In short
UK “sovereign AI” strategy and the revolving-door between government and AI suppliers (especially Anthropic), plus whether new rules are needed; debate over data-center NIMBY protests (electricity, water, pollution, land, disclosure); and whether enterprise software incumbents (Salesforce, Workday, Oracle) are reasserting value as frontier AI model layers commoditize.
Guests
Matt Clifford (UK AI policy architect; led Frontier AI Task Force → AI Safety Institute; previously McKinsey; co-founded Code First Girls; joined Anthropic after leaving government). Jonathan (co-host; former Meta global affairs president Nick Clegg is discussed; no separate guest identity given).
Key claims
Clifford’s move may undermine UK sovereignty; AI Safety Institute was blocked from inspecting Anthropic code; Sovereign AI Fund (£500m taxpayer-backed) ties to these interests; data centers shift costs to communities; models are becoming interchangeable, so incumbents owning customer/workflow/data regain advantage.
Notable examples
Bletchley Park AI Summit; ARIA/£800m; Times reports on conflict of interest; US data-center protests/moratoriums; IEA electricity doubling estimate; Salesforce Agentforce growth; Silver Lake bid for Workday.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOThe Revolving Door of AI Policy
0:45 to 4:28
Discussion on the implications of public servants moving to tech companies, particularly in AI.
“But I think we have to ask ourselves the question, as flagged in previous episodes, is Matt Clifford joining Anthropic making a joke of the UK's sovereign AI strategy?”
Concerns Over AI Strategy Outsourcing
4:28 to 8:32
Exploration of the potential conflicts and instability in AI governance as leaders transition to private roles.
“Now, not to take too much of a victory lap, but I have to say a lot of this move was predicted on this very pod back in May.”
NIMBYism and Data Centers
8:32 to 13:00
Debate on local community impacts from data centers and the costs borne by residents versus tech companies.
“It's back to that old problem, you know, just keep the Brits where they need to be.”
Energy Demand and Infrastructure Challenges
13:00 to 14:00
Discussion on the rising energy demands of data centers and the infrastructure limitations in responding to this growth.
Power Supply Challenges
14:00 to 15:00
Understanding the limitations of current power grids to meet demand.
“So where is all this power going to come from?”
Political Implications of Energy Infrastructure
15:00 to 16:00
Exploring the political ramifications of energy supply for data centers.
“That's been bought by Microsoft, is my understanding.”
Environmental Concerns of Data Centers
16:00 to 17:00
Discussing the environmental impact of data centers on water and air quality.
“where is that cooling water going to come from?”
Local Community Reactions
17:00 to 18:00
Examining community responses to proposed data center projects.
“You know, diesel generators as backups, that's going to be more air pollution.”
Job Market and Skills Gap
18:00 to 19:00
Discussing the job market implications and skills gap in the data center industry.
“That's really the ultimate self-interest.”
Global Data Center Protests
19:00 to 21:00
Highlighting global protests against data centers and their implications.
“And so fewer NIMBYs to be worried about, perhaps, and more space.”
Show all 16 chapters
AI's Economic Dilemma
21:00 to 23:00
The balance between AI infrastructure growth and resource management.
“in Tarno, which is quite near Mumbai, and in South Africa.”
Value Shift in AI Market
23:00 to 24:00
Analyzing the changing perceptions of value in the AI market.
“You know, this fashion for AI, where we are seeing previously solid SaaS enterprise software companies completely fall out of love with their existing business and pour resources into infrastructure.”
The Role of Major Tech Companies
24:00 to 26:00
Discussing how major tech companies navigate the AI landscape.
“This one's going to run and run, I suspect.”
SaaS Market Dynamics
26:00 to 28:00
Exploring dynamics in the SaaS market amidst the AI evolution.
“But actually, to your point, these things where people run actual businesses, you know, we used to call them wet concrete because once the concrete sets, you ain't going nowhere.”
The Evolving Landscape of AI and Enterprise Software
28:00 to 36:20
Learn how AI is reshaping the enterprise software industry and the opportunities it presents.
“don't own the customer relationship, Salesforce does.”
The Evolving Landscape of AI and Enterprise Software
36:23 to 37:42
Learn how AI is reshaping the enterprise software industry and the opportunities it presents.
“Building a technology company is brutal.”
Transcript
Automatic transcript. May contain errors.0:00Welcome to The Difference Engine, the show for tech founders, investors and innovators.
0:11In this episode of The Difference Engine, we examine the apparently increasing phenomenon of highly influential public servants moving to technology suppliers. Is this acceptable? And is it time for new regulations to be put in place to protect national sovereignty? And despite the AI free-for-all, are the old guard of Salesforce, Workday and Oracle coming back stronger than ever? And are data center NIMBYs a useful protector of all our freedoms or just Luddites in the AI data center realm?
0:44We're feeling a little bit aggrieved and a little bit, what would we say, validated? Indeed we are. Let's not be smug. But I think we have to ask ourselves the question, as flagged in previous episodes, is Matt Clifford joining Anthropic making a joke of the UK's sovereign AI strategy? And wider picture, is the British state and certainly the bits that relate to AI, is it for sale? Clifford, let's face it, you know, British startups are now going to be competing against the UK's best connected AI lobbyist who was meant to be their champion against US tech dominance. All right, let's dive right in.
1:25In a single fortnight, the man who literally wrote Britain's AI strategy was forced out of his government post, or some would say moved on, because he's got a brand new sparkling, I assume seven-figure job at possibly the world's most controversial AI company, a company that in this same week did not do the courtesy of the British government and the British people of sharing its code for checking with our Safety Institute. It seems like, Jonathan, the door between Westminster and Silicon Valley isn't just revolving, it's spinning. It's rotating at high velocity. I mean, to be fair, this sort of track from government influence into private industry and other influential bodies is not new.
2:13I mean, you know, we are History Bus. Ernest Marples, Ministry for Transport and Labour Government in the mid-60s, had a family company of which he was a major shareholder that specialised in building roads. What did he actually do as a transport minister? He presided over the Beeching Report, which entailed essentially destruction of the British railway system. hmm i wonder who might benefit from that completely coincidental now let's look at mr clifford a little bit he went to uh cambridge as did um as did nick clegg i think um did a bit of political science at mit as one does if if one has the um wherewithal and look to get in and the brains let's be fair and he's clearly brainy because mckinsey a favored consultant of many a government and many large uh companies where he next rocked up and then sort of reinvented himself as a startup venture entrepreneur, although to be fair, entrepreneurs first, is backed by, you know, Reid from LinkedIn and a bunch of other folks.
3:12And then infigled himself to be the head of the UK's AI policy, the architect thereof. To be fair to the man, he did do Code First Girls. He and Alice Bentnick did go on to establish, you know, getting more women's technology careers. Maybe we should look at that in another episode, and see how well that went. And maybe that's where the kind and public side of things sort of peters out a little bit. Maybe the anthropic move, interesting name, is sort of payoff because clearly of all the things Mac has done, coding ain't one of them. No, no, no, no, no. But he did become very interested in AI from about 2023, some sort of Damascene conversion.
4:01And he did help design the UK's Frontier AI Task Force, which subsequently became the AI Safety Institute. The very ones that Anthropic did not really release its code to last week, the very same people. Which is such a pity, because he did play a leading role in preparing the UK's 2023 AI Summit at the historical home of cybersecurity and codebreaking, Bletchley Park. Indeed. Now, not to take too much of a victory lap, but I have to say a lot of this move was predicted on this very pod back in May. Please play the clip. But something else is sort of interesting about this governmental tech journey.
4:49A lot of the people who got voted in, given some very nice sinecures. and I think you're wondering about the new head of the wealth fund here, Paul, and other names. I mean, we've seen people like Poppy Gustafson go from the heights of tech into government and then disappear after a few months. The undoubtedly super talented Matt Clifford authored the government's AI opportunities action plan within six months and then left for, in inverted commas, personal reasons And another great talent, Dr. Gene Innes, who was head of the Turing AI Institute, stepped down after government told, in this case, the charity, but have a look at where the charity gets its funds from, to focus on defence research.
5:40So that sort of instability across the wider administration and quangos somewhat reinforced the sense of a government struggling to reconcile their sort of avowed technocratic ambition with actual institutional discipline. so there you have it and i think um another benefactor of a large silicon valley big tech company who was also formerly deputy prime minister or co-prime minister of this country nick clegg i think you're off to see him tonight aren't you jonathan i will be sitting and eyeballing mr clegg as of 6 30 this evening yeah great so he he in case you don't know just to fill you in was deputy prime minister and then spent seven years as met as president of global affairs.
6:28And as previously stated, there's a very interesting book about global affairs and just how important that is at Meta. He's now, I think he's a gun for hire. Got to keep himself doing something. I mean, he only trousered 20 million plus, allegedly, from those years at Meta. I've personally seen him doing some lobbying work in a private members club that shall remain nameless. Now, speaking of tall flowers, Poppy Gustafsson, the Dark Trace co-founder under Mike Lynch, RIP, was Minister for Investment for not quite a year and left after the autumn 2025 ratio. Haven't seen her reappear anywhere, but similar sort of thing, moving between, in this case, private and the private industry and the government.
7:17And then we come to the interesting case of the Sovereign AI Fund. Oh, would this be Tony rearing his ugly head again? Some would call this man a war criminal, not us. No, no, no. But his daughter-in-law, the lovely Suzanne Ashman, has recently been appointed as the managing partner of the UK government's, see the theme here, Sovereign AI Fund. That is a£500 million venture fund. And guess who puts the£500 million in, Jonathan? Can you guess? Would that be private sector money, Paul? That would be what they call taxpayer-backed. Stuff that's come out of our wallets, you mean. That stuff. Indeed.
7:54Right. Let's not worry about the wolf at the door of the Blair family house. So I guess really what we care about in this pod is categories. And the whole premise here is that we as Europeans, and in our case, Brits, really need to compete with the best in the world, right? So we need to grow our own. We need to have thought leaders. We need to take big, bold swings at AI. And my fear is we're sort of, with moves like this, selling our politicians as what the US think we are, which is just rule takers and not rule makers and certainly not pioneers. It's back to that old problem, you know, just keep the Brits where they need to be.
8:37Given that we're a little suspicious about this boys and girls club, you know, is this revolving door in any way preventable? Or is it just reality? I think it's preventable if you, as we did, correctly predict that this revolving door exists. One is speculating here. No idea if this was pre-planned or whatever. But just the stories in The Times. on the 2nd of September, Mark Selman, great writer there, wrote a story entitled Conflict of Interest Row as Chair of Government Lab Joins Anthropic. That's the 2nd of September. Five days later, research agency chief quits after taking lobbyist role at Anthropic.
9:20It ain't a good look, is it? That's what sounds to me like there was a bit of cake in eating it going on there, hoping that nobody would notice that the delicious sweet was being consumed. Yeah, and this is from the resignation note on LinkedIn. I want to ensure my role at Anthropic doesn't become a distraction from ARIA's incredible work. Just a reminder, ARIA is the£800 million commitment that the government made to AI. But later in that same resignation, and very thoughtfully written resignation note, we see the Secretary of State has asked me to stay on for a short transitory period until the 6th of November.
9:58I'm sure that's not unpaid to launch a process to select a new chair i.e they had done nothing about it had not anticipated this might be a problem but this comes in the week when um at Trump's request we read in the financial times the AI safety institute of the UK the very thing that you mentioned uh earlier on was um precluded from inspecting the latest anthropic models code so um you know we're being frozen out as a country if these guys are supposed to be helping us out it ain't looking too good um it just to me it plays back into this role as um sort of a distraction and a um you know a subordinate the point about all of this paul is is that these sorts of actions are deemed to be perfectly reasonable, you know, through precedence before.
10:52They are perfectly legal. They are, shall we say, normal business practice. The problem I have with it is it just isn't a good look, and people should think about that because it makes all sorts of questions be asked. And those are questions which we shouldn't have to ask. And we should not be suspicious of the motives of people who hold prestigious, nationally important positions suddenly resigning over some spurious issue and then suddenly magically reappearing, batting for the other side. And you'll remember a few episodes ago, we had the redoubtable Christine Kifara on, who has some very trenchant views about how Europe should be ensuring its own economic future.
11:46Because what we're seeing here is that AI strategy is being effectively outsourced to people who are absolutely on the anthropic meta and other behemoth payrolls. So I think it's time for some fairly direct conversations to be had around UK government, civil service and industry about what is actually in the UK's best interests. And I think there's a few people need to have a good look at themselves because what they're doing, even if perfectly legal, perfectly reasonable from a commercial point of view, is a very bad look indeed.
12:30Not in my backyard or NIMBYs get painted with a very bad picture, but with a presidential midterm elections coming up, this has blown up very big in the States. This is the rebellion around data centers. Right. I think there's a really basic argument here, Paul. It's that, why should local communities bear the environmental and economic costs of infrastructure whose benefits accrue largely to a very small group of global technology companies. I think that last bit is the most important piece because if you think about why should anybody bear the costs of a community that is in fact the basis of the entire insurance industry pools of risk shared amongst people most of whom the vast majority of whom will not benefit for the good of the the wider whole but i think uh you know the benefactors here not being the wider whole being a very small group of um mostly almost exclusively american big tech this is uh this is set to run yeah i mean there are many many issues to be resolved here um not least the sheer question of data center electricity consumption right so if you look at the iea it the latest estimates is that data center consumption of electricity will double from 485 terawatt hours in 2025, last year, to 950 terawatt hours by 2030, right?
14:02So where is all this power going to come from? Because most grids are incapable at the moment of supplying power demand for the existing businesses, homes, and so on. and we simply, for instance, in the UK, and it's pretty much true in the States, cannot even interconnect our new power generation sources into the grid. Yeah, so I think the power argument is possibly the one that's baked in the most truth, but there's a lot of emotion. Well, it's a political question, Paul. It's a political question. It's about who pays for those new substations, transmission lines, generation capacity. Because are we going to get another situation here where we have the state subsidizing massively rich companies and their shareholders.
14:49Right. Well, I think that is pretty much on the money here. So in the States, as you've seen, you know, is it Three Mile Island, the New York nuclear plant that nearly burnt down in the Jane Fonda movie? That's been bought by Microsoft, is my understanding. You look at Grok or X or whatever you want to call Twitter these days. And Elon is providing his own power, albeit, you know, maybe not very clean power. It's extremely unclean power. It's extremely unclean form of gas turbine generation. There you go. So the power thing is sort of uniquely UK, because I think in the US, it's a state by state thing, because that's where the country is set up.
15:31And, you know, there are pockets of people saying nay, pockets of people saying yes. Just to be clear, the The UK government is 100 % behind, at least publicly, behind as many data centers as possible. And their rationale is we don't want to be left behind in this AI race. So there's a big problem with electricity supply. There's a massive problem with water consumption. We've just been through the hottest summer in recorded history. We're seeing data centers being proposed in drought-prone areas where they'll consume millions of gallons a day. where is that cooling water going to come from? And if it's preferentially supplied to data centers, who suffers because of it?
16:18It's not just human beings suffering, it's the whole environment being remodeled. Now, just to give the counter on that, the counter from the US data center folks, and I have no means of fully checking this, is that a lot of the systems, certainly the new ones, are completely sealed. Once they've got the water that they need, they don't do all this evaporation stuff it's a sealed system they use electricity to also cool the water so the electricity consumption goes up now um there's a saying normal for norfolk we won't even go into why that is that why that is a saying but um slightly it could you be regionalist i guess you can i don't think being really pissed off by them by the have you heard the noise that actually comes out of a major data center from the cooling equipment You know, diesel generators as backups, that's going to be more air pollution.
17:06You know, there is going to be wastewater from these facilities. Where does that go? Is the water and sewage system set up to deal with that? These buildings are absolutely huge. I saw a diagram the other day of one proposed in the States, and its footprint is literally the distance between the northern suburbs of Paris and the southern suburbs of Paris. Not just the middle of the suburbs, but the end of the suburbs. These are huge. Also, massive emissions from constructions of these projects and the visual impact. If you've got one of these being built next to your house, your life savings, your life's work that's been put into property, they're going to be wiped away.
17:52Now, that last comment is sounding a bit nimby-ish, I have to be honest. Yeah, right. Not in my backyard. I'm going to be financially wiped out. That's really the ultimate self-interest. I'm going to be wiped out. Oh, and maybe I'll become dependent on the state. Yeah, there may be some compensation available, as they say. But there's 5 ,000 angry Norfolkers, if that's the right word, because the petition against this Norwich APEC scheme. There's another up in arms situation in Scotland, in the borders. There's quite a lot. There's a growing spot. And I've got to be honest, in this office alone, you've got both sides of the equation of this brewing argument.
18:32It's not really surfaced in the UK yet in the same way as it has in the States. But I feel like it's an issue that time is coming. And I have to say, the places that is both Scotland and Norfolk in those instances, and let's not forget, just down the road, we've got a massive 100 million data center for Microsoft here in Acton in West London. But these two places, Norfolk and Scotland, they're not what you call in the main high density areas. And so fewer NIMBYs to be worried about, perhaps, and more space. And per the construction point, maybe a few construction jobs. I take the point that they don't need a lot of people to run when they're up there.
19:12But those are high skilled jobs. High skilled jobs that we actually don't have the people trained for in the UK. That's really going to work, isn't it? Now, in the States, this thing's much, much hotter, right? There's 142 protests, 42 US states, 500 plus jurisdictions looking at this. And we know Americans love a lawyer, right? They love a lawyer. Lots of bans going on. Mawadi in New York is trying to pass a statewide moratorium. God bless him. Is Man's Army actually being prudent on this? Because the reality is governments are approving projects faster than they or anybody else can assess the cumulative impact.
19:54This whole thing has been driven by a money machine. And let's not face it, the people who are lobbying for these data centers and the people that will construct these data centers are some of the companies that spend the most on lobbying in the world. Would that be why they hire ex-government officials? Probably, right? That might be something to do with it. So, yeah, I totally agree. There's a lot of money running around, and we've all just got to hope that that money pyramid doesn't fall, because if it does, we're looking at some pretty bad financial situations globally. But let's be clear about this.
20:33this isn't just a US or UK phenomenon. You know, the opposition parties in Finland are really warning about strained electricity supply and the pushing up of electricity prices. In India, which, you know, a country that has rapidly industrialized and moved forward in the last few decades, you know, people are protesting against a hyperscale facility in Tarno, which is quite near Mumbai, and in South Africa. Because there's a lot of outsourcing of IT services to South Africa, you think, right, well, what's going to happen when big AI centers come in? But again, it's water, electricity, land, and frankly, inadequate disclosure.
21:26These things are being rammed through on the pretense that they're essential for you know the future well-being of the global economy yeah exactly and so let's make sure that we're very clear as we make our minds on this is a very live debate as i say i think it's coming to the uk hasn't hit here as in europe as hard as it has and it clearly is in the states and the reason for that i would say is the midterms but you know the the counter narrative to all of that just to be fair here is that you know we're all using a lot of ai and we don't want to be the guys waiting as the drip comes out of the tap for our AI prompts and our responses.
22:03We don't want to hamper our country. So it's a real dilemma in terms of economic progress, and it's going to be fascinating. I think it's also an exercise in fairness, ensuring that everybody benefits at the moment. And I think it is too tipped towards very powerful interests at the moment. should a reasonable position on this be if these are going to be part of critical future infrastructure by all means build them but make the companies that are building them and using them pay the full cost of those facilities so use genuinely additional clean energy which would be a good opportunity to drive innovation minimize water consumption which you said paul maybe some of them are doing it but i suspect those are fairly unproven systems as yet and there are many many pressures in the market that they will just want to get those data centers up and running and then worry about water consumption later they need to honestly disclose their resource use i think um and we should not as nations allow AI infrastructure, the fashion for it, because this is like a black tulip phenomenon.
23:26You know, this fashion for AI, where we are seeing previously solid SaaS enterprise software companies completely fall out of love with their existing business and pour resources into infrastructure. The danger is that they're going to crowd out individual households and absolutely other industries. So this could cause a massive distortion in the economy, which benefits nobody. Right. So I think you can hear, listeners, that we are, between ourselves, somewhat debating where this is going to land. Love to hear what you guys think. Please hit us up in the socials. This one's going to run and run, I suspect.
24:10it. It is the clichéest cliché of all to say data is the new oil. It's the sort of nonsense that politicians who know nothing about tech often trump out. But to stretch the analogy to breaking point, data might be the new oil if you own the well. And some of the people that own the well have been around in the world of enterprise software for a while. I'm thinking specifically of Salesforce. So 20 plus, yeah, sure. And this month, I believe Larry leapfrogged a bunch of people, Larry Ellison, the founder and CEO, I think, I know, probably chairman of Oracle these days to become the world's richest man.
24:50So something is going on that tells me this saspocalypse doomsterism that all of these great big enterprise software companies have had it, are just a little bit wrong. Yeah, but I think what we're trying to get into here is, is the market, or has the market been, I think it still is, valuing the intelligence layer of AI whilst undervaluing the systems that actually own the enterprise data, the workflows, the infrastructure and, you know, this is the big one, the customer relationships on which that intelligence depends. Yeah, you know, and Salesforce is very, very interesting, right? So it is back in favor with the street.
25:39It had a great Q2. Shares are up, you know, almost a quarter. Its Agent Force product, which I believe is a partnership with Anthropic, need to check that, has grown over 240%. And, you know, the SaaSpocalypse seems to be laughing it off, which, you know, great for the old guy's rule thing. But actually, to your point, these things where people run actual businesses, you know, we used to call them wet concrete because once the concrete sets, you ain't going nowhere. These things do own the data, the customer records, the financial records, everything you need to comply. there's a lot of stickiness in in what it is that they do and maybe um the sell-off of all of these of um companies was a little bit too uh hasty you know if you think about an ai stack it's data which then has infrastructure sitting on it which has models on it then applications and then ultimately the business workflows.
26:46All this frothiness in the last year or so has been about the model layer, right? So when we talk about open AI, anthropic, et cetera, this is simply the model layer. Now, it should be pretty obvious to anybody who's been vaguely sentient for the last few months and uses AI themselves because they'll recognize this in their own behavior, that the models are becoming interchangeable commodities. And what do we know about commoditization? It drives down value. And more of evidence of that is from Workday. A lot of people forgot about Workday. It wasn't the most original idea. Let's put it that way.
27:25Does anyone need another CRM? Does anyone need another system of record? Because that's what we're talking about here. Well, Reuters mentioned that Silver Lake, the massive private equity company, one of the dons of the industry, was trying to take workday private. Now, these guys, Silver Lake et al., don't take crazy gambles. They're looking for bargains. They make money on being smarter than anybody else and seeing value where others don't. So this may be a counter-cyclical bet, but the reality is that, again, the people that own the models don't own the customer relationship, Salesforce does.
28:08Oracle does. SAP does. Microsoft does. Yeah, and Workday does. And so Workday's shares jumped 30 % on the rumor. And ultimately, the board there decided to say, thanks, but no thanks. So your bargain hunter comes in, makes a somewhat lowball deal, allegedly. And then it's rejected because I think the tide is turning a little bit from SaaSpocalypse. Oh, my God, AI is going to consume all before it. into a more nuanced, sensible, mature view of where the value lies. Right. So just a very simple example of that. So we're saying Silver Lake may be doing something counterintuitive, right? And that is because the conventional argument is AI will replace Salesforce, whoever, because an AI agent, in the case of Salesforce, can build its own CRM, right?
29:02People have tried. But there's an equally plausible opposite scenario, which is AI makes, Salesforce's case, the operating system through which millions of companies deploy agents. Sure. I think let's put the category lens on this. So if some of the old school winners of yore are back in favor, what is it that they've got to do so they don't get continually depositioned by new category players coming in to say, yeah, that's legacy. That's the old way of doing things. I've got something new. And one of the things that I've noticed picking up, certainly even from clients, you'll see a lot of professional services companies talking now about not just data sovereignty, but AI sovereignty.
29:48And this is a reinvention that these huge systems of records will have to go through. But it opens up all sorts of funky new category plays around AI sovereignty and data sovereignty. And that, I think, is where these large companies will move to. Yeah, I mean, if you just think about the way that AI and traditional enterprise software is going to have to interact, we've got loads, layers and layers of new category opportunities. So if you just think about it, so if you're trying to move to an agentic company, say you've got Salesforce, right? So the AI agent needs to access the company's underlying data.
30:33You're going to need an application for that. It needs permission to act in itself. It needs to know which employee is allowed to approve something. it needs to update the crm needs to create the invoice it needs to record what it did it needs an audit trail right so if you think about this it's the enterprise software layer that's becoming the control plane for ai ai is just an engine you need the operators to really understand how to use that engine and in terms of the category language that's being used we hear a lot of this phrase harness, which to me sounds very similar to a group of companies that did very well, but never really captured the imagination of the investment.
Read the full transcript
31:18A lot of them ended up in private equity hands. We used to call this sort of stuff middleware, which was a dreadful category. Who the hell wants to be mid anything? But these harnesses, this middleware, this mitigation, this AI accountancy layer, and I mean accountability, not in the financial sense, the accountability of AI, that is a rich vein. And I totally expect to see somebody or maybe a couple of people having a swing at that layer. You've got a few contenders, Snowflake potentially and others that could really do something here. And I'm really fascinated now that all the hype from the Saspocalypse and everyone's dead and some of the old guard coming back.
31:59It's just setting it up beautifully for some new categories to be created. That can be done around the traditional enterprise software arena. I just want to return a little bit to your point about how the markets are starting to re-evaluate these companies. It's worthwhile just going back to some fundamentals here. So let's call them frontier AI companies, people who are doing the big LLMs and the infrastructure. They may get extraordinary revenue growth, as we're seeing with the big LLM players, but they're also consuming enormous amounts of capital. Now, this is a bet like no other. So, you know, you've got GPUs, data centers, electricity, researchers, model training, inference, safety, sales, and, of course, the great big subsidies to acquire users at this stage.
33:03I don't think that's even the – that's not even the worrying bit. The worrying bit is the derivatives, right? It's the swapsy deals. It's the access to GPUs. I mean, this is the bit that worries me, not the fundamentals. Like, if you're actually getting a chip, fair play, even if you're paying over the odds. But if you're on a derivative play and selling space or doing some back deals to get access further up the list, that's when it gets murky. Right. And let's not get away from the fact that, you know, if you use the term frontier, because there's enormous technological uncertainty behind this.
33:36And what the market is starting to say is, or what it has said, actually, is that it was putting its bet on a particular model company. And I've already said in this episode that the models are becoming commoditized because they believe that that particular model is going to capture an enormous proportion of the future value of intelligence. Right? That's a very big assumption. And history says that's not normally the way it goes down. You might get one or two leaders taking maybe up to 80%, but there's plenty to play for. If you compare that to established enterprise software companies who have shown historically their ability to innovate and change the nature of their companies, they've already got the customers, the recurring revenue, they've got data, they've got the distribution, they've got the switching costs, which you'll never forget, it's difficult to change in any market.
34:32and a load of cash flow. So just look at Salesforce in what, 2026? And when they reported their financial year, they had$41.5 billion of revenue and$15 billion of operating cash flow. Those are huge numbers. And its AI database business is growing rapidly. So they are benefiting from it without having to do this massive spending on the infrastructure. So they're just sitting there benefiting from it. So I can't remember the, you probably remember the name for the guys that go out and prospect for oil in... The 49ers. There you go. It's a better time to be sat on a big old data well than it is going out sort of like digging.
35:18So we're all set up for this. I think it's fascinating. You know, part of us is like, great, the old guys are getting it back. But there's plenty of opportunities once all the noise dies down. and it looks like the smart money has sort of, some of the smart money has already figured this out. Right, that's true. You know, history suggests, and we do look at a bit of history on this channel, as you know, that owning distribution and infrastructure can be at least as valuable as owning the underlying technology, which is Paul's point about the 49ers. You know, just think about it. Google didn't invent the internet.
35:51Microsoft didn't invent the PC. Visa doesn't manufacture anything and Amazon doesn't manufacture most of what it sells. But, you know, the economic value migrates towards whoever controls the interface between technology and customers. So don't write off the big enterprise players yet.
36:16Thank 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.
36:48Building a technology company is brutal. We know that. It's years of early mornings, late nights, working weekends, long hours, lost sleep and neglected loved ones. You keep going because of a relentless belief in what you're creating. At Hambleton Partners, our senior advisors have walked that path, experiencing at first hand the challenges that technology leaders face every day. We know selling your tech business isn't just a transaction, it's one of the most important life decisions you'll ever make. That's why we take a different approach to technology M &A. We combine deep transaction expertise with hard-won industry savvy, helping founders navigate every stage of the transaction process with empathy, insight and the confidence born of deep experience.
37:35When the time is right to explore what's next, work with advisors who understand every step of the journey. Hambleton Partners.
From the publisher
What happens when the person shaping a country’s technology strategy suddenly becomes the person building the companies it was seeking to regulate and compete with?
To explore that question, we need look no further than the curious case of Matt Clifford.
In this episode, we explore the growing movement of technology talent between government and industry, the risks it creates, and what it means for UK AI sovereignty.
Also in this episode, we’ll take on te great data centre debate and ask why the SaaS champions of old could rise up once again.
00:45 Matt Clifford AI, For Sale: The British State
12:30 The Great Data Centre Debate
24:11 SaaS and The Comeback Kids
There is more information on how to design your category on our blog
Follow us on LinkedIn:
Want to create a podcast for your business or brand? Contact Flamingo Media to make it happen.
