75. 30,000 employees gone overnight: What can we learn from Oracle’s cull?

0:00Welcome to The Difference Engine, the show for tech founders, investors and innovators.

0:09so paul what's coming up today we're swapping our beautiful green bottles of mythos for called mythos is it the end of the cyber security boom or just fantastic pr but first are 30 000 jobs simply collateral damage in the race the us ai dominance one company certainly seems to think so

0:32so um oracle is one of those companies that you uh if you didn't know you didn't know but it silently became one of the largest companies in the world uh jonathan and i've been dealing with um it its markets its competitors its categories for years something interesting happened the other day at six o 'clock in the morning 30 000 employees of oracle six o 'clock in the morning 30 000 oracle employees, fired. No warning, no manager chat, nada. Just your role is gone. And if you'd like any money from us, sign here immediately. Now, to put this into context, Oracle has just hit record profits. Say that again.

1:1030 ,000 people sacked. Oracle has just hit record profits. The stock rose on this news, 5 % up. But their CEO and all-round interesting bond villain, Larry Ellison needed$8 to$10 billion cash because this man has built up a$156 billion AI data center gamble. That's the sort of overdraft which even Jonathan would blanch at. And he's gone on to fire nearly one in five workers for the simple purpose, some would say, allegedly, to make the balance sheet with this large debt look investable. That's not innovation, guys that's financial engineering which has human collateral i'm afraid to say one also wonders you know why are we talking about this um well what about you know uk and european tech leaders can't happen here can it could it happen here well i might do anything's possible in this particular world so that 300 billion open ai deal they've done justified 50 billion dollars in debt um open And AI walked away because the chips were already obsolete, resulting in a negative$10 billion cash flow.

2:26So what would you do about it? Well, he'd fire 30 ,000 people, not because the tech replaced them, no matter what the message to Wall Street, but because Larry's motherlode just collapsed. You know, the press release said, investing in AI's future. Oh, come on. The truth, we borrowed more than we could service. Perhaps that's a lesson to certain national economies not too far from here. Anyway, sitting here in our little concrete tower in West London, and having dealt very happily with American tech firms for well over 40 years, or 70 to 80 between us, this to us, and certainly to me, looks like American tech at its worst, right?

3:14So the shareholders win. Stock was at 5 % on layoff day. Executives win. So Larry's fortune now about$200 billion. And the workers, a sort of phrase I never thought would come out of my mind. But the people that actually make the company run lose. And that's 30 ,000. Imagine 30 ,000 people. You know, that's about at least 10 English villages worth. And their careers vaporized before breakfast. What must have that felt like? What did they tell their families? Well, I know what they've been saying on LinkedIn, and it ain't good. Somebody say it's casino capitalism dressed up as disruption. The US loves to moralize about innovation, but when debt calls come home, workers are the collateral.

3:59Yeah, we are beginning to sound a little bit like fully paid up members of the union. However, I'm going to try and take the opposite side of this. So, you know, these are big bets, which a company at the age of Oracle and, dare I say, a gentleman with the venerability of Mr. Ellison would not typically make, you know, the typical wisdom is you made your money, you know, vest in peace and go and enjoy your spoils. But this is a big bold move. And yes, there was a kerfuffle with the data center, which was being built for OpenAI. And the deal is looking a little bit flaky. But the bets are down.

4:40That money has been borrowed and it needs paying back. So if you were a startup, you could suck one in five people from your business. And let's look at it, ex-sacked 80 % of their workforce and their business. So is this the US model or is this something even bigger? The end of times, thanks to AI. the US model is all about over leveraging, betting big. And unfortunately, that sometimes means firing en masse. Now over here, we don't do that. We have very much these days tighter labor laws. Some would say strong worker protection. Other would say over strong worker protections. And we do have a culture that values transparency.

5:29Like look me in the eyes when you're shooting me in the back sort of thing. the temptation is real because UK scale-ups are looking as you said at what goes on the US we're chasing AI hype we've had a lot of venture debt and we are as a country signing lots of very large cloud contracts. UK venture debt equals fixed repayments, missed covenants, lenders seize assets. Yeah I mean I guess if you've got checks and balances you can. Yeah so if you're a US and you make a single customer bet a uk reality check you know if your 50 revenue is from one hyperscaler and they pivot you're sunk and and you know that thing about email layoffs out of the blue well you know if you're based in the uk uk law and this is pretty much reflected across europe needs 90 days of consultation for you know more than 100 redundancies you know do it via email well you've been for about 2.7 billion in unfair dismissal awards there are a lot of unfair dismissal awards as you say um so the u.s treats these layoffs um rather cruelly i would say as financial instruments in the uk that's not just unethical it's sort of illegal you know that sort of answers the question why don't we let wall street efficiency creep back into british boardrooms be careful what you wish for right you know us tech in 2026 is a bubble um you know the obr which is the Office of Budget Responsibility, warns of a 26 billion tax hole if it bursts.

6:55The Bank of England says valuations are the most stretched since 2008. Are we in Europe following America off the cliff? Yeah. Or ironically, could a nice recession reset everything for us? These are interesting points all brought up from this one sort of severe act of corporate decisiveness. So what should UK leaders do instead, Paul? Well, it'd be nice if they didn't send everyone a six o 'clock in the morning pink slip. I think there's three things they could do. One is to, let's audit the capital you have before you bet. Because we know a lot of this bet was put on with, you know, the, let's just say the scenarios weren't fully planned out.

7:41There was a lot of risk in that. If you're going to bet big, you do need to look at things in terms of negative cash flow. And how many months of burn do you have if the bet goes wrong? It's just an obvious thing to think about, right? Have you tied your debt to what you hope are future outputs of AI, or are you spending so much on CapEx? And when we talk about AI CapEx, let's be really clear, we're talking about Nvidia mostly. Is your debt tied to that? And, you know, we've seen people who will not take on customers unless they can promise in a legal document they have the NVIDIA chips that they say they've got.

8:25So, you know, this is very, very risky stuff. And classically for any business, why would you let a single customer equal more than 30 % of your revenue? That's how you get in trouble. Yeah, and from sort of one soft idea, the idea of money, to another soft idea, which is the idea of people. And we would advocate that certainly in the UK, you need to build human-centric redundancy protocols. That is about giving. It is about being decent, I think, from our point of view. So give 60 to 90 days notice. Absolutely not an out-of-the-blue 6am email. you know give out placement support if if it is the case that your company structure has to change radically the least you can do is help the people that have got you there in the first place to support them as they move on and that means career coaching um you know funding networking and so on i think the other thing is can you just be honest you know we're pivoting to ai this role no longer fits versus you're obsolete yeah and i think you know if they were being completely honest What's wrong with we overhired during COVID and we need to rein it back a bit, just as a demonstration of honesty.

9:41And allegedly, I have no proof of this. I read on LinkedIn. This isn't something that may or may not be true. But allegedly, there was some discrimination against those folks who had the most amount of RSU still to vest. So those guys who's had a potential liability on the company, some say, allegedly, on LinkedIn, those were the folks that it would be most beneficial for Oracle to get rid of first. And if that is true, we have no proof it's true, were that true, that would be a reason that would sort your balance sheet out faster than just letting people go. But with 60 to 90 days, as Jonathan said over here, then it's only one quarter's worth of money.

10:29Surely that's a more human way to deal with things. I think there's another issue here, which I think starts to become moral. Gosh, look at us, all high-emiting. I know. I know. How could we be talking this way? It is not only moral, but it can be an element of productivity and motivation and ultimately company success to align executive pay with the rest of the company. And of course, the Japanese have been doing this for years. So align the pay as a multiple of what your average worker takes home or your lowest paid worker, not just stock. So CEO compensation also aligned with cash flow, not just that stock price.

11:12The casino starts with the stock price, isn't it really? Of course it does. And people become much more focused on that than they do on all the things which are supposed to influence that stock price. So I think also some of the personal guarantees should be capped. You know, don't mortgage your home for company debt. I mean, really do not. And be honest with the employees about the risks, not just the investors. I mean, if you are asking your employees to shoulder risk in terms of their salary, their career, their future, their savings, and so on, be absolutely clear about the risks. In summary then, I didn't think this would land us here.

11:57We're sounding like something between some union conveners and gospel preachers. But the fact is, these layoffs, these Oracle layoffs, were not, despite what anyone says, about AI. They are about collateral they're about uh leverage they're about chips on the board they're about taking risk and ultimately perhaps paying for a bet that the people affected by the outcome of the bet weren't up for didn't bet on didn't see 34 years of their career ending this way any advice then for all european listeners weirdly um as the world appears to be tearing itself apart um you know we have we have in the uk definitely benefited from and um from from us tech and you know have really followed on its tail coats but as things have got a bit more extreme um the better view for uk tech is don't copy the us playbook you know audit your debt respect your people and build things to last not to speculate and we've talked about that in terms of where Europe should go in terms of building its business you know go into specialist niches build moats that will last for a long time and add real value because the next time a tech giant fires masses of employees it shouldn't be shocking it should be unthinkable

13:40well guys in this increasingly volatile world many of us are looking for digital sovereignty and now more than ever our national health providers those in the law and order community governments and public sector organizations want more control over their communications they want a digitally sovereign solution which can never be interrupted based on external demands no matter how powerful they are, and a service which means they are not reliant on a single vendor. Opting for open standard software means a competitive vendor ecosystem and solutions which are designed to interoperate with other organisations' infrastructure.

14:14This is where Matrix, the secure communications open standard, plays a vital role. Most European governments already use Matrix-based solutions for communications and Element is the leading provider of Matrix-based secure communication solutions. It's already working with the European Commission, NATO, the United Nations and governments in France, Germany, the Netherlands, Sweden and many others. Element provides sovereign, interoperable and end-to-end communications. If you're responsible for a nation's trusted and resilient communications, take a look at the Matrix Open Standard and Element at element.io.

14:47We've used Element ourselves and find it fast, secure and so easy to adopt. If you want digital sovereignty, you have to check out Element. See our show notes for more information and get on board. with the new secure world of communications today. So Jono, if I say the word mythos to you, you're probably imagining your toes in the sand, those golden Greek sun rays burning into your skin. Oh, a lovely green can with condensation falling down it. Beautiful. But no, we're talking about a different type of mythos. Oh God, sorry. I've woken up, my bubble has gone. Which brings me to my question. Is Claude Mythos the end of the cybersecurity boom or just fantastic PR?

15:31That's sort of interesting, isn't it? I mean, you know, is it a self-selected all-American group of vendors tipped off by the anti-Trump current AI sweetheart anthropic? Are these the winners in the cyber race? And until recently, it seemed the only fast-growing Uber category outside of AI. Now, they could be looking like they're included in the all-so-rounds. um so is the consolidation of cyber now or no no no the fact ai is tip the arms race so the bad guys is not new uh just making it more obvious you know no need to behave as a uh anthropic has it's its leadership is looking for pr kudos to build on the anti-open ai sentiments you know um wasn't sam eltman's house actually firebombed in recent days ridiculous why are people's firebombing people's houses with their babies in them that's crazy right you know This is probably just performative scaremongering to try and capture momentum, you know, tip public and lawmaker opinions into Anthropik's favor.

16:34Right. So we've got a debate here, haven't we? Like, you know, as you said, is this just the best guys in AI doing the best thing for the world because they're the bestest ever? or is it just canny PR and one company, albeit the guys that seem to be in the race at the moment, Anthropic, basically creating the scenario that they win the new category of post-AI cybersecurity or whatever they call it. So one could say that what Anthropic has done with Claude Mythos, and in fact the broader positioning around Claude as a security-aware AI system, I see a new version out recently. What they're saying here is it's not just about incremental improvement as the cyber arms race sort of has been.

17:18This is a moment in the sand. It implies a shift from all types of cybersecurity tooling, which is fragmented. And that is the message that people like CrowdStrike have been saying for years. And that's maybe why they made the cut for the Anthropic special attention people alongside Google, Microsoft, Palo Alto and a few others. It may be that this incremental shift to consolidate to platforms has now been superseded by a one-time shift that this Mythos product and the project that goes alongside it has now pointed out. And this threatens the promise that has kept lots of cyber vendors growing for the last few years, that the pie is growing and that the complexity requires many specialized tools.

18:07And if you're a big player in this market, that all of the specialist tools you need are available on my platform and not somebody else's. Right. So let's break all this down and see what this scenario means. So I think the first thing, if we do subscribe to the idea that suddenly cybersecurity is now part of AI and all those cybersecurity firms can just disappear, what we're seeing here is a self -selected winner's circle. Bully for them. Yeah, well, the optics matter on this. If the ecosystem is being highlighted as largely US vendors, already well capitalized, closely aligned with frontier AI labs like Anthropic, then it starts to look less like an open market and more like a curated inner circle.

18:54Yeah, what you said earlier, this group was effectively tipped off or at least aligned early, reinforces the idea that the next phase of cybersecurity isn't being competed for. It's already pre-allocated. Now, that has to be toxic for a long tail of vendors. If buyers believe the real future is concentrated amongst a handful of AI line players, everybody else immediately looks second tier. They do. And let's not forget what Peter Thiel's advice is here. He hates competition in markets, hates it. Be different, but be in a market of one. So this is looking a little bit like a self-selected winner.

19:35So I do agree. Now, the second way to look at this from what Anthropik's trying to do here is trying to achieve what they call narrative capture. And it's using its friends in the media and experts and of course a lot of those um media friends do uh back anthropic and it's anti-trump let's say anti-defense department or department of war whatever the hell you want to call it uh starts uh and so on bbc news night we've got these so-called experts um forgive me they are largely academics jonathan um so you know talking about what where this takes us all But if the experts only seek to validate this AI-led consolidation thesis, what's the point of that?

20:20Because that'll just frame these vendors as the inevitable winners as appointed by one player in the market. And as you say, it will kill innovation and stop new cybersecurity categories breaking out. because independent or at least non-US ecosystem, bearing in mind all of these guys are US, will be excluded. That's when you get narrative lock-in. And that's the point. It's not just technology driving consolidation anymore. It's perception. And in enterprise tech markets, we can both attest perceptions move budgets faster than reality. Oh, yeah, yeah, yeah, right. And I think the third thing we need to think of in this scenario is that AI does indeed collapse the category.

21:04So if you think about cybersecurity's growth, it's depended on ever-expanding threat surfaces, tool sprawl, you know, from EDR to SOAR, et cetera, and the need for human analysts to stitch it all together. But if AI systems like Claude unify detection, reasoning and response, you know, reduce the need for multiple vendors and abstract away complexity, then cybersecurity stops being a tool chain market and becomes a platform market right now that's the inflection point when booms end when dozens of categories collapse into three to five dominant platforms yeah and and yeah another possibility the fourth possibility if you will from all of this is those not included now look like also runs and that is um you know it almost sounds sounds like a famous world leader that we know that never denies and always believes their own story.

22:01What you see there, and this is classic pattern of hyper growth and consolidation, is explosive growth at phase one, hundreds of vendors, lots of categories, second phase, platform emergence, and some sort of consolidation. Might be a few vendors, could be one. And then finally, and this is with your investment hat on, the capital flight to those perceived category winners. Now, if buyers and investors start believing that, then only these AI-native, anthropic-approved, model-integrated security players matter, and everyone else fades out or gets acquired. And that means, unfortunately, funding dries up for all the rest of us, and that alone can end a boom.

22:43Yeah, regardless of actual demand, I suspect. And then, of course, we've got this sort of nastier thing, which has come in as technology has matured, and that's the sort of political and ideological undertones on this. So bringing in the idea of Anthropic as the current AI sweetheart of anti-Trump circles adds a more controversial, but may I say it, rhetorically more powerful layer. So it sort of suggests alignment between political media and tech elites, which we've seen assembling. And it implies a gatekeeping of which companies are legitimized. And it ultimately frames the outcome as orchestrated rather than in some way organic.

23:27Now, even if that's overstated as an idea, to me, it strengthens the argument that this is less a data-driven shift and more a coordinated narrative shaping the market. All right. So the bottom line on all of this yes case, meaning that yes, this is the end of the cybersecurity boom. What's the bottom line on this? Right. I guess if you put all this together, then the argument sort of becomes like this. So AI via players like Anthropic is collapsing cybersecurity complexity. Now, a small aligned group of vendors is being positioned as the future. Media and expert narratives are enforcing that position, and everybody else is being implicitly downgraded or exclusive.

24:13And in conclusion, then, if we believe that, yes, this is the end of the cybersecurity boom, this isn't just good PR, then it's the moment where the cybersecurity boom peaks and flips into consolidation, where growth concentrates, vendor diversity shrinks, and the category loses its status as the last end of the cybersecurity boom. It's a dramatic story, but I'm not sure it holds up well because it's a stronger case that what's actually going on here is excellent positioning by Anthropic. And this is not, in fact, a structural turning point. So let's now examine the case for Claude is not the end of the cybersecurity boom.

24:48Yeah, no, I think so. I think a stronger case, this is an excellent, this is just, frankly, excellent positioning by Anthropic, and it's not a structural turning point. I mean, to start with, this idea that AI tipped the balance to attackers sort of isn't new. The notion that AI advantages bad actors has been circulating for years. You know, we've seen things like automated phishing, malware generation, and social engineering. And they have been sort of, if this is a thing you can say, been improving steadily since early GPT era systems. And ransomware groups were already operating like efficient SaaS businesses before Claude scale models even appeared.

25:36And the reality, and we've seen this, and it's a core message from many of the security firms we've worked with, is security teams have always been outnumbered and reactive. So to this extent, Claude Mythos doesn't reveal a new reality. It makes an existing asymmetry more visible and therefore easier to communicate, right? So to me, that's a messaging win, not a market-ending shift. Speaking about messaging, I'd say also fear narratives are just a classic way to shape markets. This is a familiar FUD playbook here. So emphasize a fast-escalating threat. Oh, my God. Frame the current defenses as inadequate.

26:15What is that shit? And then position your approach as the obvious and necessary evolution. now anthropics is leaning heavily into that platform by amplifying the ai driven risk and talking about how the current crop of cyber security was built pre-ai and then that means you you know by deduction surely you need something from anthropic to fix all that it's not unusual it's just how um you know major platform shifts get take hold and it is though a classic strategic piece of PR? Again, you're much closer to strategic PR than evidence of a collapse or collapsing industry. This performative scaremongering angle, it's plausible to think this is about momentum building.

26:59I mean, Anthropic sits in a competitive triangle with players like OpenAI, where differentiation is hard. And, you know, they can lean into safety, governance, responsible deployment sort of gives it gives it a distinctive lane you know framing ai as a dangerous but manageable if done their way there's two things i mean it nudges public opinion and policymakers towards stricter controls good get them in harness and positions anthropic as the trusted vendor in that regulated future that that's not accidental i think it's narrative shaping aimed at regulatory advantage. Yeah, I agree. And I think the timing here is odd.

27:40Just after it's had its run-ins, shall we say, with the Department of War and the fact that it wants to be perceived by certain folks in the market as being on the moral high ground, which brings us on to the anti-open AI undercurrent which we live in. And again, nobody is standing up for folks that throw firebombs at folks' houses with kids and so on. It's ridiculous. But there is a bunch of turbulence. And Sam Altman's been man enough to talk about this, which would amplify the fact that this is the time when an open AI undercurrent would push you further in the market. So whether or not the specific incidents are directly relevant, there's definitely increased scrutiny on AI firms, even in good old, merry old England.

28:28This gives an opening. This cracks a little bit of a chink of light for rivals to contrast themselves and anthropic benefits from leaning into we take risks more seriously and others are moving too fast hence the you know hence the alleged stopping of um the model being proliferated that's classic competitive positioning from us from a sense of power uh but it's not evidence that cyber security as a sector is obsolete or anything like it so why doesn't this end the cyber boom then so even if ai increases hacker capability, as of course it will, it doesn't eliminate demand, it does the opposite.

29:06So more sophisticated attacks means more spending on defense from those attacks. Faster cyber, faster attack cycles means greater need for automation tools. AI generated threats means AI driven detection and response. So, you know, if you think about history, which we're very pond of doing here on the Difference Engine, every escalation in cyber offense has expanded the markets. You know, cloud, more security vendors. Mobile, more security vendors. Zero trust, more security vendors, right? So there's no reason AI would break that pattern. If anything, it just widens the attack surface and increases budget.

29:43And if I could move from history to economics for a second, I heard this today. I thought this was genius. A CEO, somebody we'll have on the pod very shortly, mentioned this. if a problem is worth solving because it's such a bad cybersecurity problem the market would solve it the reason that there are still cybersecurity holes is for good reason or bad the market has decided that stuff ain't worth fixing and if what seems it becomes worth fixing that's when there's a cybersecurity need and that's when some new innovation starts let's move on to uh the fragmentation that this is designed to cause this whole announcement and this whole framing of the haves and the have-nots in cybersecurity.

30:24So the idea that AI is going to collapse cybersecurity into a few platforms is very appealing, especially to those guys on the right side of the red rope, if you see what I mean. Because it's unrealistic, though, also because enterprises have already today got deeply embedded, heterogeneous cybersecurity sites. They ain't going out and changing those tomorrow. The regulations that have to be dealt with vary from region to region, in some cases making entire cybersecurity solutions not fit for that region. So this is not a one time in life, you know, stake in the shift in the sand, whatever the phrase I'm looking for is.