Enterprise AI, Sovereign AI and MWC Barcelona Trends

This week, we consider the announcements from MWC in Barcelona and pre-GTC announcements from around the enterprise AI industry. This episode of Utilizing AI features Nick Patience, Brad Shimmin, and myself, Steven Foskett from the Futurum Group, discussing this globe-spanning conference no longer just focused on telecom. We also talk about NeoClouds following the incredible two billion dollar funding round secured by nScale, perhaps Europe's largest funding round ever.

We then turn to the current impracticality of local models for most users and finish with what we're looking forward to at GTC. Welcome to Utilizing AI, the podcast focused on practical applications of artificial intelligence from the Futurum Group. Every Wednesday, we explore news and use cases of the ways in which AI is transforming enterprise IT and the industries it serves.

I'm your host, Steven Foskett, President of the Tech Field Day business unit here at Futurum, and before we dive into this discussion, let's meet who's on the panel today. Brad? Hi, everyone.

Brad Shimmin. I am the VP Practice Lead for Data Intelligence Analytics and Infrastructure at Futurum, and, as you imagine, I like data, I like AI. Hi, I'm Nick Patience.

I'm the AI Platforms Practice Lead, so I get to focus on the, the things that are unique about AI. Absolutely. And of course, we've got a bunch of other great folks that you've seen here on the podcast and hopefully, some more folks from Futurum will be joining us in the next c- in the, in the coming weeks.

Uh, note to listeners, I know that GTC is going on this week as you're hearing this, but we are actually recording this, the previous week, so it, uh... we will be covering that in the future, but everything we're saying here is not yet defined by the earth-shaking announcements that you are probably, maybe, could be hearing, this week. Instead, what we're going to do here is we're gonna talk about some of the news as we know it, the week before GTC, and that is going to be, I guess, headlined by what came out of Mobile World Congress, MWC.

Now, one of the ways that AI is affecting all of us is that it is, I was gonna say infecting, but it is appearing everywhere. Um, and Mobile World Congress is unique because it is both an enterprise conference as well as almost in a consumer edge, environment, 'cause you do see a lot of, of things that we don't always cover here. Um, Nick, let me throw it to you first.

Um, talk to me a little bit about what AI-related tech did you see at Mobile World Congress, and what did you think of it? Yeah, Steven, you're right. It is, it is a kind of combination.

It's one of these shows throughout the year that's, I mean, not quite on the scale of GTC, but in terms of that has become a kind of you-need-to-go-to shows. Um, whether you're, you know, the, the mobile, name and the word of the beginning is, is now slightly misleading. It's becoming a bit more kind of general purpose.

So yeah, I was... I'm not a particularly, you know, I don't get particularly excited about phones, so I wasn't, like, going around, figuring out who's got the latest, latest models and, and things like that. Um, I was there focused on, on AI as ever, and I think there's a few things that, that jumped out at me.

Um, so sovereignty, so sovereign, sovereign AI is a, a key part of my, is my, my coverage, and I think there was, there was a lot of folks on that. There was a big announcement from Telefonica on, on that, and there was, various panels. I went to one with, Telefonica CEO, D-Deutsche Telekom CEO, and a couple of others.

And it was kind of, it was kind of like European sovereignty in, in a microcosm. There was a lot of, "Why can't we all just get along and do things? " type message.

So it's quite similar. I hadn't been there f- I hadn't been to MWC for about a dozen years, and last time I went was probably, you know, the, you know, 5G is around on the, on the horizon. Now 6G is on the horizon, and there was still kind of bickering about the same kind of things, but also promising pretty much the same things, that 5G was gonna, you know, 5G was gonna change the nature of operators, and it was, yeah, it's not just about phones.

It can do all sorts of things. Do you remember the Internet of Things was supposed to be based on 5G? Um, what happened to that, you may ask.

Um, and now 6G's coming along, and it's kind of sound-sounding pretty similar. Now, that's not to say that the AI has not moved along massively since then. Obviously, obviously, it has.

We all know that. But so I think there's, there's a, there's a lot of opportunities for, European telcos, to, to some extent, I don't wanna overplay this, take on the US hyperscalers. They certainly s-see themselves in that role.

Um, and then that's where obviously the, the, the chip companies come in, like, like NVIDIA, the equipment makers like Nokia and Ericsson, and all the software layers, are all kind of like, you know, hump, you know, buzzing around this trying to, you know, see where to, to, to place their bets. So I think there's, you know, there was a lot of cool things there. There was, some of the stands there were just out of this world in terms of, you know, the, the production values.

Uh, it's not like any, I don't... It's not like any other s- show I've ever really been to in terms of that and the amount of money that is l-lavished on this. Um, and there were, you know, huge presence from China.

Um, all the k- all the kind of Chinese hardware companies of all k-kind of stripes you can imagine were there. Um, but also, you know, you know, the likes of, you know, Qualcomm and, and, Nokia and Ericsson and, and others had, had, had big shows, and then every country basically is represented. You go, you know, there's the, there's the British stand, there's the Estonian stand, there's the Argentinian stand.

Um, and so, so it's, it's a, it's an interesting mix of, core technology, interesting AI stuff, which is quite cutting edge, telco stuff, which sometimes doesn't feel quite as cutting edge, and then geopolitics, all wrap- all wrapped in. So it was, yeah, it was quite, it was quite an, quite an experience. Yeah.

Actually, I stand corrected. Um, I just looked it up, and indeed, it's just MWC now. It's not Mobile World Congress.

Uh, reminds me of RSA, the former- Yeah ... RSA Conference is now just RSAC. Uh, I guess GTC is just GTC.

Yeah. Uh, we're, we're, we're post-modern now. Things don't stand for other things.

Uh, but you're right, it has always been such a cool conference. Because of the telecom angle, it does always bring in a lot of international flavor, and that is, I'm sure, a refreshing, a g- g- gust of fresh air to people who are tired of hearing this, sort of Silicon Valley focused, narrative around all this stuff. Brad, talk- Mm ...

what, what, what is your impression? Well, fir- first and foremost, yeah, it is a, a sight to behold, this conference, and any, any conference that has people movers built in has to be, has to be big. Uh, it's, it's like its own airport if you will.

But, yeah, I think, you know, a- as you said, Steven and Nick both, telecom plays a big role here, and it is, you know, so a lot of people, huh, boring compared to, you know, shiny new toys. But it is, I think, also very heartening to see these vendors, looking to extend their role. I mean, yes, th- th- a lot of the news coming out of the show from the telecom providers was the shocking realization that AI might help them to optimize, automate, and, and streamline their, their networks to, you know, increase value and performance.

That's their job. But, they, they seem to be quite aware of, of the fact that innovation happens everywhere, and so much so, I think it, it was, you know, shocking to me a little bit to see just how much that, that's played out on the mobile device part of this spectrum, end of the spectrum, because what was it? There was, uh...

Trying to think of the name of it now. There was a phone that actually had a little pop-up camera on it that would... A, a robot camera, they, they actually called it.

It was Honors, Honors robot phone. That's right. Yeah.

And it would actually dance for you, and follow you around the room as, I guess, you, you, practiced your TikTok video dance sessions, whatever it's built for. But my goodness, I, I think the funk, the strangeness, is, was certainly on display at the device end with AI really serving to drive a lot of that innovation. Yeah.

I actually was in one of the kind of mini keynote sessions where the Honor CEO, was, was, was showing that one. It was, some of it, some of it felt, like, incredibly innovative there. Some of it also felt like, remember when we used to add cameras to phones?

Um- ... you know, it, it kind of looked a bit like that, except obviously- Yeah ... it moves and it follows you, follows you around, and things like that.

A couple other things I think stood out, I guess, there was y- when I talked earlier about the, the telcos positioning themselves as AI infrastructure providers, I, I, I don't wanna kind of, you know, knock it as, as kind of irrelevant, 'cause I don't think it is. It's, it's... So it's not only the telcos, but it's obviously all their suppliers.

So, you know, obviously NVIDIA has been... is, is pushing the telcos in that direction. We had a, I had a really interesting show floor tour with the, the, the person who runs, she's sort of GM for marketing for, for telcos for, for NVIDIA, and we were going around, you know, Nokia, Ericsson, the big, the big kind of companies, but also some of these much smaller, providers that are taking, like, DGX Spark boxes and building base stations for very little money.

I mean, we're not talking six-figure amounts here. We're talking very little money, and it kind of, it almost kind of reminds me, you know, not being too tasteless about it, it's kind of drones versus, you know, very expensive missiles, and it's kind of a, a similar thing. They're, they're very disruptive are those DGX Spark boxes from NVIDIA, which are, are, are quite disruptive in that, in that world.

Um, and I think, you know... I mean, Cisco was talking about inferencing grids, you know, in, in some of the kind of... I went to a kind of private session with the CEO of, of Cisco and, and, and a few of the executives there, which can't go into too much detail, but they, yeah, they're talk- they're talking about how important their stack is.

Um, and NVIDIA, sorry, Nokia framed this as a kind of AI super cycle, where, you know, obviously the data is much more bursty in, in, in, in, telecommunications and, and then in, obviously in Agentic it will be much more bursty as well, so they, they think they, they have a, they have a play there. And then finally, I think, we hear a lot from Jensen Huang of, of NVIDIA and others about physical AI, and this show, I think, that's the direction it will, you know, partly go in. Obviously, there will always be the consumer play around the phones.

Um, but I think, you know, this is, this is again where, where Nokia was talking about the traffic differences, and, and, and things like that. Um, NVIDIA had kind of partnerships with SK Telecom of, of, of South Korea, BT, NTT Docomo, and this is where obviously 6G, and going back to my point about 6G and the great promise of it, it's I guess for IoT with 5G, you think physical AI, with 6G. I think that's, that's the pitch.

Yeah, you, you know, like you said, Jensen has said many times over the last year that physical AI, the physicality of AI is the future, and certainly you can see that reflected in how companies are building out automation for their supply chain as well as factories, et cetera. But it, it's interesting to see, you know, when you walk around MWC, just how many robots there are or were, you know, at the show on display, and none more interesting to me, by the way, than the DinoBot, which was a, a velociraptor. Um, cheap price tag, only €24,000, to pick this up.

But, you could have your own, you know, auto- au- autonomous velociraptor to follow you around and, and, scare off the neighborhood dogs if, if needed. Is that the key use case, you think? I think so.

Or, or to play parties and bar mitzvahs, bat, bat mitzvahs, I guess. It would be really good for that. Cool.

Yeah. So it's... So we, should we switch gears, Brad?

Actually, before we do, I wanna- Mm-hmm ... I wanna just chat briefly about Lenovo in particular, which had, as, you know, for those of you listening to this podcast, before we came on the air, we were talking a little bit about booth space at the show and just how expensive it is and how varied the vendors were in terms of how they spent their money. And Lenovo's was, was one of the more, or, or less extravagant ones comparatively.

But I, I have the feeling, like with when you look at some of what they rolled out at the show, in particular their ThinkBook Modular. If any of you haven't seen this, please look it up because it is, it is fascinating. Last year we had, their laptop that would flip up and make a really awkward looking laptop for the coffee shop to show off the fact that you have a very high vertical screen.

This year, with the ThinkBook Modular, it's basically the back of the screen is a s- is a screen, which is a second screen that you can pull out, set next to the laptop. The, keyboard itself can be removed from the laptop, and that screen could be placed on that partition, making basically a, a hinged side-by-side screen that you could put up in the coffee shop to show people just how you've, you know, your money is, is being spent. Um, but what struck me as, as really cool about this isn't, isn't this modularity for the keyboard and screen, but the fact that Lenovo now looks very much, like, I, I guess you, you would call it their rival Framework, the Framework laptop, which is very much, you know, plug and play.

I need a USB-C port, plug it in. I need, an adapter for SCSI, plug it in. Uh, and this, this laptop from Lenovo is built just like that, and I, I very much applaud their work here in moving us forward in terms of building laptops that are extensible.

And it's more than extensibility actually, it's, it's repairability. Um, for those of you who follow Lenovo, and, and perhaps are ThinkPad bigots like I am, you may have noted that with their recent, T14 release, I think it's Gen 7, that, they received a 10 out of 10 repairability score. They've always been great to repair, aside from a, a few models.

But my goodness, to build a, a, a ruggedized, you know, enterprise grade laptop that you can repair any way you see fit, new keyboard, new RAM, new hard drive, new monitor, whatever you want, that is impressive. So, I, I tip my hat to Lenovo here. Yeah.

Sounds it. I actually saw a video that, that, that, the laptop you're talking about on, on, on Twitter. I was, basically I probably walked past the Lenovo booth many, many times.

I didn't go look at it. But yeah, it was- 'Cause it's not flashy. Yeah.

It was quite, it was quite something. But yeah. Cool.

Well, it is interesting because so many people are, you know, nerds, like us are very excited about, like, the Framework, laptops because they're, repairable. But it's really exciting to see a major brand leaning into that and- Yeah ... you know, not using proprietary screws and everything.

I mean- Yeah ... you know, it's all well and good that Apple stuff lasts a good long time because it's so well built, but boy, once it goes, it goes. And, and, and yet, you know, these things are, you know, they, they're, they're not just, repairable, but, somewhat upgradable too.

And that's a- Yeah ... that's a question that I have too about, you know, some of the AI laptops and, and machines that are coming out there. Um, you know, what happens when, state-of-the-art advances and hardware requirements, need more?

Um, you know, d- I- I- is it just gonna be a throwaway culture? Are we gonna be able to upgrade these things like Lenovo's suggesting? Well, if prices continue as they are for silicon, I, I'm going to say nothing will be thrown away.

As we're seeing in the gaming community, older devices are, are being utilized and reutilized regularly right now. So I think this repairability and upgradability is, is gonna play a huge factor as we move forward for how hardware is built. You know, if, if, for instance, I don't need, you know, DDR5, I can get by with a slower RAM, maybe, you know, I have a laptop that can accommodate both, so I can use chips from five years ago and I'm still humming along.

Just a thought, Lenovo. Make it happen. I think that may end up being easier said than done on that particular point.

Probably. Right. I can dream.

Yeah. Yeah. But with the price of RAM, I mean, that's another interesting aspect.

Um, we also got, last week, Apple introduced the next generation of their M5 Pro and Max. Yeah. And, one point that, went really under the radar is that they've canceled the high RAM version of the, of the laptops.

I think that they had a 256 gig RAM option or something like that, and now it's down to 128. Um, what I've heard in, from insiders is that essentially it would've been priced out of the market because those chips- Mm ... are just not, you know, they're, they're just too cheap.

Which is incredible because, if you look at the, the pricing of those models, Apple has not raised prices. I mean, their RAM was always extremely expensive. Yeah.

It was always very expensive to go with the bigger RAM model. It is still very expensive, but actually no more expensive. It is the same to the dollar- ...

as it was with previous generations. I wonder if they'll be able to continue that when it comes to RAM and SSD in the future, and I wonder, did you happen to notice, are other, vendors, either cutting back on RAM or raising prices due to the RAM crunch? Did you see anything with that, Nick?

I, I haven't seen anyone, dropping the, the amount of RAM back as- No ... except Apple actually. No, I haven't.

Right? With the M- But as a, as an, as an Apple user, I know exactly what you mean. You've gotta buy your RAM up front, because doing it later is basically, technically i-impossible.

With the M series, it's physically impossible. Yeah. Right.

Right. Yeah. $200 a step.

It's, that's not cheap, but cheaper than a DGX Spark. Yeah. But that being said, you know, it is, it is interesting that they've managed to hold the line.

So you didn't hear a lot of, um... 'Cause in briefings and so on, I do hear a lot of vendors privately talking about the RAM crunch, and the expense of RAM, and the lack of availability. Did you see that at MWC?

I didn't, but I must admit, I probably wasn't looking for it. I w- I was, I was looking at, the different, different layers of the stack than, than the RAM and the, and the bus. You're whistling past that graveyard.

Mm-hmm. I do hear, outside of MWC from the, the storage players qui-quite a bit of discussion about this in terms of long-term options and what they, what they want, how they want to approach it. And by and large, it is, it is very much keyed around optimizing what, what they have to, to look for the bottlenecks and to squash the bottlenecks.

That's, that's the key. So in other news, we've been hearing a lot about NeoClouds. In fact, I was amazed to see, the term NeoCloud and Enscale specifically mentioned, in above the fold on, you know, New York Times, Wa- you know, Washington Post, you know, sort of, mainstream news sources this week talking about, what apparently is the largest C round raise in history for a European company.

Nick, what's your perspective on Enscale? Yeah. It's, it is interesting.

" And they wouldn't be particularly, uh... That's a quite justifiable position to, to have because this company didn't really exist, two and a half years ago. Uh, so Enscale has raised a $2 billion, a th- Series C round, which it says is the largest, Series C round in European history.

About six months ago, Mistral raised a $2 billion C-C round as well. So it obviously... There, there might be some currency exchange fluctuation going on there, but it doesn't, it doesn't really matter.

Um, it's, it's a lot of money. And I think the... You know, this also gives it a valuation of, just under $15 billion.

Um, and then we've got, you know, we've got, you know, Aker ASA, and 890 Industries and, and some of these companies, these, these, these funders, coming in again for the company. So the company was es-essentially spun out of what was originally a Norwegian, Bitcoin mining firm. And so, you know, this is a, you know, this is a fa-familiar tale, with different countries inserted, I guess.

And so the, the company, was part... Last September in London when Jensen Huang came over, and sort of anointed his, his favorites, among the European, or specifically UK startups, you know, they were one, and they said, "Yeah, I'm gonna invest in you," had the prime minister on the stage and all this kind of stuff. But at that point, they'd only raised $185 million.

So I saw in the announcements, Enscale has partnered with Microsoft. It's gonna... Back then, this was.

They're gonna build these data centers. " They haven't got any money. So of course they're- Other people's data centers, Nick?

Well, no. They- Is that how you do it? Well, they raised some debt.

They did some debt funding in the late, late of 2025, and now they've raised, this money. So, putting aside... And it-- Also, they, they announced a, a couple of, well, sort of three board members, Susan Decker, Sheryl Sandberg, who, who we all know, and Nick Clegg, who's the former deputy prime minister of this country and, um- Oh ...

was, obviously on Meta, a policy guy for, for, for many, many years. So they're kind of, sort of throwing... They're trying to throw their weight around a little bit, in terms of, in terms of, the building.

And they're also part of the Stargate Norway, you know, star- project last year. But anyway, it's, it's, it's interesting, I guess, from, you know, kind of broader, you know, sovereignty, issues. They're kind of, kind of threading a needle a bit though because they're kind of being a European sovereign play, while being sort of deeply embedded in sort of US companies like, like Nvidia, you know, providing with, chips.

Like Microsoft is partner in, in data center. Well, Microsoft's not doing the build, but Microsoft's, helping fund these things. So it's...

So, so it's a kind of... It's an, it's an interesting angle, and it's a very similar story, to the likes of CoreWeave, in the US, Nebius in, in the Netherlands, and Iren, which is an originally Australian company. I know it's floated in the US.

Um, so it's, so it's kind of... It's an interesting play. But one of the numbers, that, that kind of caught my eye was, the company said in the next five years, so that's, what, 2031, it needs to raise another $45 billion.

So- To do what? For data centers. So, so that's quite a lot of money.

Yeah. And so we've gone from, like, 185 million in September of last year to, you know... I, I, it's raised, you know, f-a few billion now, including this two billion round.

And so it's... There's a lot of, you know... I mean, we know the story obviously of the 690 billion, committed CapEx from the, from the US hyperscalers, plus including Meta, who's not really the same business, but put them in the same bucket for this.

Um, but there's a lot of this going on in, in, like, in other countries, I guess, is the point. You know, we, we've heard about some, some of the ones in the Middle East in the past. Um, obviously, that's a slightly different situation now perhaps.

Um, but there's, you know, there's British ones here, there's, there's French ones, there's German ones. There's... So it's, so it's a bigger story out there of, of, of what people are kind of calling the circa, you know, economy of, of, of this.

And so, you know, obviously be tracking them, pretty closely as to see, see what, see how, what they're actually, how they, how they're gonna be getting on, over the next, few months. And I have to say, I, I don't think that it's just pure speculation, although that does play a massive role in both the large scale data center build-outs and the NeoCloud build-outs. Um, but if, if you look at how the AI marketplace itself is building, the software that would ostensibly run in said data centers, you can see this shift right now toward, what I would call more a-ambient inferencing, which if, if you think about, you know, even, even, last week with, ASUS, which now has the NUC, NUC name for the small, you know, devices, desktop, you know, PCs, that they stole from Intel, that, basic- basically, um-They, they are used Editor's note, nobody stole anything here.

Go ahead. Right. Sorry.

They, they, are now using, which I think is a wonderful brand, by the way. Um, anyway, the fact that what they're promoting is, this device, not as something that you do your spreadsheets on, but instead something you run your Claude bots or your, your OpenClau, or now your Microsoft Copilot CoWork, on twenty-four/seven as an agent that just runs in the background, I think speaks volumes to what the vendor, landscape anticipates for AI in the workplace. And that is, as I just mentioned, this ambient, intelligence, but it's, it's more than that too.

Um, as we've known and we've talked about on this show for a while now, there, there are different classes of models that are your, you know, frontier scale that you basically, your Geminis and Anthropic's, Claudes, et cetera, that you are inferencing from a hyperscaler. Um, and then there is a smaller class of model that is more like what we've seen recently from the, the Qwen team, that is, you know, geared towards smaller twenty-seven billion, you know, on down, let's say. And those are all, you know, giving, builders in the enterprise much more optionality in terms of where they run their inferencing, where they spend their inferencing dollar.

So, you might, for instance, have a plan with Anthropic, you might have a plan with Google, and you would have the ability to, to then supplement or extend that through Neo Cloud partnerships, or running locally as well, all three, that you would use to, to sort of optimize your own spend on inferencing to route the request to the right model at the right time, at the right price. So we're seeing this, this sort of, you know, move toward being able to, to tailor your inferencing spend to the task at hand, and that's why we see such an opportunity for these, these Neo Cloud providers in particular. Yeah.

And quite... And you mentioned, Qwen there, Brad. There's, been some, some developments, haven't there?

Yeah, there sure have. It's, it's such a well-regarded family of models and has been for a long time, especially for code generation. 5 release that they just came out with, the, the entire, you know, AI, you know, community was a quite abuzz, were they not, at, at just how good this family of models, you know, is.

And, it was very strange that shortly thereafter, as in, what, what was it, Steven? 5, and then suddenly, oops, gone. Yeah.

But a- and, and, you know, to, to their credit, well, you would imagine they would do this, but, Alibaba ver- was very quick to say, "Look, look, look, we're not shutting this down. We... This is going to continue.

" And, at MWC, there was a booth that they had, with the Qwen glasses, which compete with the Meta, Ray-Ban glasses. Uh, it was very much on display and had people talking about it, even though it was monochrome in green. I...

F- So fantastic that the text is in hacker green, by the way, in your, in your heads up on the glasses. But, it, it's, you know, is making some waves, and not just regionally, but globally. Mm.

And just talking of people who've left, interesting lab companies, just coming back and coming back to, to Europe for a second, if I may. Um, so it... I mean, f- 10 years ago, you know, European rounds were like, you know, "Here's 5 million euros, here's 10 million euros," et cetera, et cetera.

And we've already just talked about Enscound, there's 2 billion. 03 billion, obviously trying to translate the euros to get it over, seed round, which was quite something. Uh, and they're based in Paris, and they're gonna have office...

people in, in New York and Singapore and, and various other places. And, the Go- the former Google, the DeepMind guy, David Silver, he raised a billion dollar seed round last month- Mm ... led by Sequoia.

Um, he's in, he's in London for to develop, his, his company called, it's called Ineffable Intelligence, to build some sort of human, you know, super intelligence thing. So it's, it's, it's, um... So some of these people are leaving, amicably, some of these people are leaving maybe, in Yann LeCun's case, perhaps slightly less a-amicably.

Um, but they're chas- There's a lot of money chasing them, all over, all over the world, no matter where they might be. Yeah. One thing that, that I wanna fi- kind of, I guess, maybe finish on here since we're getting to the, the end of time, not literally the end of time, but the end of this episode's time.

Hopefully. Um, maybe we're getting to the end of time too, I don't know. Um, but, I stirred up a little bit of a hornet's nest in social media this week by, suggesting that, local, LLMs are bunk.

Um, not generally, but, specifically with regard to, agentic applications like OpenClau. " Um, a- and, and I'm raising my hand here and saying, "Point of order. Point of order.

No one is running OpenClau locally, even though they think they are, because they're all using the, large SaaS, models online to actually do the, the work, the AI work," because no one has shown, or at least I haven't seen-Anyone who has been able to build a workstation powerful enough to handle the kind of context windows and the large size models required to do the actual, any, any decent job of actually OpenClawing. Now, you could certainly use some local models, and I've seen some people try to do it, but the problem is right now the hardware just can't handle it, and the, SaaS models are so good, so much better that essentially every FAQ out there, including my own experiments with that, have shown me that you just can't run it locally. Now, do you think that's gonna change, or do you think that, these, alleged local captive-owned private agentic experiences built around OpenClaw, and, and similar technologies, do you think that that is actually gonna be able to be run locally?

Um, may-maybe, Brad? Yeah. I think if, if we can all afford DGX Sparks, absolutely.

Um, there's no reason why you couldn't run that locally with, with a large enough model. But that's, as you, as you rightfully pointed out, Steven, is, is the problem. It's, it's simply physics in terms of the VRAM you need, and the context window that these agents require in order to function properly over time, and that is the real problem.

I'm sure, you know, with, with a, a model that, that has a diminutive context window, you could offload certain tasks that your local OpenClaw agent might do, such as organizing your, your inbox or your, in downloads folder. Do you really need a frontier model to do that? No.

So you could route to that local model to offload from the cloud-borne agentic model where appropriate, necessary, and I, I would think that we will see more and more, sort of more mature implementations of this idea of a local agentic system that will incorporate those kinds of tactics, if you will, or best practices. But for right now, yeah, you're, you're, you know, if you're gonna run this on GPT OSS twenty, you know, two hundred, is it two or... No, it's twenty billion parameters, for example, you are going to have a very bad experience.

Yeah. Just full stop. And, and, and not only that, but it's, it's just really practically not going to work.

" And the tr- the thing is, I've tried it. I tried to run it with Ollama locally. I tried to run it with, you know, DeepSeek and, and Qwen, and some of these other models locally.

Um, they're too slow. They don't have the context window. They don't have the, reasoning ability to get a functional OpenClaw running.

Now, again, maybe if you've got a DGX Spark, I don't. Maybe if you have a, quad Mac Studio setup like Jeff Kuerling showed up- Daisy chain twelve of them ... on his blog.

Yes. You know? Yes.

That, then that's gonna work. But for me, in my experiments, it absolutely didn't work. " So, so we'll see if it is in the future.

Again, I would love to see that become practical, but I think that that's gonna come head to head with the RAM crisis, at least for the next year. Yep. So before we go, again, as I mentioned at the top, this week as you're listening to this is GTC.

Now, you probably noticed that we didn't, put our, you know, our crystal, our Palantirs, I'm sorry to mix metaphors there, and, look forward into the future of what GTC will bring because, you know, we can't really talk about it even if we were pre-briefed or even if we do have guesses. Uh, but we definitely will be, coming back to you next week with another episode of Utilizing AI. Before we go, though, Nick, what do, what do you look forward to, that you can talk about as of now?

Yeah. So I mean, obviously, the, we know there's gonna be some, some sort of chip announcements. I'm not-- I, I don't have privileged information on that, and I wouldn't obviously disclose it if I did.

One thing, though, that I suspect Jensen might, will be talking about is this five-layer cake, metaphor that he introduced it at Davos, and he's actually just, they've just published a, a blog post on it, NVIDIA have. And it's the five layers are, are energy and chip/uh, semiconductor manufacturing, computing infrastructure, including data centers, foundation models, and then applications. And what his, his argument is essentially is that, you, you as a, whether as an enterprise or, a government or an investor, really need to think about AI in, in all these levels.

Um, and all that means is not just software, it's, and it's not just the models. We've always said, you know, we all know the models are not the applications, and the applications themselves are not enough. It requires all these things, to be looked at in, in sort of in a holistic, holistic way.

So it's kind of framing for, for investors, for enterprise leaders, and for policymakers, and I think that's partly, you know, NVIDIA's become a lot more active in the, in the lobbying and the policymaking, kind of domain over the last couple of years. And I think, you know, we're, we're talking about here for some sort of, you know, the way we look at energy and data centers and the interaction between those and the foundation models, just trying to put all that together, so to make people think about things like that. It's hard.

I don't know what's gonna be his keynote at this point where I'm, where I'm talking, but I, I'd be surprised if that's not mentioned. Yeah, and I would love to see, some mention and some work around their support of the open source community. Uh, they have been, despite, you know, having a lot of locked down IP, and rightfully so, it is, it is theirs, they have been quite active in promoting and supporting the open source community.

And I think that, before we came on the air, we were actually chatting about this a little bit in it, that, Jensen has some, sessions lined up for the show in which he's going to bring to the stage and, and talk, speak with the open source model makers about what they need, what can they look to NVIDIA to support them with. So I like that. Yeah, it's a key-- That's a key session for them, I know.

And, Jensen does one of these a year. " This year is about open. He say he's got the LangChain, Cursor, Reflection, Perplexity, Thinking Machines, Mistral, OpenEvidence AI too, and Black Forest Labs.

Um, and so for them it's, it's great to have him be the moderator, and it's, it's, it's a huge draw, and I think it's on the, on the, on the Wednesday. So yeah, that, that's, I think that'll be like a, a highlight of the, of the entire show. Well, we will definitely be, covering GTC next week on this.

Again, we record, you know, on, on, on Tuesdays ahead of a Wednesday, episode. That's, you know... So, so it's gonna be a little bit, a little bit delayed.

Uh, but, if you want your fix, on GTC news, I would definitely keep an eye on the, analysts, researchers, and community around the Futurum Group. Uh, we will be covering this live on our social media. We will be posting articles on the Futurum Group, website, as well as the various Techstrong sites, including, Techstrong AI.

So do keep an eye on that. And of course, we will react on the nep- next episode of, utilizing tech or utilizing AI here. Um, before we go, I do wanna give y'all a chance to, pitch what you're, what y'all are working on.

Um, I, I, I take it that neither of you is going to GTC, but I know that we do have some Futurum analysts going there, from, from the tech field day team. Alistair Cook will be there. Um, Nick, what, what are you working on, and, who else is going to GTC, if you know?

So Bre- Brendan Burke, our semiconductor lead, is going to GTC. I normally go, but I'm training for a marathon, and it's incredibly disruptive to that, so I can't. Um, but Brendan Burke will be there, and he'll be reporting back.

Uh, in terms of what I'm working on, I've just published, my new data set, the AI Decision Maker Survey. Um, so that's eight hundred plus decision makers in North America, Western Europe, and parts of, Asia. And so we're looking in there into, you know, use cases, challenges, budgets, and of course, no kind of survey or these kind of reports is, is complete without a, a maturity model.

e. the most mature, kind of companies are. So watch this space.

There'll be, you'll be seeing m-more about that soon. Yeah. And likewise, I'll be publishing, our decision maker for data intelligence, analytics, and infrastructure as well.

Uh, I don't, I don't have a five-stage model. Um, may-maybe a two-stage rocket. Uh, but we'll, we'll we'll need to compare notes, Nick, on, on that one.

But, yeah, I'm very much looking forward to publishing that, in the next week or two. Uh, also next week I'm going to be at Microsoft's, Fabcom. So the week you're watching this, I will be roaming around thinking about all kinds of knowledge graphs versus context graphs.

So, stay tuned for more on that. I'll be writing about what I, what I see and hear there. Excellent.

Um, thank you for that. And, I appreciate, having you both join me on this episode and every episode. I mean, it's always a lot of fun to talk to you and get your perspectives on what's going on in this crazy industry.

Um, and thank you everyone for listening as well. Um, this is, of course, Utilizing AI. Uh, and if you enjoyed it, we would love for you to subscribe.

Um, the easiest and most, you know, best way you get to see our pretty faces is if you subscribe on YouTube. ai, and the Techstrong TV app, which is available in most app stores. Uh, this podcast is brought to you by the experts and analysts from the Futurum Group, where insights meet AI.

Uh, thanks for listening, and we will catch you next week.