something in half an hour. It looks amazing. And then you try to fix everything and add, add, and add, and it's not so glamorous as it was in the beginning.

So by adding the dreaming feature, I hope the idea will be that they're not going to lose context, they're not going to start messing up with what they're doing, and it's going to be a better view on what's going on. The part I am not sure, because sometimes you have question that you ask Anthropic, that it is pointed. You want to check the price of this, or somebody asks you something, or you're going on a street and you saw a sign or building and you want to check what it is.

You don't care about this anymore. You're done. So would this also have a lot of baggage and a history about it that nobody's going to use?

And how exactly we need to know that it's dreaming about the right things. Well, you mentioned that. I'm reminded of, I think it was the second Arthur Clarke movie on "2001," "2010," where HAL asks the computer scientist will he dream, right?

And now we have AI agentic dreaming. But here's how I look at this. I think we are running smack dab into a wall of what AI really costs, right?

Originally, it was, could it do this? And now I think we're starting to see, well, at what cost could it do-- It could do it, but what's it going to cost? Mm-hmm.

And without some sort of efficiencies being built in, without something giving in this model, it's going to severely hinder the uptake of AI because as even, I read an executive from NVIDIA said this, right? If it's cheaper for the human to do the task, use the human. Yeah.

Oh, go ahead, Mark. I'm sorry. No, no.

So, is this part of that, John? Is this maybe the beginnings of that? So I was thinking about the cost element, and again, this made me think about Philip K.

D**k, right? In "Blade Runner," a replicant, and dreaming and thinking. But given the scarcity of computing, if compute scarcity is driving product design, I'm wondering how the pricing and the governance evolve for agents that run for hours or days at a time.

That's the economic part that I'm sure Anthropic is thinking through, but that's one thing that stuck in my mind as well, is how this shifts even more so the competitive landscape between Anthropic and OpenAI. One thing that Dario Amodei, the CEO of Anthropic, said during the announcement of this new feature, he pointed out something that just left me a little bit startled. He said the company, and I want to throw this out, this is on the business side, they're innovating at such a good pace and such a smart pace.

They cited an 80 times annualized increase in revenue and usage in Q1 this year versus what was expected to be a 10 times growth. It's pretty phenomenal, and I think it's part of what's happening with these types of features and the types of things they're doing to push the ball forward. I think they want to catch the agentic AI wave, and this is again, a feature which is meant for agents and orchestrated agents because- Mm-hmm ...

imagine if an agent gracefully degrades or fails, it has a better kind of possibility to orchestrate behaviors which are more production ecosystem friendly, right? So that's point number one. Second point which I am thinking is that it is moving people like me away from, why should I build my own?

Because this is not only one agent or orchestrated agents working together, it's also a systems problem. And we will talk about this in the AI Field Day section as well, that we need behavioral logs, right? We need to assess those logs.

We need to have some kind of assessment of the feedback loop and the self-modification which is happening. And the more you do it by yourself, the cognitive load is gradually increasing, right? So you need to have specialists or the companies who can do this kind of system integration of these kind of features, especially for your orchestrated workflows.

So I am thinking of more of why not to go to build my own orchestrated workflows rather than buying it from a specialized organization which can make use of these in-demand features. And hopefully it will not orchestrate our agents which are compute hungry or they're architected well. So we need to also be mindful of the cost.

Yeah, and I think the cost is only just starting to come due. So to Garima's point, this is something we spoke about a lot last week and as well to this article. Every one of these features, whether it is the sort of internal monologues that are happening now with a lot of the advanced frontier models or this dreaming feature or some of the other features they've added here, I'll call out another one that was mentioned in this article is essentially a keep working on it until you achieve the required level of output.

All of these features eat up tokens like crazy. And then as Allen was pointing out, at the same time, what we're seeing is the same companies, literally the same companies, switching from flat price models to per token models because frankly, people are way, way overusing... the assumed token levels.

Well, it's no surprise. 3 to 3X as many tokens as previous models did for basically the same request. And these features, like this dreaming feature and so on, use even more tokens.

And not only that, but they're going to autonomously use those tokens. So it's not like it's sitting there waiting for you to do something. It's going to do this constantly on its own, and it's going to charge you for that.

And so I think that all of this is part of a trend among these companies to figure out how to make these models more profitable. You talked about 80X revenue growth. Well, that's interesting.

Yeah. I'd like to think about that. But that money's coming from someone, and it's probably coming from you.

Yeah. The unpredictability of the cost would raise alarms, right? If you're the customer.

You don't have a fixed idea of how much this is going to cost if the autonomous agents are racking up the bill. But this is, I think, the normal evolution of technology. Granted, this is as big a technological revolution as I've seen.

But initially, it's always just make it work. Then it's make it work efficiently or make the economics of it work. I think you skipped a couple steps in there.

Yeah. I went right to the bottom line, as they say. But make no mistake, that's where we are now, the bottom line.

We saw it with cloud. Remember the whole FinOps thing, right? Just because you can, doesn't mean you should.

Well, there's another way to look at this, Alan, the business model for growth, because I think you're right. I think you're exactly right on this. With technology, the first thing that happens is can we do it?

Right. But when it comes to the business models coming out of a lot of these VC-funded startups, especially these days, the next step is put it out there for free, no matter what it costs us to deliver. Then the next step is try to get subscriptions and recurring revenue going so that we can show user-based growth and paid user-based growth.

And then the next step is flip those subscription users to per-use users or bigger subscription users in order to lock them in as long-term customers and show recurring revenue. And then, and only then, do we think about making the whole stack profitable on a per-transaction basis. And so whether it's cloud or whether it's any of these other things that we've seen, the last thing they think about is making it profitable.

The problem here is that AI, just like the rest of the AI economic world, skews all of these things just so incredibly that the let people use it for free stage was burning up billions of dollars a month. The switch it to try to get people hooked on it stage is burning up tens of billions of dollars a month. And now they're going to try to flip to the let's make this thing profitable stage.

And I think that's going to be a real reckoning because unlike cloud, which was always profitable on a per-transaction basis, this has never been profitable, not even remotely profitable on a per-transaction basis. And once they turn that lever that says, "Chunk, make money," I think a lot of the people on the other end of that transaction are going to be like, "Whoa, there. " Well, this is why, and again, there were a series of reports or studies about that whole question about AI ROI and how it's not translating so far.

I know it's early, but I'm going to say, this is becoming more and more of an issue, and it's become more top of mind, this whole financial model. Yeah. One more thing which is different from what we have seen in the past technology curves.

We have SLA for a service which we buy, right? In this case, unfortunately, we don't have an SLA of any reliability or any outcome because of the non-deterministic nature of that technology. How long we can survive with no SLAs being handshaked is a question mark in my mind because we are talking about real technology differentiation.

People are making profitable businesses out of it, but what is the end customer getting? Are we compromising reliability? Are we undermining risks?

These are the questions which we need to answer. Look, maybe yes to all of those things, Garima, but I think for most end user companies, you've got CEOs and exec teams and boards sitting there saying, "How can we do more with less? " Right?

" I'll be diplomatic. " I put out an article this morning. I followed the company GitLab, I think from the day it was founded or came out of YC, Y CombinatorWhat a radical company.

I don't know how many of you are familiar with the GitLab story, but Garima, you are. You're a DevOps person, right? But all remote from day one.

They didn't even have an address to mail your bills to and stuff, right? There was no office, a post office box. All remote.

Literally an open book. It wasn't just the software that was open, the whole company was open. You could look up and see who everyone that worked there, who they reported to, in some cases, what salary bands were they in.

The entire roadmap was laid out, everything. It was totally open. And the founding team led by Sid Sijbrandij did an amazing job.

Sid had his own situation he dealt with, and he moved on. A lot of the original exec team moved on. They put out a thing last week, their new CEO, who came over from...

Well, at one point he was with New Relic. And they're calling it GitLab Act Two. I don't know if you saw this, Garima, but basically, without giving the exact numbers, they're going to be going through a massive reorg.

Major headcount losses. They're going to pull out of certain jurisdictions and leave it to partners. But more importantly, the thing is framed as positioning GitLab for an agentic AI future.

And look, I applaud the openness in coming out and saying it. I think a lot of companies are doing it without being quite as transparent. Mm-hmm.

But it's going to be interesting to see what this really means. Mm-hmm. Right, Steven, I'd love to have them at a Tech Field Day and talk about it, right?

Because it's going to be interesting to see what it really means and what the goal is and how they're getting there. And if it's successful, like the original GitLab, we'll see a lot of people copying or mimicking that kind of path. But they're out there blazing a new path, certainly, that's very different.

And in my article, I put in some former GitLabbers lamenting about the loss of that great culture that was there. I don't know if it necessarily means you got to lose the culture. But anyway, it's a decent article, and it's an interesting story to follow.

Speaking of interesting stories to follow, Steve and Garima, you both mentioned you were out last week, I guess in Santa Clara, right, Steven? Yeah, San Jose, close enough. San Jose, excuse me, for Tech Field Day.

Steven, why don't you kick it off, and Garima, feel free to jump in. Yeah, I'll give the facts, and Garima can give the perspective as someone who attended and joined us for the whole thing. First off, it was wonderful to have a bit of a Techstrong gang reunion.

Garima grabbed a selfie with me and Guy Currier and herself together out there, and I hope we can see more of the folks as well. So yeah, this was AI Field Day, which is our sort of gold standard of what Field Day can be right now, and it really proved itself out. I said in the introduction to the delegates that again and again, what happens at Field Day is there's always a dark horse company that you don't expect a lot from but comes in and wows you.

There's always, I guess, a front-runner company that you're maybe a little disappointed in. And then there's always a lot of lessons. And I actually advise the delegates whenever they come to Field Day to listen just as much to the presentations about the topics and technology as it is for the themes and what they're saying.

And certainly, this was a lot of agentic discussion, a lot of autonomous AI discussion, a lot of data, and how to integrate data with applications. We got some really fabulous discussions from some of the companies on how to do this. But I think the thing that really linked us all together, a few times ago, I remember saying here on the gang that AI is, we're trying to figure out how to make it real.

I think that that's really where we're at now. We're making it real, and we have people who are doing real, productive, useful, and yes, efficient and profitable applications. But at the same time, we're still living in a time of just this intense hype around things that frankly aren't real, whether that's super intelligence or agentic AI or let's say self-driving cars.

You know what I mean, Garima? So, it's one of those things where I think all of us are seeing the world change in front of us and at the same time trying to figure out how to make productive use of this. So Garima, let me turn it over to you.

You attended for the first time, I guess. What was your perspective on the event? Yeah, before I go to the AI Field Day, because I have a lot to say there, but I would like to appreciate Alan Shimmel, the Techstrong Group initiatives, and the Techstrong Group as a multiplier of innovation.

Because you could see that myself, Gaye, Kate, Stefan, we are all- That's right. Kate Scarcella's there, yeah. Exactly.

So we all come from the Techstrong connection, and then we led this connection into the AI Field Day. So thank you, Alan, for actually creating this kind of- My pleasure ... platform for all of us to kind of integrate.

Now, coming back to the AI Field Day, and my expectation was it's another conference, right? So I went with that notion into the AI Field Day. But I got to realize my dopamine was increasing every hour when I was attending that AI Field Day, just because how legends think.

Stefan, I would say that you have created an incredible space where real discussions happen. And my key takeaways now from AI Field Day, there were a lot of companies there, some very well-known, like Ciscos of the world, Hammerspaces of the world, thescalability of the world. And then there were not so known companies as well, which were incredibly strong in what they were doing.

My three key takeaways from this AI Field Day was, the first thing which I should articulate is that AI is a systems problem. AI is not living in isolation. AI has different facets.

Infrastructure. The AI stack comprises of infrastructure, it comprises of the hardware and not so known hardware components, which I actually came to know about. The AI on Edge, how it lives and how it brings value and how it should be system architected.

So that's another facet. And the software on top of it, of course. Needless to say, a lot of agentic discussions.

So of course, AI is a systems problem. Start to think about that. The second key takeaway for me is that it's not product-centric anymore, it's platform-centric.

So it's not one product, one hardware, one chip, one software vendor. So you'll have to agnostically look at the ecosystem and how you integrate with the AI stack in a meaningful way. Right?

And the third part is also very interesting, and this is my key takeaway from this, is that why not to build agentic AI workflows by ourselves? The build versus buy decision making, we'll have to think about it very clearly in our heads, because the cognitive load of creating your own stack is much more than what you think when you bring a partner. So we have seen in this AI Field Day how partners are building an ecosystem which we can bring, right?

Which will save us hundreds of hours. So start thinking about your 18 to 24 months of roadmap and who should you bring. The build versus buy decision, you'll have to be very selective in what you do yourself and how you can integrate into the ecosystem with partners.

Like Selector, for example, was an underdog. It came with a very pragmatic approach how to run network operations. Similarly, Cisco was a surprise package for me because they showed something called AI Canvas.

We in the telco space are craving for that kind of functionality. So I think I have seen a lot of these innovative solutions. But now over to you, Stefan, for more.

I can keep on talking about this because I'm very excited. Yeah, it is always very exciting. And again, to me, the most exciting thing, Garima, is when folks like you and Kate and Guy and, I don't know, Ryan Booth and some of these other folks who have completely different perspectives come together and sit together and roll around how can we approach these things.

One of the things that we did this time as well is I got a chance to sit down with Garima as well as some of the other folks to talk about their perspectives on topics generally in sort of an interview format. And it's always fun to hear the perspective that people are coming from. To have Kate sitting in there, and she is constantly thinking in terms of information security and what does this mean for information security, but also how can I bring my InfoSec perspective to what I'm hearing here?

And it was fun. There were so many instances where she used that perspective as the metaphor to help her to understand something that was completely unrelated to information security, and that helped us all to see it more clearly. And you mentioned some of the surprises.

I think the biggest surprise for most was to have Solidigm, of all people, they make SSDs, come in with an absolutely fascinating deep dive that was an anatomy of an AI query all the way through the prompt and the tokenization and the expansion of tokens and stuff. And to our previous discussion here as well, we could instantly see how a request that is maybe a half a dozen tokens becomes using thousands of tokens on the back end, and how that could be optimized or maybe can't be optimized. So I just can't wait to share these things.

We're going to be posting them actually to the Tech Field Day YouTube channel today, but you can actually see them already in the Tectron TV app. One more thing I think we should mention is the presentation from Guy, because he presented the Futuron Group research on where AI is moving and how the edge AI infrastructure space becomes extremely important, where Solidigm fits in, right? So I think there is some material available on the Futuron Group with that perspective.

Do watch out for that. Absolutely. I'm wondering, we spoke about chips before in the beginning.

We have Cisco making a lot of announcement. Juniper and HPE, I think, finished the merge, and I think HPE just announced a first AI switch as well during the same time. So where are we going with AI and switching and edge?

What is the mean AI switch? You know what? Cisco Live is in about two weeks, a little less.

Steven, I think you guys are doing a Tech Field Day there, aren't you? We are indeed. Yeah.

And we will be covering... Thank you, Evgeny. We will be covering exactly that in a few weeks here at Cisco Live.

Evgeny's speculation. I should know in an hour or two. I might be there with a video crew, Steven.

Oh, that'd be awesome. That'd be awesome. And yeah, and if any of our listeners are going to be at Cisco Live, do connect with Tom Hollingsworth and the crew.

Of course, we would be thrilled to have you join us as well for some of those sessions. Evgeny, anything else? No, I just wanted to see what are you guys thinking about?

What does it mean, AI switch? Look, I remember when the first cloud switches came out and with virtualization and all of these things, and it was heady times. Anyway, we're about out of time, though, today, and so that's going to wrap up today's Techstrong gang.

Big thank you to our gang members, Evgeny, Garima, John, and Steven. And Garima, to your point earlier, it's I who thank you for coming on here again to the gang because you all make this richer. We couldn't do it without you.

Otherwise, it's just Mike, I, and John talking to ourselves. Secondly, if you like what you saw today, do subscribe to our Techstrong TV YouTube channel or the Tech Field Day YouTube channel or our Techstrong OTT app that Steven mentioned. tv.

tv site and the OTT site now, and so hopefully it's a little bit more intuitive and gets you to the information you want quicker. We do gang live every weekday, live at noon, of course, Eastern Time, and you can watch it on any of those channels as well as any of our Techstrong sites, along, as we call it, the Techstrong network. But that's it for now.

Thank you for watching. I'm Alan Schimmel. Thank you.

Thank you for being part of the gang. We'll be here tomorrow.