This is an AI Transcription. It’s pretty good, but please forgive any errors.
[00:00:00] Jonathan: Welcome to The Difference Engine, the show for tech founders, investors, and innovators.
So Paul, what's coming up on The Difference Engine today?
[00:00:14] Paul: We'll be waving goodbye to Software as a Service, SAS, and saying hello to Agents as a Service, ARS.
[00:00:22] Jonathan: Cheeky! We'll be taking a trip across the channel and asking what the Paris AI Summit tells us about the evolving relationship between politicians and tech.
[00:00:31] Paul: Mais maintenant, en première, but first, let's dive in! Let's take a moment to read NVIDIA's palm. Is it looking a little bit Cisco? Okay, so it's becoming increasingly clear, much as we love him, NVIDIA's future might not be as assured as it appeared just a few months ago. And it's not just about the shock arrival of our eastern friends DeepSeek and the massive pummeling, record pummeling, it got in its valuation.
The
[00:00:58] Jonathan: Great Pivoter. From, uh, GPU. Nothing to be ashamed about. Well, it claims to have invented that in 1999. Did they party? Uh, like it was then. Um, uh, so graphics delivery platform. Uh, pivoting to the, uh, an AI enabled chip and ecosystem category lever, leader. I mean, NVIDIA now faces growing challenges. that could reshape its role in the AI and IT world in the coming years.
But what do we usually do when the future looks, um, you know, a little foggy? What do we normally do then? Well, we look at our crystal ball. We look in our crystal ball and we look at what happened in the past. So, are there any lessons for NVIDIA already in existence? I feel there are. There indeed may be.
As I was riffling through the internet the other day, Uh, I found this amazing graph, uh, from New Normal Consulting. If you look at this graph, you see that the share price growth and fall of two companies is very, very similar. If you plot Cisco's 1990 to 2002 share price performance and Nvidia's 2014 to the present, the correlation is almost 100.
Percent. Yeah, it's extraordinary. Now that is scary. So we sort of ask ourselves here that um, category lit and category driven stock market bubbles Do they always follow the same path and could nvidia decline after a short period of nasdaq field expansion? Not that
[00:02:30] Paul: we're saying cisco's necessarily in decline.
That's that's been a very well managed company, albeit it's laid off more people than a small country over the
[00:02:39] Jonathan: years. Well, in the last year, it's, um, it lost money or its performance was questionable in four quarters. This quarter has done a lot better, but more of that later. Um, so let's try and have a look at why NVIDIA may or may not follow the path of Cisco and face a sudden slowdown, um, in growth for the market of its technologies.
From which it may never fully recover. And that's GPUs in NVIDIA's case, and it was internet routers in Cisco's case. So let's put this into context back in late March, 2000, at the height of the dot com boom, which we absolutely lived every second of a quarter of a century. Yeah, I know, but it, my PTSD from that period still haunts me.
I think
[00:03:22] Paul: I remember the pink slips.
[00:03:23] Jonathan: Um, yeah. So Cisco did become the world's most valuable. Company and had a capitalization of more than 500 billion Now 25 years later its market capitalization is currently less than half of that In sheer dollar terms never mind the impact of inflation
[00:03:40] Paul: and to be fair though.
It's probably shared a fair bit of Dividends in those times it has
[00:03:45] Jonathan: and you know, this is not about saying cisco is in any way a bad company This is about showing how the rationality of the markets actually act and how it affects leading tech companies So we know that cisco went from an out and out category leader or being regarded as an out and out category leader and internet routing To a collapse in market belief of endless growth and demand for that technology.
It then had to chase one category leadership opportunity after another in the hope of regaining its form of valuation. And I think I've
[00:04:16] Paul: told you guys before that, um, Cisco did interview me for a job in their consumer division. Uh, when I asked what that was, it was Webex, the Palm device, if you remember that and some other nonsense.
Uh, and I literally said, are you joking?
[00:04:29] Jonathan: Yeah. Didn't get the job, funnily enough. Despite some serious missteps in a. Cisco had great leadership and it still continued to prosper as it built out infrastructure, frankly in virtually any country on the planet you could name. But it never again regained a valuation based on those early market assumptions about future growth.
Right. And so even now it's following the trends, it's talking about empowering service providers to reimagine networks for AI. Um, and claims that it's Q2, uh, and it's latest Q2 results have just come out, that it's AI infrastructure orders exceeded 350 million in the quarter, bringing that total to roughly, 700 million for the first half of the year.
So go Cisco
[00:05:14] Paul: really seem to add up to me. So if they're doing seven 50 million of revenue and the market caps, did you say 500 million, something like
[00:05:22] Jonathan: that? That's a, it's a very complex relationship. I think what we're, what we're trying to say here. Is that they are surfing the latest wave and making sure everybody knows about it?
So but we do know that in that the revenues were declining quarter by quarter last year And there was layoff as you said after layoff after layoff, but they have pulled in some better numbers in the last quarter But that's tech for you. So go back to nvidia for a moment, you know the genius of jason Dominating the airspace is that he's done it Like having a golf club it you know He's done it with a tool singularly ill designed for the purpose and vidya had been we all remember this the go to company for GPUs and PC gaming My performance computing, but GPUs were not designed with artificial intelligence in mind.
The clues in the name, graphical processing unit. Yeah. So the GPUs were able to take over those AI workloads simply because there was no other viable alternative. It would have been
[00:06:21] Paul: better if they call them general, general purpose,
[00:06:23] Jonathan: general purpose. Yeah. But there are, they did have an ability to handle large volumes of data that made them effective, despite the fact they were not optimized.
And in the absence of effective competition, Jason and his company certainly made hay while the sun shone.
[00:06:40] Paul: There's no doubt about it. This is a company whose brand has accelerated to the front of the forefront of everybody. Even those who don't know anything about the tech industry, they know who NVIDIA are.
[00:06:50] Jonathan: Don't underestimate this genius. This massive category pivot delivered first mover advantage and ecosystem dominance. It's currently the de facto standard in AI infrastructure with CUDA, which is their Pleat unified device architecture, I think
[00:07:05] Paul: that is the most,
[00:07:07] Jonathan: but that in itself has been around for 2006.
So if anybody thinks that Jason woke up one morning and went right, we're now an AI company, they're deluded that strategy had been going strike after strike after strike design the category. Good. And guess what? CUDA is deeply integrated with popular machine learning frameworks like, I guess, TensorFlow, PyTorch, and that gives NVIDIA a strong boat that competitors are struggling to bridge.
It's got 800 architecture, the H100 architecture, and the And there's an upcoming Blackwell architecture. So NVIDIA has secured a dominant position in AI cloud computing and, you know, is dominant in data centers. OpenAI, Meta, Microsoft rely heavily on those, uh, NVIDIA chips for training and deploying large language models.
[00:07:54] Paul: And I think, um, you know, just an example of that, we were speaking recently to a company, um, selling into the AI data center space, uh, and they said, Unless you can prove they will not even take you on as a customer, unless you can prove you have your allotment, not even got them received, but they want to see written proof you've got your allotment of NVIDIA chips.
That's a lot of power.
[00:08:15] Jonathan: Yeah, and you know, even if alternatives to NVIDIA's GPUs do emerge, moving away from that NVIDIA ecosystem creates significant cost, effort, uh, time barriers.
[00:08:26] Paul: In the case I just mentioned, they're going to say, Awfully nice of you to take an interest in our, in our storage, uh, infrastructure product.
You're, you're not a serious customer because you can't show me that NVIDIA is going to ship chips to you. It's amazing. And
[00:08:39] Jonathan: that's absolutely true. And even if those GPUs aren't really up too much, the history of tech is replete with examples of ultimately second rate technologies having first class market share.
And again, NVIDIA has built a powerful brand that may be assuring to corporate boards trying to navigate rapid and risky. Change.
[00:08:57] Paul: Yeah. And I think there's a point here. We often see this in category. Um, categories led by engineering type folks, the people going, my stuff's got better engineered. It's just, it's got better feeds and speeds.
It doesn't matter if you win their hearts and minds of boards. And if you have sales guys saying, if you haven't got an Nvidia allotment, We're not selling to you that permeates up. That's that's category. And
[00:09:18] Jonathan: it's true. And, you know, and videos is nothing if not an engineering led company, even though it's doing some very, very clever marketing.
It's got a 30, 000 30, 000 strong workforce, and they're mostly engineers. Those people can build. application specific libraries. So within that, there's countless specialists built to build even the most obscure application.
[00:09:42] Paul: It certainly has a momentum now. Yeah, that's for sure.
[00:09:44] Jonathan: And it's talking to thousands of customers, um, about what they're going to need in future.
So that's a powerful feedback loop. Um, and again, this sort of discussion is going to reassure any corporate board who care much more about. What something does than what it's built on. And they care deeply whether they think the company is going to be around in the future. Again, a massively strong barrier to startup competition.
[00:10:08] Paul: So just a sidebar here. So, um, obviously you, they say all boats rise. On a rising tide, um, and there is a confusion, uh, between, uh, the large LLM players, uh, you know, the meta open eyes a slash Microsoft amongst the forefront, but also Anthropic, Mistral, Coheir, blah, blah, blah, blah. Um, I was speaking to the CEO of a very successful mid tier.
BI business intelligence company this week. And they, he serves, his business serves the largest banks you can imagine as customers. He's originally a coder. And so perhaps not unnaturally for him, he spent his Christmas holidays, believe this or not, writing his own transformer conclusion. It ain't that hard.
And he said about building this transformer, which is what the competitive mode of the LLM companies is. Let's remember, he said, anyone could do it. Now that's just a humble boast. Yep. Everyone. could not do that. However, the point is that competitive moat that they think they've got boasting about how much money they're spending on the latest AI project and how much money you need to compete with them.
That bubble sort of got a bit burst with deep seek the way he sees it. That's not, it's not as hard as they make out to build transformers.
[00:11:21] Jonathan: This is almost certainly true. And there are some other issues which NVIDIA should be worried about. We will cover those in a moment. Even that revolutionary potential of DeepSea's apparent efficiency.
The basic demand for AI compute power is continuing to grow. Yeah, so
[00:11:36] Paul: he's not talking about the people that get confused with NVIDIA, the other people in the so called AI hype cycle. He's distinguishing between the guys who got something, that's NVIDIA, and the guys who got what, in his view, Nothing.
[00:11:51] Jonathan: Which is a very interesting position for NVIDIA to be in. Yeah, great. Great position for them to be in. But we know that the push towards real time generative AI and enterprise agent based applications could, could keep the NVIDIA SuperTanker revenue growth going for quite a while. And lastly, the other thing which people tend to forget about NVIDIA, it isn't just a GPU company anymore.
It's expanding into cloud computing via, uh, NVIDIA DGX cloud networking through, was it Mellanox, which they acquired as an acquisition? I think I've
[00:12:24] Paul: heard Jensen say that was the most underrated acquisition in terms of, um, the market recognizing it, that he knows of.
[00:12:31] Jonathan: I'm not quite so sure about NVIDIA drive though, and I'm much less convinced about the mass application of autonomous vehicles, as I think, frankly, they're a solution looking for the problem, but everybody makes mistakes.
So. Those diversified revenue streams could provide some shelter against potential GP market downturn, although there are formidable incumbent competitors in both cloud and networking to be fought. On the cloudy side of things, there is a potential quadruple whammy. For NVIDIA to deal with now financial market over expectations.
We've talked about But there is a paradigmatic shift in processor technology and distribute Distributing computing models going on there are new AI environments appearing and there's some rather weird political and regulatory uncertainty that needs to be navigated at the moment, so Could we be seeing the limits of Jason Huang's amazing and highly lucrative pivot?
I would say, you know, as AI demands have grown, and this is happening on a day to day basis, so have the inefficiencies of relying on GPUs for underlying infrastructure. Uh, not least the energy demands of what will, what will. Undoubtedly power the next industrial revolution. So building AI infrastructure with GPUs comes with some significant downsides.
Uh, power consumption and cooling. GPUs by any standard are massive energy guzzlers that require elaborate liquid cooling and data centers. And they're both expensive and resource heavy, you know. Already some major companies are believed to be rethinking their investments in GPUs because of issues like overheating.
Uh, and we're here, we're here with some even pausing buying NVIDIA's Blackwell
[00:14:21] Paul: GPU. Will that allow more people to get onto the waiting list and move up the waiting list? I'm sure that's going to be fine.
[00:14:26] Jonathan: It sort of is, but I'm sort of thinking about the movement from steam engines to diesel electric traction on the railways.
You know, NVIDIA has established the whole infrastructure as a standard, the basic way of moving all that data about. GPUs are possibly their steam engines. They do a great job, but they're very, very wasteful. So we already know that GPUs are going to be up against neural processing units and they are promising to.
revolutionize the space by slashing energy consumption. People are talking about more than 90%. Uh, the same apparently is true of, uh, Google's tensor processing units, and they're already deployed in Google, Google cloud. Um, and they're also particularly designed for deep learning. Um, we also know that NPUs have the potential to outperform GPUs and specialized AI computational ability.
Um. Imprints tasks are much better, and this, to me, indicates a shift in the industry. This could reduce GPE use and reduce it to a historical stopgap.
[00:15:28] Paul: Well, um, is it time for a newsflash then?
[00:15:30] Jonathan: Oh, go on then. What's happening? So,
[00:15:32] Paul: um, the newsflash that, that, that many of us have been waiting for is, uh, comes from the East.
Darling of the, of the, of the British tech. Uh, industry arm, which is in a lot of these data centers you're talking about already has made the announcement. It's about to launch his own chip. And I quote the financial times here that arm to launch a chip in a move that could up end the semiconductor industry.
[00:15:55] Jonathan: That's, that's quite a claim. And what is this claim based?
[00:15:58] Paul: A new chip that arms producing. And let's remember who owns. Uh, this talk about category leaders, um, master son, uh, is, is the guy he's legendary, um, soft bank, soft vision fund. And allegedly this arm chip already has its first customer, which is somebody you just mentioned earlier.
Meta. Not a bad first customer,
[00:16:19] Jonathan: not bad. And we'll see what happens in then. So, I mean, this just basically showed us that the market is still very much, much in flux. So the rise of, and from our point of view, the rise of the MPUs and the TPUs will create opportunity for new category creation because cost structures are going to be revolutionized and industries are going to demand more optimized solutions.
We also know. Apart from this, you know, latest news, the major players still not to be discounted Intel, AMD, Qualcomm, we're all investing in NPUs and that could challenge NVIDIA's stranglehold on the AI chip market. Um, and there are other. Developments, which threaten the primacy of GPUs, AI accelerators, you know, you've got the Apple M series, Tesla's dojo, Amazon's Tranium, um, and the building of custom AI processes optimized for workloads.
We've got field programmable gate arrays, FPGAs, and of course, good old ASICs, and they provide specialist AI acceleration with greater efficiency in some applications.
[00:17:22] Paul: So once the magic wears off a little bit. You do tend to see that. And I've seen lots of pictures. I'm sure we all have on the internet of the latest Apple M, um, mini max, just stacked up like a tiny little data center on your desk running deep seek.
I mean, absolutely. What is that? But other than an analogy for a bunch of, um, Nvidia chips in a data center, uh, and open AI. And it's, it's, it's. Pretty amazing.
[00:17:50] Jonathan: Yeah. So, so still on that, that sort of hardware orientated and end of the ecosystem, um, you know, you've got those, um, you know, ironically NVIDIA's own digits, desktop supercomputer, you know, reducing reliance on the highly managed centralized GPU clusters.
And thirsty, as you said. Yeah. And then you've got other things like edge distributed AI models that will make a difference. And there's also. The evolving nature of artificial intelligence itself, and specifically the push towards generalized large language models, um, it's becoming clearly less practical and more costly, and that, as we've referred to a number of times, has been underlined by the arrival of the much less hungry.
AI tools, such as, of course, you just mentioned it, DeepSea. So
[00:18:39] Paul: I think these are prime times for category creation. These are the times that if you're a category designer, you absolutely pray for. They don't come around very often. Really is the time when we can figure out, you know, what can we learn from history and how do we take it forward?
Yeah. So when you look back at all of these factors in combination, NVIDIA's future valuation journey could well, as you said at the start, Professor Simnett, resemble. The valuation of Cisco, the curves could be up and down in the same way.
[00:19:08] Jonathan: Cisco dominated the networking market for the years of the dot com boom.
Um, before facing what was a black swan. Anybody who knew the history would know that the dot com bubble would end at some point. The issue is nobody ever wants to believe that and nobody knows when it's going to quite happen. You're saying that's what's happening now. It is potentially what is happening now.
Um, you know, as those new tech innovations and technologies. Which were forced to emerge because of the collapse of the dot com boom, Cisco struggled to adapt. Um, and it's category leading position weakened over time, as did it's market capitalization, never ever hit those earlier highs. Now, NVIDIA, I think has possibly learned from the past, it's well aware of the threats to it's dominance.
It's
[00:19:51] Paul: pivoted massively once.
[00:19:53] Jonathan: Yeah, it's already, uh, it's already adapting by investing in NPUs, expanding its AI software ecosystem and being pretty aggressive with its M& A opportunities. It's pivoted successfully before, but the question remains, can it continue innovating fast enough in the current rapidly shifting market to defend its category leadership position and critically its valuation?
We know it could invest in all the other technologies everybody's. Coming in, but whatever happens, decline won't, won't come from the impact of a single and what I heard called the other day, yank, which I love yet another NVIDIA killer chip developer arm, by the way, um, but from NVIDIA, increasing having to fight on several fronts in an ever more complex competitive environment.
And that means new category creators could take away that overall category advantage. They currently enjoy now. I wouldn't bet against Jason, but then again, neither would I bet against history. All right, let's get some popcorn.
You know what? This really grinds my gears.
[00:21:04] Paul: What's grinding your gears this week? Freshly back from Paris, I missed out on, uh, the grandiose Paris AI Action Summit. Um, and it seems, uh, and certainly in some quarters, I didn't miss that much. Um, Bloomberg, never really known for being that edgy, described the Paris AI Summit in a sort of Valentine Day massacre assessment as, and I quote, here's the headline, Paris summits.
Pinky promises are not enough. I don't know what they mean by pinky, but I'm not really going to explore it. I
[00:21:36] Jonathan: think, I think our friends across the pond might be able to explain to
[00:21:39] Paul: us. Possibly. And then, um, this is the subhead. This is, this is where it gets really bitchy. Yeah. The world needed. Concrete commitments on safety instead of an investment pitch from France.
[00:21:51] Jonathan: Wow. Sounds more like the concrete is on the boots on this one. I guess
[00:21:55] Paul: when an American owned, uh, famously by the guy that was the mayor of New York owned a media outlet says that, um, you know, we don't need an investment pitch from Europe. The knives are a little bit out there.
[00:22:07] Jonathan: But is there something out there where the category designers could move in?
[00:22:12] Paul: Yeah, I think so. I mean, um, it looks, you can always look out for change when, good or bad, uh, certain things happen. Is it the taxi driver? The taxi driver is a classic. Another bad sign is when you see people, and I guess if you've got a few miles on the clock and you look at LinkedIn, you know who the A players you've and then the B players and when the B players start.
Boasting about hiring B players because they, the famous, um, line, uh, that I recall is a players, high A players, B players, high C players,
[00:22:42] Jonathan: but frankly, there are some fatuous titles starting to appear on LinkedIn at
[00:22:46] Paul: the moment. Ridiculous. Let's go through them. So if I say to you, growth hacker. What do you say?
Uh,
[00:22:52] Jonathan: probably some, somebody doing low level marketing, demand generation or something.
[00:22:56] Paul: Yeah, yeah, yeah. It doesn't sound as good, does it, really, as growth hacking? But it's the reality, right? Yeah. One man's growth hacker, another man's marketer. Uh, how about prompt engineer?
[00:23:06] Jonathan: That would be somebody who's probably just.
The bout done the first thing on chat GPT.
[00:23:14] Paul: Chat GPT. Yeah. I just call it an AI user. I love the fact that people were saying we're all going to be prompt engineers. Nah, just AI users. Yep. And then, um, another favorite of mine and it normally has an M in the middle, but sometimes has an F in the middle.
Yeah. Fractional CXO. What's a fractional CXO in your mind?
[00:23:30] Jonathan: Probably a jack of all trades without a job currently.
[00:23:33] Paul: There you go. Open to work. Yeah, that's it. So, but I think the worst. The worst of all, and that's what the, the, uh, AI, um, action summit reminds me of is when the politicians get involved. That's a real low point.
[00:23:45] Jonathan: Yeah. I mean, it's sort of difficult to tell the difference between big tech and politics at the moment anyway, isn't it? It is, but I like my tech tech and I like
[00:23:51] Paul: my politics. Not so
[00:23:52] Jonathan: much purity. We like purity.
[00:23:55] Paul: Let's keep it. Let's keep it real. So the latest insanity, the thing that's driving me absolutely crazy.
And let's be honest, it's tried to hush, hush, been hushed up a little bit is the UK government. Saying to Apple, and let's remember the UK is the home of the so called online safety bill. Oh yeah. Um, which in its very wording cannot technically achieve what it wants to do. But the UK government, it seems now by the back door, and I'll just use this word advisedly and repeatedly.
Allegedly. Allegedly. Allegedly. Allegedly. In secret. Talks, uh, and using legal injunctions, allegedly the UK government has realized how badly the online safety bill is going to miss online safety act, I should say, and they just figured out it would see how encryption works. Oh, so it's trying to ask Apple, the world's largest, most profitable company ever, if it would bind awfully, if it would just backdoor.
all of our messages. So, uh, the BBC has, um, in its ever so nice way, Dear Zoe has said, Apple has previously said, um, it would pull all encryption services like its security service from the UK rather than yield to, uh, comply with government demands. However, allegedly, this is all subject to an injunction.
So we're not even supposed to know That this is all about. I mean, does that sound like big brother to you or I
[00:25:17] Jonathan: don't know? I seem to remember the UK government getting rather exercised about potential backdoors in huawei's equipment Not that long ago,
[00:25:24] Paul: which is the thorn of a horn of a dilemma here.
Like which one do we go for? Yeah so that's the the um, bbc being ever so careful and ever so politically correct saying that they are They politely asked, and, um, let's not ask too many questions. Now let's look at the other side of the pond. Let's see what we get from USA Today. USA Today, uh, calls out Britain, and it says, Britain's order is a slippery slope that could lead to other countries demanding the same sort of access to U.
S. encrypted storage services. Right, so now it's, now it's less of like, The government and privacy and saving kiddies. Now it's more like a trade war. It's just getting messy. They go on to say it's a very concerning assertion of extra territorial power. What other countries might also demand that access?
Who knows? I mean, right. So do we really want politicians? And tech mixing up to this extent.
[00:26:19] Jonathan: Well, you know, tech, tech has now become a powerful, uh, instrument of policy. Yeah. Uh, as much as it's a powerful instrument of economic growth. Yeah. So this collision has been a long time coming, but it's well and truly here now.
[00:26:32] Paul: Yeah. I, I just don't know if we can ever get back to the stuff we like because it all is becoming a bit entangled up. Um, they've got their noses in the trough. They think they know what they're doing. God help us. And just as one final point that, um, Some of us who have watched the witnesses for a while are finding it hard to disentangle the two.
The self styled, uh, UK's oldest IT analyst. That'll be Mr. Hallway. I speak of Richard Hallway, MBE. MBE, yes. He of TechMarketView fame, he's the founder. And in comments with another industry veteran, Mr. Stephen Kelly.
[00:27:08] Jonathan: Strangely, also appears to be in charge of the body that recommends whether people get MBEs or not.
[00:27:15] Paul: Is that right? It is true. You have me at a disadvantage, sir. I did not know that. As far as I know, Stephen, uh, is the, um, is the CEO of Serata at this point in time. Also has had
[00:27:26] Jonathan: jobs with the government. Former CEO of Sage, former CEO of Micro Focus.
[00:27:29] Paul: And so these two had an interesting, uh, dialogue on LinkedIn.
Uh, where, um, Stephen was talking about the, um, state of politics and the latest, uh, taxation moves, the job tax that's, uh, come here in the UK and, um, prefaced all of that as one would with, I don't want to really talk about politics because he's a tech guy, Richard. The UK's oldest IT analyst said, come off it a little bit, he said, Stephen,
[00:27:57] Jonathan: I think he waited straight in, straight in,
[00:27:59] Paul: and Stephen says, you say, I'm not making any political point.
But at some point, you have to stand up and be counted. He then goes on to say, Richard, I did not vote for the current government. And I think it's been a disaster, blah, blah, blah, blah, blah, blah. We're both staunchly against politics and tech colliding, but I don't know, it seems like it's, it's going to happen.
What's really
[00:28:19] Jonathan: chilling is the complete and utter lack of tech experience of people who are now presuming to know better. Yeah. Trying to put particular rails in, whereas the tech industry has created its own rails in the past and, and, and the politicians were quite happy to let that happen.
[00:28:35] Paul: I mean, I'm quite confident that, that, and that, you know, you should take advice from you and I on tech matters.
I'm not sure. You want us doing open heart surgery or flying your plane home from your holidays. So if there's any way at all, we can disentangle this, that would be nice. That would be great.
[00:28:56] Jonathan: So in an earlier episode, we discussed the launch of NVIDIA's Digit's desktop supercomputer and much as we were impressed with the potential impact of NVIDIA's 3, 000 box, it was a fact that the company was planning to release several AI Software tools that really caught our eye Now the release of these tools are a signal of a move that could dramatically reshape the enterprise software landscape It's more than just a trend It's a paradigm shift that has a potential to change everything and drive a new era in enterprise tech and exciting And if not a little scary, think about how this could go.
It's all about opportunity.
[00:29:34] Paul: Yeah, and it looks as though there's a lot of, um, scaredy cats in the world of SAS. We think they should be not terrified, but a little bit concerned.
[00:29:43] Jonathan: Yeah, as we know that the timing of new category design and deployment is always crucial. And it seems we're at the point where software as a service, or SaaS, as we have known it and loved it for many years, is nearing the peak of its growth curve.
As we've mentioned before, SaaS is ripe for reinvention, both to climb the next S curve of innovation and to unlock fresh revenue streams.
[00:30:06] Paul: What we're witnessing is a changing of the guard, a start of an S curve. Something that really will shake things up from a tech point of view.
[00:30:15] Jonathan: So, I mean, in fact, those of you have been reading the press in, in recent weeks, we'll know that Microsoft CEO, Satya Nadella has hinted himself that the era of traditional SAS is coming to an end, giving way to something new.
Yeah, and
[00:30:30] Paul: not only him, Jason Lemkin, uh, famously of the SASTRA. Um, event and that's the man that
[00:30:35] Jonathan: has to do a pivot.
[00:30:37] Paul: Yeah. Um, I think he knows it though. In fairness to Jason, he's a canny guy. Yeah. Um, so is it overrated? Is it overhyped? Are we really at something new? Do you think?
[00:30:48] Jonathan: Maybe I think we should say welcome to the age of agents as a service.
Oh, now, now, now, now given. Given SAS was pronounced SASS, this could give us a problem. To take our Q from the same acronym for the American Association of the Advancement of Science, that is actually shortened to triple A,
[00:31:09] Paul: S. Well, that saves them the problem, doesn't it? Because for us, for it to be ours, we need an R, which we don't have in as a service.
But there
[00:31:17] Jonathan: we go. But you know, in joining application analytics or even accounting as a service in the world of IT, should we pronounce it? A A A S, because if we go A A A S, that would leave us all, us Brits, sounding like a bunch of Liverpudlians. Haha, indeed. Anyway, I think we need to make our minds up, because if we are to believe the propaganda coming out of the Action Summit, then the, uh, move to Agentech was front and centre.
[00:31:42] Paul: Yeah. And, you know, is that another something that's getting a bit hyped?
[00:31:47] Jonathan: I think we are careering towards an inflection point. You know, the future of enterprise software might and I think will look dramatically different from what we used to. You know, instead of selling software, the next generation of companies will sell.
Intelligent agents in this new landscape AI will become the central driving force of the enterprise
[00:32:08] Paul: will it disappear as a thing? Well, I think it slowly
[00:32:11] Jonathan: will because we're seeing a new driving force in the enterprise acting as a catalyst for Potentially Unprecedented levels of automation.
[00:32:21] Paul: Yeah, and does this also move the owner of the software away further still from IT?
Because the drivers, the people who are buying these agentic software products are likely to be line of business.
[00:32:35] Jonathan: Yeah, but of course, you know, the IT is going to have to integrate. These things, I mean, let's not forget this is a technology which has to be integrated within any particular business or organization, but the key thing about this is the movement from the automation of production, whether that's been a production of real goods or the production of knowledge to the automation of decisions.
The world of automation, as we know it, is shifting. You know, it's no longer about business management being the ones guiding the automation process, sort of what you were referring to. It's the automation itself that's going to get automated. All right. So in this sort of strange new world, the idea is that the AI agents will seamlessly integrate with business processes, so they will handle the complex tasks that we're all used to with minimal human intervention.
[00:33:26] Paul: Right. And that does sound like to me, at least, less of a role for your traditional IT sort of mechanics.
[00:33:33] Jonathan: Yeah, I think it is. And I think there's a, there's an analogy perhaps coming up on that. So here's what I think we think is coming. You know, those traditional software interfaces, which we've grown accustomed to, you know, those dashboards full of complex UIs, endless menus, and require hours of training to use effectively, will fade into the background.
What are we going to get in their place? Conversations with our AI agents, you know, a Nirvana where instead of laboring through 15 spreadsheets to generate a quarterly forecast, we'll just simply tell the AI agent generate Q4 forecast and it'll happen.
[00:34:08] Paul: But I think there's also a nuance in here, right? So, um, you and I are quite prolific users of AI, um, maybe not the chatbots, but certainly all of the LLMs that are out there.
The beauty for me of that is, especially if you're an inquiring mind and you've got some sense of logic, is you can ask a question on a question. Right? So, whereas the spreadsheets and the dashboards, et cetera, you get in, you get stuck at a certain point as a non technical person, and then you're screwed.
You know, it is a help desk. It's a, it's a follow up question, but with something that you can interpret the answers from. Like a chatbot? Yeah. You can ask question after question, and I guess that's what agen X's doing. It's taking out the need to go back to, to mummy and ask the questions. You just figure it out as you go along.
It's really enabling in that sense.
[00:34:55] Jonathan: It is. I mean, I mean, essentially that whole business logic is being compressed. So those entire workflows, when you've just described a number of workflows there, um, that used to require bespoke code, well, should. In the future be as simple as issuing a command. Now that whole complexity in the background, which we've, you know, we've had to manage less of as IT has developed, but it's still there, will, will be hidden from view.
Yeah, and those agents won't be confined to working with a single database or system, which we point them at. They'll operate right across the tech stacks, integrating one would think effortlessly, if we're to believe the hype. And work in harmony with the existing infrastructure. Because
[00:35:40] Paul: I'm asking my questions my way, I get personalized answers to me.
Now, this is anything like this, which is the technology moving on, you know, massively is a gift for a decent category designer.
[00:35:51] Jonathan: It is just like that. You know, what we're saying here is like. Enterprise computing, Alexa, you know, people have to, uh, grab the idea of category creation in, in an environment where once you used to spend time navigating and managing convoluted software, and frankly, I think some of these Irritating AI, uh, co pilots could just disappear.
Most of which
[00:36:19] Paul: are useless, I'm looking at you, Gemini. Um,
[00:36:21] Jonathan: you know, we can devote ourselves to more strategic endeavors, perhaps improving our businesses or delivering a more efficient service. Right. Or who knows? I mean, who knows? Even enjoying all that endless machine age leisure time. We were repeatedly promised.
[00:36:36] Paul: That's never coming. And by the way, you'll spend your leisure time playing on AI games. So yeah, but
[00:36:40] Jonathan: the point from a category creation point of view, this is not just about adding AI features to existing software suites, which we've seen ad infinitum over the last two or three years. This is a complete re imagining of enterprise software.
So. You know, we have got this true paradigm shift going on at the moment and it should be as transformative and probably be more so than the connectivity we got from the rise of the internet, the flexibility of cloud computing, of course, triple A S. Wouldn't be possible without the maturation of both of those.
Um, and also, it certainly won't signal an end to privacy and security concerns. They may only get worse. Definitely, 100%. Yeah, but don't think either that this is a wave of back to ground zero creative destruction. Yeah, there's a reason Satya is a bit excited about this. And I think that's because The big players, the Microsofts, the Googles are perfectly positioned to own the entire stack.
Well, they'd certainly have you think that, wouldn't they? They would, but they could from, you know, going from enabling software infrastructure to general enterprise applications. I think they, because of their distribution, are likely to dominate general purpose AI agents and that will serve. The basic needs of businesses and organizations across the, across the globe.
It's sort of much like they currently dominate desktop productivity software.
[00:38:09] Paul: Yeah, that was a scrappy fight. Um, so I, I predict a scrappy fight for those guys. Um, not least of which because of the deep seat moment.
[00:38:16] Jonathan: Yeah. Well, I think while they're having that, having that fight, there's a critical point that many people might overlook.
Right. So, so while they're fighting there for dominance in that horizontal AI space, the real opportunity. Lies in the niches. That's where the real innovation take place, takes place and where the new category creation will unfold
[00:38:37] Paul: for sure. And I, you know, I, for one, I'm all for that. Um, but if you think about it, if you're the person in charge of a business, it might not seem too exciting.
To just do a bit of narrow AI. I think it's terribly exciting. And I think the mistake is to go the other way and try and do something general. As you say, the big guys got that nailed.
[00:38:55] Jonathan: Yeah, you're dead right. Because I mean, you know, let's face it, as people who run businesses, you know, in the, in the world of.
Enterprise software as users. We don't actually want a generic software that sort of understands our industry or our particular tasks. And we want agents that deeply understand the specific businesses and our particular workflows and our particular pain points. And if you again, going back in history.
IBM, who was the big generalist supplier at the time, didn't build the ERP category. SAP did. We need to remember that. Um, so we want agents that can help us gain a competitive edge, crease our margins, and just have that direct impact on the bottom line. It's not about. It's not just basic improvements of existing processes.
It's about transformation of industries.
[00:39:53] Paul: Ironically, the AI summits in France and they call it action, but we think the action might be elsewhere.
[00:39:59] Jonathan: Yes, I think the action is elsewhere. The action is going to be on where that category opportunity lies. So while those big players are focused on enabling general purpose AI for that.
Mass adopting audience, huge opportunity for smaller, nimbler entrance to carve out niches and build specialized AI solutions tailored towards particular industries or particular business needs. Um, you know, I'm convinced that those who are quick to understand this shift and move to meet it, have the opportunity to build strong competitive modes long before the incumbents evolve their business models to catch up.
Um, But I think there is a reality that the world of enterprise software, as we know, it is rapidly evolving. Um, SAS, SAS, while still relevant, might soon be a relic of the past. The future is undoubtedly intelligent agents, automation, and specialized AI that truly understand specific use cases. So, the question I think is, for listeners, will your company be ready to create the new categories in this rapidly evolving world?
Approaching AAAS tech era.
Thank you for listening. If you want to learn more about category design, head to becategorical. com. If you need help designing and dominating your category, then get in touch. Contact details are in the show notes.