Futurum Media, AI Jobs and NVIDIA’s New Rivals

Hey everyone, happy Thursday. Whether it's raining or sunny or windy or whatever at your place, it's the dawn of a new era, because today we launched Futurum Media, and we're going to kick our show coverage off with it today. We've got an expanded panel.

" I'm looking where Alice is played by Ann B. Davis, but we don't have her. But let me introduce you to who we do have.

First of all, our regular Thursday gang members. We've got my friend Drew Gutstein, CISO extraordinaire from up in New York. Drew, good to see you.

Our roving correspondent here at Techstrong, Jamie Hampton. Jamie, it's good to see you. Thank you.

Our man about Silicon Valley, Jon Swartz, and his giant shirt. Jon, welcome. Hi.

Our special guest today, the Futurum Media brain trust, the light is just shining from them. Making his debut on Techstrong gang is our friend Lee Sellers. Lee, welcome to Techstrong Gang.

Thank you. Holding the Paul Lynn square in the middle, or whoever he wants to be, our good friend Steven Foskett. Still a little hurting over his socks, but Steven, we appreciate you being here.

Of course. Speaking of Red Sox, he's up in Boston. The one and only Dan O'Brien.

Dan, looking dapper. Thank you. And the man in the high castle up there, Mike Bizard.

Mike, welcome. I'm trying to get this over Steven's head, so. It is.

It's fitting right over his head. Okay. Let's not rub it in.

But guys, seriously, what a great day. We are launching Futurum Media. Dan, in a lot of ways, this was kind of your little brainchild.

Not that you have a little brain, but it was a little idea in your brain. So I'm going to let you go first. What's Futurum Media?

Yeah. Thanks, Alan. So Futurum Media is really bringing together all the capabilities we've always had inside Futurum, but bringing them into a much more integrated portfolio, much more integrated client experience for our clients.

And really, I think we've always had this fundamental belief that Futurum Research, and all the great data and analytics piece that we have, along with all of our media and our reach and our distribution and the audience that we have on the other end, that those two assets were always better together. I think that's been a fundamental belief in how we built out the company here at Futurum Group. And, I think we've really now pulled the media side together in a way that allows us to take all that great content that's coming out of ETR, the custom research team, Signal65, and really take our clients on that full campaign journey, the full activation.

Events, web media, digital, video. I think every one of our clients has a story they're trying to tell, and a lot of the work we do on the research side is helping them understand who they're trying to tell it to and help them tell it better, help them tell it in a way that resonates more. And they're all trying to reach an audience, and I think we have a lot of that audience.

That's the folks that are consuming this content right here. That's the Techstrong Gang audience. com and Security Boulevard.

The many folks who follow us on social media, YouTube, all the places that we show up across Futurum. So, to me, yeah, Futurum Media is really about that go-to-market activation side. Working for tech vendors to help them get the story that they're trying to tell out to the audiences that they're trying to tell it to.

Excellent. Hey, Lee, you're new to the gang. Hello.

Give people your take. What gets you excited about this? Yeah.

Well, first of all, apparently, this is the first time I've ever been on live TV, so I just wanted to call that. It's very exciting. Want to say hello to anyone at home, Dan?

I want to say hello to everyone. I mean, the reality is the next audience is not human. I mean, it's the AI that's standing between the vendor and the buyer.

And when you look at Futurum Group, which we're proud members of that organization, this company, this analyst firm, was born out of the AI age. We understand that buyers are using AI to do their research, create their shortlists, get their list of requirements, and using that as much as anything, as any tool, in order to make buying decisions. And we're ready to help support that reality.

We understand who the audience is. We know how to create that third-party credible content that's engaging, that's going to pique the interest of all the agents out there that are doing the hard work of researching technology solutions. I mean, this is really an exciting time to bring all these capabilities together under a media umbrella.

It's definitely a different organization from the previous media organizations that half of the people on this panel have spent decades supporting. So this is great. Good stuff.

Steven, I know you have your views on it, and I wanted to get that out there. Yeah. Let's be true to Techstrong Gang because it wouldn't be Techstrong Gang if I didn't say, "Now, Lee, I completely disagree with you on this.

" No. No, the truth is that- Good luck ... for me, the exciting thing here is that it lets us really stay true to who we are.

Tech Field Day is a special thing, and it gets to stay Tech Field Day. Techstrong is a special thing. It gets to stay Techstrong.

And yet, we all get to do a lot more. And I think that that's the most exciting part of it. We don't have to try to be everything under the Tech Field Day brand or the Techstrong brand or Security Boulevard or anything.

We're able to work together really nicely. And personally, I love the fact that it lets me get on Techstrong Gang and the Futurum podcast more often. I think this is my sixth recording of the week, which is pretty exciting, too.

Good for you. And you know you have an open invite any day, Steven. Every day.

I don't. Now, Lee, you do, too. Look, now that we're Futurum Media, it's- That'll be great ...

it's all good. But let me kind of put the cherry on here if I can. " But when I first met Daniel Newman and wanted to put this together, put Techstrong as part of Futurum, I think shortly thereafter, the next two people I met were Steven and Lee.

And we've been talking about putting it together because we're stronger together than we are in individual silos. We've been talking about it for two years. It took Dan and the rest of Futurum and two years for us to put this together, but we've put it together.

What I'm excited about it is, as I think all of you have hit, it's not just the media piece of this. Futurum Media is a piece of Futurum Group. And there isn't a competitor in the market today that goes from this site, at Signal65, let's call it, our testing lab, that has a whole bunch of Nvidia equipment and all kinds of other equipment sitting there that we could run real-world tests on what's going on in silicon today, hardware, software, everything.

Through to the analyst, through to ETR with a 10,000-strong, $2 trillion IT track of spending on IT, right? Into this Futurum intelligence portal, which is the AI brain of Futurum, if you will. Right?

Then out the other side into media and amplification and all of the special things like Tech Field Day and Visible Impact. Help sponsors, help vendors bring that message to market, the Techstrong amplification. And let's not leave out Six Five Media as well, right?

Doing that kind of executive analyst to sponsor thing. There's not another game in town that has that breadth That vision, that's now reality. And that's what I'm pumped about.

John, Mike, you guys, when you said some of us have been here decades, that's you two. Have you ever seen anything like this? What do you think?

Well, John and I were talking about this just before the show started, and we really like ETR, which is a relatively new asset for Futurum because it gives us access to data and analysis that's a little more forward thinking. And we've talked about us and a forward strategy with Futurum on a previous show. But John, you've been looking at research data as it relates to journalism for a long time now.

What's your take on what's going on here? So one of the pet peeves that you and I, Mike, have had for years, I believe, and most other tech reporters, is that by the time you get to some of this market research, it's old, it's obsolete. It's three months, four months old.

I think what I really like about ETR and Futurum information is that it is timely, and it spins forward. It looks forward, it doesn't look back. And that's what I really need.

I need people to tell me, not just where the market is, but where it's headed and where they think it might happen, because we're trying to anticipate that. So I do like it. And I've noticed, and this is not blowing smoke to you, Lee, or to Dan, we've been doing a lot of stories off of the ETR research lately, and I find it really interesting and insightful, and I honestly mean that.

And because there is a sea of stuff that you and I might get every day that we just kind of discard. Mm. Well, just a couple of weeks, John.

We had Dario Amodei with the AI pacing argument, and a couple of days later, we had 100 CIOs from large North American enterprises weigh in on that whole discussion. I think that that real time- Right. I remember writing about that.

And actually, in today's episode, we were going to talk about a couple of your reports, I believe. We will. I think Jamie wrote one of them, wrote about one of them.

Some really interesting spend data going on in terms of the shift to cyber. Yeah, look forward to get into it. This does raise an interesting point, though, getting it back to the relationship with Futurum Media, and that is, and we've been doing this a long time, I think to break the news to everyone, but the lead is dead.

A spreadsheet of names is not pipeline, it's inventory. So how do you move that into real demand? It's understanding buying signals, it's understanding what's going on in the marketplace.

It's understanding how the big thinkers, the CXO suite, where are they moving the markets right now, and how can you align your product solutions to fit the moment? And ETR is exactly the tool that companies can leverage when they're building their go-to-market plans, their strategies, and their messaging, and it's all wrapped together, and that's an extraordinarily unique capability that we bring to market. Well, spent investors have been using their data to place their bets for a decade.

It's nice to see tech vendors starting to place their bets around that signal, too. I will quibble with one thing that Lee said. As much as AI agents are involved in the research process these days, it's still people who sign the check to buy the thing.

Today, that's the case. We've got the robots. My Wells Fargo account somebody signed a check.

Jumpers doing the software, Mike. Yeah. Well, but here's the thing to remember here.

This is the dawn of a new era. I grew up in my career looking at reports that were based on data from six months ago, a year ago. We've used the word forward now a few times.

I want to kind of dig in there. It was about two months ago, Daniel Newman wrote a manifesto, put a whole thing online about Futurum Forward, the forward analyst. Taking a page out of the Palantir model, but combining it with Bloomberg and some of the other information places out there.

Making the analyst part of the decision team at their customers. Not showing you what happened last year, but helping you decide what's going to happen next year. And that, to me, resonated, and here at Techstrong, we've tried to be a forward media company.

What does it mean to be forward media? And Futurum Media is going to be forward. We want to help you with what's next, we're not here as historians detailing what happened six months ago or nine months ago.

Our reporting, our focus is on what's happening next. When Drew sits in that CISO chair, he doesn't care what happened last year. He wants to know a little bit of what happened last year.

" But he has to tell his board, what's their exposure next year? Quantify that exposure for me. And that's the kind of research, bedrock research that we're going to get out of Futurum, that we can amplify and work with people like Drew, that Jamie writes about every day.

And that's what this is about, really, in my mind. Just I want to do some housekeeping. I'm sorry for the shameless plug.

com/media. Forgive us, today's day one. We'll be building that asset out over time, but it's already up there and functional.

You can still go to Tech Field Day or Techstrong or all of our Techstrong sites are there. Visible impact. None of it's going away, it's just coming together.

This wasn't about making things-- I think, Stephen, you hit on that right at the beginning. Nothing's going away. It's coming together.

So really excited for Futurum Media. I don't want to keep blowing our own horn. I don't know.

It's just not the right thing to do, maybe. But I'll end it with this. I am really pumped to be doing this with these guys and these folks, but it's not just Dan and Stephen and Lee and I.

There's an army behind us. All of Futurum, all of our Field Days, Techstrong, VI team. And we're ready.

I wrote on a couple of the articles I did today. I'm pumped to get back to work on this and really knock it out of the park. But Mike, we need to hop into another segment because we're over time.

What else? Is there anything else newsworthy today? Well, yes.

But as promised, we're going to look at some research data, and there's a lot of things out there that sometimes are conflicting, some things are a paradox. And yet, as we all know, multiple things can be true at the same time. So let's start.

McKinsey's talking about how 11 million jobs in the US alone are going to need to at least evolve or change, and the people who have those jobs are going to have to grow into something else than maybe what they are today. At the same time, ETR notes that, well, in general, IT budgets and cybersecurity budgets are increasing because we're spending more on AI. However, the challenge is that we also have a limited amount of resources, and in many organizations, they are starting to cut back on people to pay for the AI.

And we're seeing this trend in other places as well, and I think that that's all going to pick up at the same time. So John, you wrote a couple of these articles- Right ... and I pulled Jamie in behind you, but what's your assessment and what's the read here?

So I'm going to give kudos to McKinsey because they, in a sense, apply nuance to the report. Rather than jumping down the rabbit hole of widespread job losses in AI, they're suggesting that the challenges over the coming decade are going to be more about worker mobility than job scarcity. In fact, they actually think the US is going to be projected to feature more overall job openings in 2035 than today through AI innovation, economic growth, and a shrinking labor pool because this is an aging country.

The top level takeaway from this report was that AI and automation were going to require roughly 11 million US workers to move into entirely new occupations in about a decade. But they also project that automation could displace the equivalent of 36 million jobs, but economic growth and new industries would accelerate demand by 40 million. So in a sense, it's a nice even-handed look.

And one thing I do want to point out, too, is that- In terms of what these jobs are, they found a sharp imbalance in how technology displacement is going to affect the labor force. So more, three out of four workers are going to be forced to transition in these three categories: office and administrative support, retail and consumer sales, and transportation and logistics. And another big takeaway is that lower wage workers and those without college degrees are going to face a significantly greater risk of needing to change occupations.

So in a sense, it's kind of a work in progress, and it shows that AI is not destroying the job market, it's just evolving the market, and you have a significant impact. And I think I'm going to pass the baton to Jamie, if that's okay, because I think she wrote about an ETR report that I found- Absolutely. And Jamie, let me ask you this about your story and your impressions on it, but when I read it, it was a clear call that a lot of this stuff is happening first in IT, and we're starting to see help desk positions and people who are hired as consultants and all those things are getting cut to fund AI.

And so my question to you is: Is the time to reinvent oneself now and not, as I watch McKinsey, they're talking the next decade, but I feel like all this stuff is happening in front of us as we speak. It is. According to the report that I wrote about, IT and help desk roles were affected at 60%.

But only 23% said that cybersecurity roles were affected. So, I'm thinking it's a good time to be in cybersecurity right now. It's always a good time to be in cybersecurity.

Yeah. Well, and that's the big question for me is, as spending budgets evolve, are companies spending enough on governance and security along with the normal AI costs? That's going to be the big question.

So Drew, I've never met a security person who said companies are spending enough on governance and security. Right? That's true.

They could always spend more. It's almost a requirement for security to make that statement. Yeah.

But I think what's interesting also, and this is maybe a nuance that wasn't fully embedded in the articles, what's changing is also the kinds of things that companies are spending their money on, right? Mm-hmm. So it's not just a change of the workforce, it's a change of the capital allocation from labor to capital, right?

And hidden in that transition are all these numbers that give tax advantages and because the sort of cost of ownership of an employee in the United States is very high compared to whatever your AI bill's going to be. So there is that sort of, it talks a lot about that kind of transition of roles and things like that, but I think it's going to be accelerated by the fact that these very large companies can actually benefit in other ways by making this transition. Aside from that, I think it's very similar to prior transitions in the Industrial Revolution context, right?

So we went through a massive workforce transition in the '80s, '70s, going from labor to manufacturing to a service economy. This is going to take our service economy to something else that we don't know what is yet. But I like the sort of message that new jobs are coming to replace the old ones.

That's actually an important thing. Maybe the service economy to the software economy. Maybe that's where we're going.

Maybe. I thought Andreessen said that already. Oh, right.

Maybe I shouldn't worry then. But you know what? I want to go to Lee, because Lee, I'll tell you one area where we've seen disruption, and that's marketing and tech.

Our friends in marketing departments in technology have really been at the, I don't mean they're the tip of the spear. I mean the tip of the spear has been pointed at them. Yeah, exactly.

And you deal with them all day. That's a big part of Visible Impact, right? Helping these folks.

Yeah, there's no question about that AI has revolutionized product marketing and how people go to market. It's everything from messaging and positioning to design and content creation, to how they determine who their ICPs are, how they're targeting the channels. It's entirely evolving in real time.

And there is no job, I mean, let's be honest, there is no job, no job on the face of this earth that can't be made more efficient through the proper use of AI. Every individual that works for a living, and even those people that don't, can utilize AI for the forces of good and not evil. And where it comes to product marketing in particular, what we're seeing is that companies are much more efficient in how they're able to determine differentiation.

They're much better at aligning content to the prospects and the customers they're trying to reach. But they need human intervention, and they need partners to be able to do this really successfully. That you need a partner to help guide you to understand the nuances of what's going on, to figure out how you can utilize these tools to be better, more efficient, more profitable, all of that.

Trying to do this in a vacuum, to use AI to figure out how to use AI will never work. And that's what we're kind of seeing right now in that temporary shift in investment. Stephen, will we, and I asked you this earlier this week, but I'm going to ask it again, but the roles inside IT and all these silos, are they going to converge and collapse?

And to Jamie's point, maybe we're all cybersecurity professionals now, and there isn't this distinction between who does what and as we have AI, because I'm going to tell AI my intent, and then a bunch of agents are going to do a bunch of different things, and I might never even see that network interface. You're on mute, Stephen. That is a good point.

I appreciate that. And after what we've been seeing lately with the agentic systems rolling out, it certainly seems that we're all going to have to be more focused on cybersecurity. But that being said, I don't think that most people are prepared to be cybersecurity professionals.

I think that it really takes professionals still, judging by the way that people are using this technology, judging by the way that this technology is using us, as we talked about on Tuesday with the launch of Meta's Muse. And that, by the way, just a little plug, that's the topic of the Security Boulevard podcast next week, where we dive in under the covers of that whole thing. I think it's really too much to expect for everybody to be cybersecurity professionals.

I look forward to the time that we have more cybersecurity awareness embedded in these AI agents and systems, which we're certainly seeing, as well as companies really kind of coming back around the corner and realizing that they need to have professionals who really know what they're doing to help head off any of the problems that could be arising from this generative future. Because as we noted here in the linked articles and so on, a great amount of code these days is being written by AI. Software developers seem to enjoy it, actually.

And that is a little bit concerning, since software developers were never really all that good at cybersecurity either. Drew, what does IT look like from your perspective five years from now? I've been advising people not to hire any engineers.

You probably don't need it now, right? So there's this sort of like, this transition is happening now, right, like we were discussing. This isn't like a far-off future thing.

And there's a massive difference between an employee who is AI capable and an employee who is not. Right? So there's another part of the job market impact here that people are sort of glossing over, is that this is a new skill that people are hiring for.

" It's like, actually, I want to see how you use AI for the interview, right? That's going to tell me a lot about how productive you can be. So it's like you can use it for good.

It actually is great when you use it that way, but it's not easy. And I think that's part of the issue and part of why I think security is safe and a lot of these tech roles are safe. The job changes.

We have a new tool, a new tool that allows us to do things faster and better. But it's still just a tool. So maybe I won't need as many security people.

Hopefully I won't. There are not enough security people in the world for what we need right now, right? So, this is a positive development, I think.

I worry about the AI backlash we're seeing, though. Right? People are, and not necessarily as it relates to folks here on the panel or our audience that are tech-savvy folk.

But when I go out into the real world, in real life, I hear people with such doom. "AI, I hate it. I hate the AI.

It's going to make us dumb. It's going to make us extinct. " Some poor guy from Haiti, over in France, his book had 150,000 people read the book.

It was up for the highest awards in France. And the French, they ran whatever the, I forgot, the most popular did you use AI to make this? And it said, yes, AI was used, and now he's disqualified.

Right? I wrote an article about this. The 150,000 people who read it, do they now think it's no good because AI was involved, but they liked it when they thought a human wrote it?

How strong is that backlash? Dan, you see it too. I know we've spoken about it.

What do you think? Yeah, there's definitely a little bit more backlash, but I think it's a temporary phenomenon. I think it's something- It's a phase.

Yeah. I think- Yeah ... it's a phase.

Because I think the value of expertise is going to go up, right? I think AI is democratizing intelligence and, in some ways, democratizing productivity a little bit. And to Drew's point, it's going to be those really highly skilled people who figure out how to use AI.

That's the workforce people want to gravitate to. I think we ultimately need less people to get the job done. But when you get people who really know what they're doing and can use AI to solve for all the BS around the edges, that's the kind of employee you're looking for these days.

I think that part of this whole thing that I find very shortsighted is that people are cutting back on so-called IT expertise, at least according to ETR. But think about it, we're going to have hundreds of thousands of AI agents running around, more applications than ever running, and somebody still has to supervise all the agents that we're using to manage all of that. So I think that all these folks who may be cutting back a particular type of headcount over here are going to be hiring another headcount over here on the other side that has the expertise to do that.

And I think there's an opportunity for many folks to reinvent themselves. But this idea that somehow or other all this stuff is automatically going to run by itself seems implausible. For now.

For now. The only way to manage this many agents that are going to be out there, Mike, is going to be smart, skilled people using AI. Yeah.

So I'll have agents managing agents, but ultimately, somebody has to manage the agents managing agents. That's probably going to be part of the new emerging kind of AI specific jobs that we're going to see in the next couple of years. We're going to see a development of that area.

So it's not all going to be doom. You're right, Alan, we're going through this phase right now, and it's always a phase of distrusting something that's new that you don't quite grasp, and you think it's a threat to you. And it's usually technology that's always finger pointed at as this is going to cost me my job, but it's going to evolve.

And as we find out and as we start embracing the technology, we'll never look back. This is the job mobility point from the study, John. Absolutely.

Yeah. Mm-hmm. You're never going to have 100% of people agree on anything.

And right now, if you don't have 30% of the population disagreeing that social media as a business model fails- Sweet ... so we've got to keep this going, otherwise what else you got? I just try to remember the market's never wrong, right?

So you can not like it, but this direction that we're going in- Yeah, of course ... feels very clear. I don't think there's any opposing force that's going to get away with it.

The shock of rapid adoption of these technologies is taking a lot of people, setting back a lot of people. We've never seen anything like this. It is.

Our book is going to get into this a little bit coming up next week when it's released. Well, this era of politicization, too, we're getting hit over the head. The general public's getting hit over the head with PAC spending and advertising and demonization of AI as a midterm strategy.

Well, the data center issue. Right. The data center issue has gone toxic.

That is a real issue. Yeah, no, it's a toxic issue. Just the fact that we're calling it an issue, Alan, I think is interesting.

Yeah. It's an opportunity. Technically, it's an information warfare campaign that Russia- Okay ...

and China are fighting now. Spoken like a cyber warrior right there. I was just reading it.

So half the people who are complaining about, by the way, I understand it's now called super intelligence, are saying that the issue is that they use it anyway. They may hate using it, but the other side of it- Well, there is that. They use it in spite of it.

And- But we saw this early on with developers, 70% of them said they weren't so sure about the security or the quality of the code, but 90% of them used it anyway. Right. And if you look at the Bear Dev survey that's on that list over here, they prefer it now because, well, they're onto more interesting things, and they have better job satisfaction because they all realized, you know what?

I hated coding in the first place. There you go. Hey, Mike, I hate to do this, but we're over time on this segment.

What's segment three today? Segment three is kind of we're moving to some AI and geopolitics, but the folks in China, I think I always pronounce this company's name wrong, but- Huawei ... Huawei.

And some of the other AI models out there said that they're going to build something that's the equivalent of what NVIDIA calls CUDA, which is essentially the middleware stack that a lot of the software built on top of the various processors runs on. And in the case of NVIDIA, it runs on GPUs from them and their new Vera CPUs. But as you look at this, it seems like the Chinese are making an effort here to unlock that layer, and in so doing, make the software much more portable across different classes of processors.

So Dan O'Brien, is China working to unravel the, basically the proprietary lock that NVIDIA has on the AI market at the moment? I think what they're doing is very logical, right? Huawei's been in the chip business for a very long time.

The last era before the data center era was the mobile era, and they got very self-sufficient at building processors and baseband chips to operate their mobile networks. They obviously moved now more into server compute and AI accelerator space. I think we've seen, and a lot of this is through distillation, we'll make that caveat, but I think we've seen China's models to be quite competitive and what they've done in open source has been generally viewed as pretty leading edge, alongside some of the top American open source model makers.

" So I think we've seen throughout the AI industry system level thinking, not just around server design and the whole rack, but really the software/hardware kind of co-design is a huge, huge thing in terms of getting efficiency out of running these AI models on top of all this AI infrastructure. And that to me is really the problem they seem to be trying to solve in this particular story. Mike, if I can, Futurum has two silicon analysts.

One's Brendan Burke who writes on tech strong semi all the time. The other fellow, I think it's out of Hong Kong, is Rolf. Well, it's out of Singapore, yeah.

Singapore. And so he covers that market really tight, and I read his stuff or watch it when he's on the TVs and stuff like that. Let's be real.

Huawei is not producing the latest generation of NVIDIA GPUs. No. I think they're seven nanometers.

Much further behind too, from their models. Yeah. They're still way behind.

The open weight models are gaining on the frontier models. There's not much of a difference between them anymore. But when it comes to the silicon, NVIDIA is still heads and shoulders above Huawei or anyone else for that matter.

That being said, CUDA is NVIDIA's secret weapon, right? Because it's CUDA that represents that AI stack, if you will. But Steven, you know more about this than me.

I thought CUDA, a lot of it was open sourced anyway, and so it's hard to attack open source because it slips through. Yes, I see Mike and Steven giving me this. Okay.

There's no open source in there. There is open source, but it's at the top of the CUDA stack and underneath the primitive's not so much. Yeah.

And to your point, yes, absolutely. This has obviously been NVIDIA's moat for a long time. And it's been literally decades that people have been trying to attack CUDA as a moat.

If you include graphics processing, then you could include OpenCL and Vulkan and ROCm, and things like that in there. But certainly AMD's HIP, Intel's oneAPI, these were ways for NVIDIA competitors to try to take on what NVIDIA was doing with CUDA, because CUDA has become just incredibly popular. We're now generations into CUDA programming, and it's got to the point where basically software developers, AI developers at this point, they just write to it.

To them, CUDA is x86. CUDA is air and water and bread. It is everything that they live in.

" The answer is right there. It's because these are for the home market that is deprived of NVIDIA hardware and deprived of CUDA. So if you are a Chinese AI developer, you're not going to develop for CUDA.

Frankly, I wouldn't be surprised if a lot of them had embraced oneAPI because it's a really great, flexible, open GPU interface. But having Huawei develop their own interface is all about that strange bifurcation of this market between China and the rest of the world, and has really very little to do with NVIDIA. And frankly, I don't think that the result of this will actually cause much damage to NVIDIA simply because it's apples and oranges.

I disagree in this regard. The models are relatively parity, but 80% of the use cases can be handled by models that are not the most advanced in the world, and most enterprises don't need that. And frankly, they don't need the latest, greatest GPU either.

So I think what they really want to have is, "I want to be able to move my software from different classes of processors as those advances are made. " Dan, what do you think? I think you're kind of spot on, Mike.

This to me is really about inference. We got to talk about training versus inference here because I think, particularly with distillation, most of the Chinese models are still really leveraging NVIDIA on the training side. But as they scale up the use cases, it really becomes more of the compute is going to be on the inference side, and I think that's where China really wants to be self-sufficient.

So making use of domestic hardware, even if it is a couple of generations behind, if they're able to optimize it really well using cheap open source models, small models, I think they can hit the economic curve that they need to hit for a lot of these use cases. And they could take a page out of the book they've done with these open weight models. What if they made the CUDA competitor open source as well, right?

So now you're onto it. I think that there's an opportunity for the foundations, pick one, to show up and, "Great, we're going to create a open source CUDA equivalent, and maybe we'll work with the Chinese for that, and maybe we'll have a bunch of Europeans work with that. " Leave it to the Chinese to democratize AI.

Just a quick point- The bizarro world we live in. But in fact- It gets better, though. Imagine this then, NVIDIA's in trouble, then the United States has to bail them out, and then we got shares of NVIDIA alongside Intel.

It gets crazy. Well, so the government will own a stake in NVIDIA, of course. Just a quick- That's where it's headed ...

point of fact. There is an open source graphics acceleration library that's been around since AMD got into this business, called OpenCL, and hardly anyone uses it . That's true.

I don't think there's an open source thing that'll just pop up and everyone will start using it. We have that. Yeah.

I think the constraint is that- You could do it ... you get better performance if the hardware designer is also designing the software. And that's never going to change.

When it will happen is, we existed in this form back when we used to buy our compilers from the machine vendors. If you had an Intel computer, you bought an Intel compiler. If you had a Sun computer, you bought a Sun compiler.

We do that because they optimize and got the most performance out of these extremely resource-constrained devices. And that's where we are right now on the AI side. So we have these extremely resource-constrained devices where there's actually benefit of coaxing out that last couple flops or whatever, that actually improves it.

But at some point, we're going to end up with a glut, right? So we won't need all that capacity, and we'll be able to do things like use OpenCL, that's maybe slightly slower, because the competitive advantages have sort of tapered down, right? And that's kind of what happened with open source compilers, right?

Everybody just got rid of the pain of having to deal with compiler differences. Nobody ever liked it. And so it was a good transition for the industry to make.

I think something like that will happen here, too. Good point, Joe. Maybe someday, we'll have storage that's optimized for those flops you're talking about, and we can call it a floppy disk.

What do you say? Flops. That would be great.

So NVIDIA, this is when the shoe's going to drop, right? So they're riding a gravy train. And of course, every company in the world wants to be NVIDIA and wants to get in there and compete, but it's really hard.

It's going to take years. They're already doing it, right? So in a few years, NVIDIA will have some pretty good competition.

By then, we might not need it . " Then everything changes in this marketplace, right? It's the indispensability trap right there.

Yeah. It becomes so indispensable that we make it run on our phones. And look, Apple's already planning for that, right?

That's Apple's plan. The edge is the number one risk to the AI infrastructure build-out. Yeah.

No doubt. Guys, the number one risk to us is we run out of time every day on this Techstrong Gang, and we're about out of time now. First of all, Jamie, Drew, I'm sorry for hijacking our usual Thursday show with Futurum Media, but thank you for playing along and joining in anyway.

I appreciate it. Lee, let's not make this the last time you're on the Gang either. Now that you're Futurum Media, you have a standing invite.

Steven, this is your third time on this week already. What are you doing tomorrow? I think I'm eating lunch.

That's what I'm doing. Okay. Well, it is.

It's right in the middle of our lunch. Dano, I appreciate you coming on. I know you're busy as heck, but you took a full 45 minutes here.

I'll be here. Got to get back on work. Absolutely.

John and Mike, as always, it's a pleasure manning the desk with you all. Thank you all for watching Techstrong Gang. You know we do this every Monday to Friday at lunchtime in Eastern Time, noon to 12:45.

tv, but that's also available excuse me, on Apple TV, Roku, Amazon, iOS, Android. If you've got a screen, you can watch this show. But we'll leave it to you to do that.

I always like to watch it live anyway. Anyway- Who has lunch anymore? What is lunch?

What is lunch? It's usually leftovers, I find. Blame AI for that.

Hey, but go enjoy your lunch. Thank you very much for spending your day with us. com/media, check it out.

You're going to be hearing more about it. It's a great day in the neighborhood. Till next time, I'm Alan Schimmel.

We're out. Futurum Media. Go-to-market execution built for technology companies.

Create stories that matter. Reach the right people. Build credibility.

Activate demand. Futurum Media.