Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

AI Outcomes

AI Outcomes artwork

Ep37 AI Bubble? Orbital Data Centers, Massive Investments, and the Future of Inference | with Thomas Odenwald

Dec 24, 202543 minAudio
ShareXLinkedInFacebookWhatsAppEmail

In Episode 37, host Ben examines the AI bubble debate with tech executive Thomas Odenwald (HPE, Orbit Fab, SAP). They recap Star Cloud's orbital Nvidia H100 launch, discussing space data centers' challenges like heat dissipation, jurisdiction, and launch costs. The core focus: Is AI in a bubble? They analyze inflated valuations, massive capex (e.g., Sam Altman's gigawatt plans), circular investments (Oracle-OpenAI-Nvidia), ROI hurdles, the shift to inference, OpenAI's $12B losses, open source rises (DeepSeek), quantum computing threats, AI layoffs (Salesforce's 10K cuts), and job market advice - upskill in AI to adapt. Essential for tech leaders and investors eyeing 2026 shifts in robotics, autonomy, and beyond. Chapters: 00:00 Introduction to the AI Bubble Discussion 00:10 Recap of Last Week's Episode with Philip Johnston 01:39 Star Cloud's Orbital Data Centers 03:39 Challenges of Space-Based Data Centers 06:01 Heat Dissipation in Space 09:14 Jurisdiction and Security Concerns 11:15 Commercial Viability and Launch Costs 16:32 The AI Bubble Debate 23:07 Understanding Productivity Metrics 23:44 Challenges in Measuring AI ROI 24:50 AI Adoption in Commercial Sectors 25:32 Investment Concerns in AI 26:35 Inference: The Future of AI 27:32 Market Reactions and Financial Engineering 29:46 Oracle's Strategic Overextension 32:13 OpenAI's Financial Struggles 33:43 The Role of Open Source in AI 34:47 The Future of AI and Quantum Computing 36:38 Impact of AI on Job Market 39:35 Preparing for an AI-Driven Future Subscribe for weekly tech insights on YouTube, Apple Podcasts, Spotify, and Buzzsprout. Like, comment on the AI bubble, and share. For sponsorships or Google Ads tips, visit our site. #AIBubble #OrbitalDataCenters #AIFuture #TechInvestments #InferenceAI #QuantumComputing #AIJobs #StarCloud #OpenAI #Nvidia #PodcastEpisode #TechPodcast #BusinessTech #AIEthics #FutureOfWork

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

SPEAKER_02We're talking about AI bubble today, right? And um it get different perspectives on it. I've got my perspective prepared, but before we get into that, I just wanted to play. I've got four clips from last week's episode because we had on AI Outcomes Philip Johnston, CEO, co-founder, CEO of StarCloud. He came on the podcast, and then like the next day, he had Elon Musk responding to his tweet, Cup Harvey, everybody, Eric Schmidt.

SPEAKER_02I was just looking on X, everyone was responding to Philip because we were just so lucky. We managed to bag this interview with him the day before the press release went out about the NVIDIA H100 GPU up in orbit. First time uh they've had a uh NVIDIA H100 up in orbit. And um it yeah, just got super lucky with timing. I suspect if I'd been two days later, Philip might have had one or two other people to speak to.

SPEAKER_02But anywho, uh, we we got there. So uh Thomas, if that's okay with you, we'll we'll take a little look back at that. And I think that does link up with bubble on the AI side of things as well.

SPEAKER_01Please, yeah, yeah. I haven't seen the podcast. I'm super interested in getting these research. All right.

SPEAKER_02What I'll do is I'll go to four clips. Each clip, I'll pause, we can have a little uh reaction to it. Keen to get your perspective. And let's go to let's go to clip number one.

SPEAKER_00So yeah, we're companies called Star Cloud Rebuilding Orbital Data Centers, initially to provide cloud and edge compute to other satellites, and then later to compete with terrestrial data centers on energy costs essentially. So um, I mean, and and and really the we're running up against these huge constraints on energy and water from terrestrial data centers. Uh, you know, I think lots of towns and cities are now rejecting huge data center proposals in their in their neighborhoods for the for the reasons of they don't want strain on the power grid and on the water supply.

SPEAKER_00And so, yeah, what we're doing is uh solving that that constraint essentially.

SPEAKER_02So let me stop it there. As a problem statement, I love the simplicity of the problem statement, albeit the solution to it incredibly complex. But power and water, Thomas, two themes of the AI revolution already, and we're in like year one and a bit, aren't we? Really? Yeah.

SPEAKER_01Yeah, so I I'm not sure if I should play devil's advocate a little bit here. And I don't know if you know that. I worked in in uh a company called Orbit Fab in the space industry for a while, which which I think was still one of the highlights in my career. Um fascinating stuff. And I a lot of young people are going into the space industry, which is great to see. And then during my HPE time, we also had one of our servers on the International Space Station, which was also a very interesting, which very interesting subject.

SPEAKER_01So, what he mentioned is is right. I mean, the strain on the grip, the water supply issue. You can now see some of these data centers from space, they're so big and so vast, and that causes problems. So, yeah, naturally, the thinking is why not go to space? No taxes, uh, a lot of real estate, unfiltered solar energy. I mean, what what would stop you from putting everything into space? But of course, there are there are some problems, which I don't know if he addressed it in his podcast.

SPEAKER_01That would be very interesting.

SPEAKER_02Yeah, let me go. We did, and I've got one of those clips specifically around heat dissipation, which is the biggest problem. But let's go to clip number two. Let me hit fly.

SPEAKER_00We started on this two years ago. Um, we started initially looking at space-based solar, and then realized if there was a cheap way of getting the energy consumption part to space, and that is data centers for most new energy projects on Earth, then you could solve one of the biggest problems with space-based solar. Space-based solar is where you have huge solar panels and you try and beam that power down, but you lose most of the power in transmission. Whereas if you can move that data center to space, then you know you can consume almost all of that power.

SPEAKER_00We've just launched our first spacecraft. It has the first NVIDIA H100 GPU or AI chip in it. It's about a hundred times more powerful AI compute than there's been in space before. And we've done a whole bunch of firsts on that. So we've trust trained the first LLM, we've uh done the first uh running the first version of Gemini in space on that spacecraft. Fantastic. Um and we'll be soon doing running high-powered inference on SAR data in collaboration with various DOW entities.

SPEAKER_02So there we go. Two years since the inception of the company Star Cloud, and uh a yeah, I guess it's a big shout out, and and Philip does this reg with regularity to Y Combinator out of that program. So many big tech companies have kind of started small and just grown massively. But there you go, Thomas, they've they've achieved something very special in such a short amount of time.

SPEAKER_01Yeah, no, I I read about uh the I think it's the H100 GPU that's running, that's running with Star Cloud now in space. He mentioned training a large language model. I was a bit surprised. I don't know, I don't think they have the capacity in space right now to train a whole language model. So yeah, it's I guess it's just a matter of time. But he mentioned dissipating heat, right? Convection doesn't work, conduction doesn't work, the only thing that works is radiation, and you need these huge radiation panels up there, which I'm sure they are capable of adding to the data centers in order to dissipate the heat and pull it down.

SPEAKER_02Let me just go and play clip three of four, which is purely about that, the biggest challenge that they are facing. Uh let's go to my slideshow. It's the heat dissipation and the dissipation of a data center. Is uh a data center is super, super hot. We have cooling issues here on planet Earth, right? That's the whole back to the whole energy driver and liquid cooling and the impact on the environment.

SPEAKER_00Yeah, yeah. So the way that a terrestrial data center keeps cold is you either blast air past the chips or you you're essentially evaporating off fresh water so you can have directed trip liquid cooling, and then that fresh water goes out to a cycle, a cooling tower, which is mostly most often is just uh evaporating fresh water. So you don't have air or fresh water in space. Um and so you need a new way of doing that. So you don't have convection or conduction essentially. And the way that you do that is um you need a large surface area that essentially glows in infrared.

SPEAKER_00Everything in the world is glowing in infrared. Like if you had a heat camera on your face right now, you'd see that your face is your face is glowing in infrared. Um and so and that what that that that just means is you're just emitting energy in uh in a yeah, in infrared, basically, into deep space. So that's the that that's what we do. We need to keep a large server at a certain temperature, and then that will uh emit that infrared. And why is that difficult? It's difficult because large surface areas need to be folded up uh for launch and then they need to deploy.

SPEAKER_00And so you can imagine you've got all this piping for a fluid which is going to go past the chips and then out to this large surface area. And so all of that piping needs to fold up and then expand out to this much larger surface area once we once we get in orbit.

SPEAKER_02And so when I look at the kind of graphics and the animation on the StarCard website, and you're seeing this giant expanse of well, I think four kilometers by one kilometer. Is that the infrared or is that the solar?

SPEAKER_00So the four four kilometers by four kilometers is the solar, and then four kilometers by one kilometer is the radiator on the back of that image there. And so yeah, you can think of it roughly one square meter of solar panel, it can produce about 200 watts per square meter in space, and one square meter of radiator can dissipate about 800 watts per square meter, maybe slightly less. And so if your surface area on the radiator is a quarter the size of on your solar panel, then you can that trade-off roughly works.

SPEAKER_02So let's stop that there. There you go. That's the heat dissipation, and technically, yeah, when something is going to be four kilometers by one kilometer up in space, and the solar plot panels even bigger, you can start seeing the challenge, but you can start seeing how there's probably a solution to that challenge, right?

SPEAKER_01Yeah, yeah. So, I mean, that's it's certainly a challenge that eventually they will solve. The other challenge is I see is definitely jurisdiction and security. I don't know if that was mentioned. I mean, who owns space? And if something goes wrong up there, I mean, how how do you can you all can you fix everything remotely? Or what if space debris hits your data center or your radiation panels? Who's fixing that? And then and then one other thing I was thinking of when I when I read about it is when I worked for my company OrbitFab, we built gas stations in space.

SPEAKER_01I think I mentioned that in one of my podcasts early on. And the reason why this was a very viable business model, and Orbit Fab is doing very well today, is satellites die because they run out of fuel, which many people don't know, right? I mean, it's they are perfectly working in working order, but they need at least a little bit of fuel to stay in the location where they are launched, because there's always gravity in the lower orbit. So I'm wondering if what the thought is on that, because you have data centers certainly in the lower orbit, the orbit, and you need to figure out how how do I fuel them to stay in the location where I want them to stay.

SPEAKER_01So I'm sure they've thought about that as well, but I haven't heard anybody talking about it yet.

SPEAKER_02No, yeah, and he did mention that the low all these satellites are in low Earth orbit. And he on space debris, he's not that concerned about that because the um the eventually the space debris you know deorbits and burns up to nothing anyway. But um, he was saying these are going to be in a higher orbit, these once they get the data centers, they'll they will operate in a higher orbit, certainly it's that size. But in terms of use case, what was interesting was he was saying, okay, because I was like, yeah, but it we've still got this lead time for you to solve this problem.

SPEAKER_02How does this remain a viable business? And actually, at the current launch costs, and that's my final uh video clip, will be specifically about his view on launch costs. But he said, right now they have a launch cost that makes commercial use viable to take the existing uh satellites that are doing things like taking images of things on planet Earth, like a battlefield or shipping lanes, and they then go around the earth and then they beam down all of the data back down to planet Earth, and all of the kind of insights and processing of that data happens down here.

SPEAKER_02What their concept is to keep the business viable is to have the existing satellites transfer that data, satellite to their existing satellite to star cloud satellite, and then they will do the inferencing up in space, and then only send down to planet Earth the specific insights, like you just need to know about which ship just got stuck in the sewers canal, and it will just send only that insight down to Earth rather than this giant data package coming down from space to Earth.

SPEAKER_02And he also said, which is really interesting, that there's also so it's expensive to get your data that way, because you're taking all of the data and not just part of it, but also you've got to potentially wait, I think he said half an hour or an hour for the satellite to pass over a base station to download its data, and they can do that pretty much instantly. So there's a really interesting like stepping stone there, which is already commercially viable.

SPEAKER_02And let's just listen to the final clip around launch costs, and then we will move on to the bigger kind of subject of AI bubble, but yeah, let's listen to Philip on launch costs. It's absolutely fascinating. Here we go.

SPEAKER_00Yeah, so right now it's about$6,000 a kilo. That is a that bit the first business model of providing edge and cloud services to other spacecraft works fine at that price. So that's the initial use case. The break-even point for competing with terrestrial data centers on energy is around$500 a kilo, so about 10x reduction of where we are. The targeted marginal launch cost that SpaceX has for Starship is$10 to$20 a kilo. And they've previously stated they're targeting customer prices of$50 to$100 a kilo.

SPEAKER_00So well within the$500 a kilo range, basically.

SPEAKER_02You've also got I presume some other options. I mean, it looks let me come off there. So Starship is key. Uh I didn't realize just how key. So to get down to that$10 a kilo,$10 to$20 a kilo launch cost, it requires Starship. And Starship exists, but it's not deployed yet into operation. So currently they are reliant upon Falcon 9, which does successfully land its booster and yeah, it's done hundreds and hundreds of flights.

SPEAKER_02And that and that current launch cost is$6,000 per kilo. And then then you just like realize, and I didn't know too much until I really looked into this before interviewing Philip. I did not realize the capacity and the step change that Starship represents. Because I'd always kind of thought, yeah, isn't that like the trip to Mars type thing? And probably it is, but actually it's come I did not realize the it's commercial uses if SpaceX can get that landing back consistently and put it into commercial operation.

SPEAKER_01Yeah. So I would be curious to see how many Falcon transports you need to build together a data center, a viable data center up there, because that's where Starship comes in, right? Because of the massive volume it presents to bring stuff up there and launch it. Meanwhile, back on Earth, which we have to deal with for the forecast, I think I think up to five between four and five percent of all of the total US energy in the US is now sucked up by data centers.

SPEAKER_01And this is just the beginning. You heard Sam Altman already announcing he wants to build one gigawatt per week, which is it's it's very hard to imagine, one gigawatt per week in in terms of new data centers. And it's even harder to imagine where should all this energy coming from. And Philip mentioned also the water supplies and everything that that comes with it, right? Real estate and so on. Yeah, interesting times. We said that many times before, but it gets even more interesting now.

SPEAKER_02Yeah, it it does. So, and then you've got AI bubble. So let's talk about AI bubble. And let me get your perspective first, and then I've kind of prepared what I think is it's been building from various sources and data sources, but I I now I think I have my perspective set.

SPEAKER_01Yeah, and I would be curious to hear your perspective. So, my perspective, because my inbox is overflowing with that topic, and everywhere you look, it's mentioned in the top headline. Is it a bubble? Is it not a bubble? Is it comparable to the dot-com bubble in the year 2000 or even the crypto bubble? And I'll come to that back three years ago. Interestingly, because I was reading up on that, and again, I was part of the dot-com bubble back in the days where I still worked for a company called SAP, and of course, the height of the dot-com bubble was here in the Silicon Valley.

SPEAKER_01And I remember walking out to my car in the evenings after work, and there were all these headhunters just standing in the parking lot approaching you. And then, of course, you had people from Webvan and Pets.com, and everybody wanted to headhunt you. And and I actually, I have to admit, I was I was I was laughed at because I still kept with SAP, and everything everybody thought I'm I'm crazy. And um, today SAP, if you if you even if you ask ChatGBT, is flagged as one of the winners coming out of the bubble.

SPEAKER_01So in interesting times, but more interestingly, I saw some statistics. Apparently, the a slight majority of investors think we are in a bubble, and a slight majority of CEOs think we are not in a bubble. So it's it's it's right. First of all, there's a split in the middle, and then it there's a different, uh, slight different tendency on in terms of the investor community versus the the business, the business community.

SPEAKER_01And bubble, the definition of bubble is massive invest, which we see clearly there is massive invest in the AI space. Inflated valuations, yes, clearly inflated valuations. If you just look at the stock increases of AI stocks that that keep the SP 500 going, or the GDP growth in the US, which since ChatGPT launched, the vast majority of every of every growth is in the AI space. Then, yes, there is definitely massive expectations.

SPEAKER_01But then the question, and that was that's for every bubble, is what's what's the path to profitability? And that's how you hear different uh schools of thought out there from the investor community as well as from the from the business community. The notes I made is like what I said, 54% of investors believe AI is now in bubble theory territory, which comes from a Forbes report. And then, of course, um I mentioned on our last podcast, we talked about the MIT report that came out.

SPEAKER_01There's another one that just came out that said 95% of roughly 300 plus AI developments that they observed are not profitable today, despite 40 billion dollars in spent. So clearly that's why people think we might be in bubble territory, because there is this massive investment with no path to profitability yet. And then, of course, and we can talk about that. We mentioned it last time already. If you if you think of what I call circular investment, what Oracle is doing with OpenAI and NVIDIA, and we mentioned that last time again, that clearly is a dangerous path.

SPEAKER_01And there were several smart people looking at the Oracle deal with OpenAI and NVIDIA, being very, very concerned about these kind of circular investments, which again caused the crypto crash in 2022 when Sam what was his name, Sam Friedman with FTX and and Alameda research when they did this kind of circular investment. Yeah, Sam bank and freed. A lot of people see that comparison or parallel to these kind of things.

SPEAKER_02Yeah, yeah. It's um it's kind of I I I I I if I was gonna sum up what my view of this whole market is we're just stupidly early. Like it's just so early. And I know what are we three years since the chat GPT moment when we all first saw that. We're three years on from that, and then there wasn't a generative AI market that we knew of. Google obviously had all of their models under wraps at that time.

SPEAKER_02Not many people were that interested in Nvidia, and so and so look at the change now, and already, already, like if I look at my investments, my equities already it was already riding the tech wave all the way through 2010s, right? Was already riding the tech wave, and then we had COVID, and then this thing came along. And so during COVID, you had 10 years of digital transformation in it in 12 months, if we remember that.

SPEAKER_02And so even the tech companies got even even larger and became an even larger chunk. Of the SP 500, to which I think the stat I heard earlier when I was taking a look at these was like it's 30% of the SP 500 is the Mag 7 stocks. Um, but the the revenue and the profit numbers are a lot higher than that. But it just overall it feels really early. I love what you said because I completely agree. When does it not become early?

SPEAKER_02And when does it not become a bubble? Is when all of the investment in all the GPUs in this whole market is monetizable by the private sector, by commercial and enterprise. And all the way through doing this year, hosting this podcast and talking to business leaders and different types of business, it's clear that no one is kind of denying, yes, productivity gains are there from the use of AI.

SPEAKER_02But just introducing Microsoft Copilot is like me giving you a smartphone. I know it will make you more productive, that seems obvious, but I can't say how what percentage of a we call it FTE, right? Full-time equivalent. Um what percentage of a person am I making more productive? And a CFO wants to see a Excel spreadsheet with the as is cost today all broken down.

SPEAKER_02We're going to introduce the new technology product, and then what are my what are the 2B costs going to be? And then the gap is the return on investment. Yeah. Well, you can't do that with either say something like um a chat GPT or lots of um lots of big companies have their own variant of a chat GPT model. Well, you can't, you're not gonna get a return on investment case. You could make one up, but it's not gonna hold any water because the return on investment case needs to have a really hard, tangible metric against it, which in the outsourcing world was how many less bodies did you need to perform a particular task?

SPEAKER_02And generative AI doesn't necessarily do that, but you kind of know in your heart of hearts, well, if I own a company and I give everyone my version of Chat GPT, will it make everyone a bit more productive? Um take some of the sum total of that, does that compound up to for me to have a better market share? Yeah, probably, but you can't put any numbers against it, and that's what I think has happened with the commercial and enterprise sectors outside of the JP Morgans this year with with AI adoption.

SPEAKER_01I 100% agree. We are we are still early, early days, it's happening so fast. The space is a bit different if you compare it to the dot-com bubble because the players are different, right? These are not the web vendors and the pet.coms, these are the googles and the medias and the Microsoft of this world. And AI is delivering real value as we speak, right? So that that that's a difference by itself. The concern is the numbers that are being thrown around, right? Like the um the CEO of IBM did some accounting examples, right?

SPEAKER_01So he said, I mentioned Sam Altman saying we need to build a gigawatt data center every week. So the IBM CEO argues that costs$80 billion. One gigawatt data center, today's price tag,$80 billion. If you do it 100 times, you need$8 trillion,$8 trillion in CapEx. And out of that,$800 million or$800 billion profit is just good enough to pay the interest rate. So this the math, the accounting math, for some people who are experts in this, it's just it just doesn't add up.

SPEAKER_01More interestingly, Andrew Ang, one of the leading voices in the AI space, saying he thinks training large language models, model training itself might be in bubble territory. But what's underinvested today is everything around inference, right? This is where the apps come in, this is where the real value comes in, that's where the business cases come in. So he thinks there is there needs to be more investment when it comes to inference. There will be different winners when it comes to the inference space.

SPEAKER_01But investors are very reluctant because they struggle. They struggle today to pick the winners in that space when it comes to inference. And even SAP Christian Klein made the announcement recently, right? Saying we we focus on inference because that's where the real value is in terms of application layers, in terms of business case, and in terms of value proposition. So that that's still very early. I mean, the whole space is very early, but in terms of the inference, it's still very early. I think in one of our earlier podcasts, I made the argument that 2025 is the year of AI inference.

SPEAKER_01Might be a bit too early. Maybe 2026 is where the breakthrough comes when it comes to inference and creating real value.

SPEAKER_02Yeah, and we're seeing, I think that's a really good point. We're seeing Google launch its TPU, its Tensor processing unit as a versus GPU, and that causes a huge storm. And the and Elon doing the same thing with the cards on board, the Teslas are purely mainly inferencing GPUs. You don't need such a you don't need an H100 for that. And so Nvidia are now, excuse me, are now and now the whole market's including NVIDIA, but now NVIDIA is joined by other players, are moving to inference.

SPEAKER_02And so it's almost like we've had these shocks. And it's almost like I as an in if I put my investor hat on, I'm still super bullish on long-term uh physical AI. Yeah, autonomy and robotics, absolutely, that's what's coming. How long will it take? I don't mind, but I I'm not de-investing in that. When I look short term, uh, hey, what stocks am I in on the AI side?

SPEAKER_02And I look at the shock waves that hit the market, and the first of those I think was Deep Seek, which we discussed when it happened. Deep Seek came along, and and Nvidia dropped down at what 88 bucks a share. Absolutely got hammered. You've had the tariff, the whole tariff tantrums with Trump throughout the year, and the shocks have come in, and then the new you mentioned the financial engineering deals have come in, and they yeah, to everybody, come on, they just look it looks really like a house of cards.

SPEAKER_02When and Google doesn't need to do those types of deals, do they? They're they're they're just like, hey, we've got this, we can run at a loss for for a very long time. And it's almost like Nvidia just growing so big, they're invested in everything. And then you've got Sam Altman making all these commitments. Where's he gonna get his money from? And then the interesting one that I just couldn't understand, Thomas, was Oracle. So I listened to the Oracle earnings results, the earnings call, the last one, a couple of weeks back, and I kind of like didn't even think much of it.

SPEAKER_02I was like, fair enough, the results seemed pretty good. Yeah, the QA wasn't too there, was nothing there that was I I really took away as like a massive negative. Next day, market crashes or not crashes corrects, and because and I was like, I had to literally go online and go, what was it in that earnings call that I didn't hear? And it's yeah, it's me not being as astute as I I should be, but it's around the financing of their commitments and around those deals, those circular deals that are being done with NVIDIA, making people nervous.

SPEAKER_01Yeah, and again, back to bubble territory, right? And that's a prime example. And again, this is Oracle. This is not Webman or Pet.com, this is Oracle.

SPEAKER_02Yeah, sorry.

SPEAKER_01A lot of people is think it's doing a strategic overextension right now through that deal. Last time we were on the podcast, they just announced a$300 billion deal with OpenAI, and the stock is rising by more than 40%. Meanwhile, it came down and keeps falling because some people did the math, right? So$300 billion open AI deal. Oracle needs to buy the GPUs from Nvidia unless they go after Google GPUs. Nevertheless, they need to spend money on that.

SPEAKER_01And meanwhile, in in Nvidia itself, invests$100 billion in OpenAI. That's what I call circular investment. But some people did the math and now questioning OpenAI is losing$12 billion, roughly$12 billion a year as we speak. And then based on this deal with Oracle, they need to pay$60 billion a year starting 2026. How do they do that? How will Oracle ever see any money getting out of that? And meanwhile, Oracle is investing heavily, and that cuts into their profits.

SPEAKER_01So people see a very, very strategic overextension on the side of Oracle and are very concerned. And hence you have these bubble discussions all over the place. And this is just one example.

SPEAKER_02Yeah, and OpenAI, and maybe we'll s put some uh meat on the bone here. OpenAI, so they built they built their chat GPT thanks to investment from Microsoft. Essentially, Microsoft saying, come use our data center technology. And and they I think they took a 10% stake in the um in OpenAI at that point. Open AI was meant to be not for profit. Then you had the whole all of a sudden it's changing to be a little bit look like look like it's for profit when it was meant to be it was meant to be open AI.

SPEAKER_02And so that whole change happened, and then it's almost like open AI is then does these deals everywhere. I can't believe the loss figure. Yeah, the running at that loss is huge, and on the one side of open AI, they've got Google, and on the other, they've got XAI, you know, these two colossus companies, or both of which would like to see the demise of OpenAI.

SPEAKER_02How does that play out for Sam Oldman? Microsoft don't seem to be stepping in to buy open AI, and Mark, you know, Microsoft are going out of business. So do OpenAI kind of have a Microsoft there also taking a look at them?

SPEAKER_01Yeah, yeah, and on top of that, we have open source, right? None of these examples that we mentioned is open source, right now. Great point. Chinese market, Deep Sea just overtook Gemini 3 again, right after Gemini 3 was announced by Google. All open source. This will be an interesting space. So, so where open on the one hand you have the open source world, on the other hand, you have the proprietary uh Silicon Valley Giants. We just had an executive order from our President Trump that there's no AI regulation anywhere to be seen, not even states themselves in the US are allowed now to do any AI regulations.

SPEAKER_01I mean, it's still it's still wild, wild west in a sense, out there. And um I don't know who picks the winners and who the winners will be next year in the coming years. It will be very, very interesting to watch. But getting the money back for investors for what they invested in in the likes of Anthropic or OpenAI or X-AI, many people don't see, don't see that path to return on investment. And on top of that, a very interesting article from the ex-CEO of Intel, Pat Gelsinger, saying, forget GPUs.

SPEAKER_01This is holding the whole AI space back. We need, and it's coming, the the breakthrough on quantum computing, and that will change the entire AI industry again and will democratize access to training in inference, and you don't need the GPUs anymore. So we'll see what we'll see what that space is is holding. Because a good friend of mine works in the quantum computing space, and they are making massive progress um as we speak in that space as well.

SPEAKER_02Yeah, that's uh I I did not know that about I knew of quantum, but I didn't realize that the progress was as advanced as it sounds like it is. It so going into 2026, it feels like there's going to be there's gonna be something seismic happen at some point in 2026. I don't think the um the appetite for AI is gonna go away. The uh in my research, I was able to find 5,000 AI agents, so I'm not sure how that works to be a smaller software provider of AI agents.

SPEAKER_02You're gonna be a lot of pressure on you to get businesses to adopt those AI agents, but the big players here, they still just seem to be in sort of gearing up mode, and as you said, you then you've got these big technology changes, and they're all gonna want to adopt those as well. So it feels uh I'm I'm super bullish, but I can totally understand the bubble fears.

SPEAKER_01Yeah. Well, and the full and the fallout in in in other terms, right? Layoffs everywhere, and some CEOs, unfortunately, also Mark Benioff from Salesforce saying, Because of AI, I need less hands. And he got criticized for using that term, I need less hands, laying off 10,000 people because AI and agendic AI will will take over, in his in his view. That's the fallout that's already coming and will continue in 2026. And that's what a lot of people fear now is that that it makes it so much harder, specifically for entry-level people, to compete and and pretty much find a job in the IT industry because AI might take over.

SPEAKER_02Yeah, that's that's really um a big concern, isn't it? And there's no sign of any legislation, there's no still no plan. We were having this conversation a year ago. There still isn't a plan for how that seismic shift is going to be gonna be handled. It's almost like it's the transition from this world to the next world, you know, the next AI agent world. But the AI agents haven't had the uptake.

SPEAKER_02I can I can totally understand how Amazon and Salesforce and those colossal companies, I can totally understand how they are able to see the automation out of roles. And they've been trying to get those roles automated out for a long time. It's nothing new to them. They've been looking at that all for as long as they've probably existed. But for a sort of regular business, it seems to have reduced the confidence in recruitment.

unknownYeah.

SPEAKER_02So if they're like, well, hang on a minute, isn't this AI thing gonna do it? Yeah, pull those job requisitions, let's just not move too quickly. It's almost like uh is pulling the job market back before the AI agents are really advanced enough to genuinely be able to remove the need for as many people as that in your business as there are today.

SPEAKER_01Yeah. Yeah. And mean and yes, and in addition to that, the gap between the everything around AI and the rest of the economy is growing, right? I mean, that's a concern by itself as well. Call it bubble or not bubble, but but the fact that we are stalling, at least in the US, but I think also in Europe, on the manufacturing side in terms of economic growth, but everything goes towards AI, all the stock growth is in the AI space, all the GDP growth is in the AI space, and then you see these examples of some people who do the math and don't find the return of invest.

SPEAKER_01So this is all this will all be very interesting to watch in the next year and in the coming year. And I I can't pick winners, I don't feel in a position to pick winners on what's happening right now.

SPEAKER_02Yeah, and yeah, and we'll we'll we'll wrap up because we're we're over on time, Thomas. But I think now's the time. Use the AI tools, consider the dynamics in the market around which job role am I going to go into, which ones am I just gonna kind of give a miss. I I'm I'm also bullish, conversely, that once we get through a sort of transition period, a whole new bunch of job roles that don't exist today will be formed.

SPEAKER_02We've seen that every major change, industrial revolution, all of that sort of thing, a whole bunch of job roles with um dot com. A whole bunch of job roles suddenly emerged. But you do need to be really, really careful. How do you be careful? Use AI. Use AI, consider what it would enable you to build as a business. I have to say, as a business owner, it's making spinning up new ideas, new concepts.

SPEAKER_02It's makes spinning those types of things up so much easier. Uh, have having these AI tools. So, yeah, there we go, Thomas.

SPEAKER_01My recommendation, this is an important point, right? And we we we kept saying that, and a lot of voices keep saying that if AI takes over your job role, there's not much you can do. But if somebody else takes over your job because that person knows AI better than you, that's something you can do to protect yourself. And my recommendation for everybody, and I talk to my kids every day on this, and finally they are they are getting there is get your skills up on AI. You need to have to be versatile versus versatile and you need to be knowledgeable in the AI space.

SPEAKER_01My daughter is an architect. I told her AI will be very important in that space. My son is in the IT industry, he's going to an MIT AI course right now. That's how you that's the only way you can protect yourself, and you should protect yourself. You need to be skilled on what's going on in the AI world, no matter which industry, which job. This is absolutely critical.

SPEAKER_02Yeah, look, I think that's a positive note, note to end on there, Thomas. So thank you very much for supporting the podcast this year. Got loads of new subscribers as well. So thank you to all that all our subscribers. And um, yeah, we'll do we'll probably have a break in the podcast for next week. I can't believe it's Christmas next week. So it's probably gonna be a break for a week in the in the podcast schedule, and then next podcast will be between Christmas and New Year.

SPEAKER_02Uh, so look at look out for that. And then yeah, lots of new guests and new conversations in in the new year. So, Thomas, you have a great Christmas, uh, you and yours, and uh and a stunning new year, and I'll see you on the other side of that.

SPEAKER_01Thanks, Ben, and thanks for inviting me. And and happy holidays to you and and the audience. And um, we'll see what's what interest interesting things happen next year. And I hope for an invite back because I know a lot of my people are listening to our episodes and your episodes without me specifically, as well. So I hope we can keep the cadence going. Um, for sure.

SPEAKER_02All right, thank you very much. Let's wrap it there. All right, bye bye, bye bye.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Data Center Hate | Nvidia Q3 Earnings | Neoclouds & Midterms Backlash | Data Centers in Space | Ep59

    Nov 20, 202646 minAudio

    Data centers are at the heart of everything in AI right now and the backlash is growing. Ben Drakes is joined by regular contributor Thomas Odenwald, a senior executive with HPE and SAP and former head of North America...

    Credit the people on this episode

  2. Jensen Out!! Nvidia CEO Steps Down | Apple 10-K AI Warning | Meta T's & C's Warning | SpaceX Ep58

    Nov 13, 202638 minAudio

    Jensen Huang stepped down as Nvidia CEO about an hour and a half to two hours before this recording and Colette Kress, Nvidia's CFO, is stepping in as interim CEO. The news was not on the docket and it took up the bulk...

    Credit the people on this episode

  3. OpenAI Protest: Filmed From the Room | OpenAI | JPMorgan | Databricks | Anthropic | SpaceX Ep57

    Nov 8, 202639 minAudio

    Protesters interrupted OpenAI's kickoff keynote at Seattle AI Week and Ben Pickett caught it on his phone. He describes more than half a dozen people repeatedly disrupting the session before security and police...

    Credit the people on this episode

  4. Ep56 30% More AI Code, So Where's the Software? | Harvard | Anthropic | Barclays | Walmart

    Oct 28, 202640 minAudio

    Two Harvard researchers looked at 718 companies. When AI coding agents came in, the code went up 30%, but finished software didn't measurably move. Where did it go? Into review: the time to review the code went up by...

    Credit the people on this episode

  5. Ep55 Frontier Models Now a Commodity? | OpenAI | Anthropic | Microsoft | Apple | GrokBot | Meta Muse

    Oct 11, 202643 minAudio

    Frontier models are turning into a commodity. When one model goes off the boil, enterprises switch, and a multimillion-dollar contract signed with one provider can look foolish in six weeks. Ben Drakes and Ben Pickett...

    Credit the people on this episode