Latest / AI in the UK / UK Datacentre Use Surges; Two UK AI Firms Secure Huge Funding Rounds
Transcript
- 0:00All right, your luck. Listen up.
- 0:16Hello and welcome to AI in the UK.
- 0:18I'm Neil Footman. And I'm Matt Wilkinson.
- 0:21Professionally, we help organisations, leaders and
- 0:24individuals learn about and implement this big scary
- 0:27behemoth called AI. There's some amazing stuff going
- 0:32on with AI in the UK that doesn't always get the attention
- 0:36it deserves. With AI in the UK, we want to
- 0:39change that. 5.9% That is how much of
- 0:57Britain's national electricity is now flowing into data
- 1:00centres. The government's own published
- 1:03figure was just 2.5. This week on AI in the UK, we're
- 1:07looking at why the gap between the official number and the real
- 1:10number matters to British AI right now.
- 1:14We've also got a London drug discovery raise that puts $2.1
- 1:17billion behind one of the country's most watched AI
- 1:21companies, with the UK Sovereign AI Fund quietly at the table.
- 1:26And Multiverse has hit a 2.1 billion valuation on the bet
- 1:30that AI is a workforce expansion story, not a workforce
- 1:35replacement story. I'm Matt Neils is with me, and
- 1:38we're starting while the lights are still on.
- 1:41Neils, over to you and the data centre numbers.
- 1:44OK. Thanks, Matt.
- 1:45Yes, as you said, the first story this week is subject you
- 1:49may be familiar with. That's data centres.
- 1:53So UK data centres, it is reported, now consume 5.9% of
- 1:58national electricity as infrastructure warnings mount.
- 2:02So a new report from the International Data Centre
- 2:05Authority, an industry body tracking global digital
- 2:08infrastructure, puts UK data centres at 5.8% of national
- 2:13electricity consumption. That figure matters because the
- 2:17IDCA identifies 5% as the threshold at which significant
- 2:21community resistance and government regulation typically
- 2:24follow. The UK has crossed it.
- 2:28The government's own earlier estimate, published earlier this
- 2:31year, put the figure at 2.5%. The gap between those two
- 2:35numbers is not a rounding error. It reflects how quickly demand
- 2:39is accelerated beyond official projections.
- 2:42It lands at a politically awkward moment.
- 2:45But let's be honest, when have we not been in a politically
- 2:47awkward moment in the last 12 months?
- 2:50So Keir Starmer has made data centre expansion a pillar of his
- 2:53growth agenda, designating the sector as critical national
- 2:57infrastructure and backing AI growth zones across the country.
- 3:02Energy Secretary Ed Miliband has described the climate impact of
- 3:05data centre expansion as inherently uncertain, and the
- 3:10Environmental Audit Committee has now launched parliamentary
- 3:13inquiry into data centre after finding that the government's
- 3:177th carbon budget failed to isolate or adequately model the
- 3:21explosive growth in driven compute demand.
- 3:24The public picture complicates this.
- 3:26Survey data from earlier this year suggests national support
- 3:30for new data centres sits at around 69, drops sharply to 56%
- 3:35when people asked about developed developments in their
- 3:38own area, with environmental concerns the most common reason
- 3:42for opposition. Separately, a Tech UK survey
- 3:45found that only percent of the public say they know much about
- 3:49data centres and 27% have never heard of them at all.
- 3:54The government is asking the public to accept a significant
- 3:57shift in national energy use around infrastructure.
- 3:59Most people cannot yet name or context.
- 4:03Germany, the Netherlands and Switzerland have already
- 4:05exceeded 9% and all three are now experiencing the regulatory
- 4:08and grid constraints, constraints that tend to follow.
- 4:12The window for getting through ahead of that is open.
- 4:15The question is whether anyone in government is looking through
- 4:18it. So I think there was, yes, we
- 4:21quoted a couple of figures there and 5.95 point 8%.
- 4:24Anyway, very close to 6% I think we can safely say, which is well
- 4:28ahead of what the projections were.
- 4:31So what? Well that perhaps that's a good
- 4:33place to start. What do you make of this figure
- 4:34this 2.5 to 5.8 Matt? Yeah, I mean, this isn't a
- 4:38failure of policy. It really seems like it's a
- 4:41failure of forecasting. Yeah.
- 4:44It seems hard to sort of comprehend how people couldn't
- 4:50see this the, the sort of the, the generative AI wave arriving
- 4:55as fast as it has, you know, over the last few years and the
- 4:58need for energy to be so great. It's just hard to hard to
- 5:04understand how we can get that projection.
- 5:06So, so incorrect from I think I saw somewhere that wait lists
- 5:11are now something like 15 years long to, you know, for an
- 5:15average connection of, you know, industrial connections to the
- 5:19grid, which is just crazy that we're, you know, that we seem to
- 5:21have an, you know, energy infrastructure that just is not
- 5:25built to cope. In in any sector?
- 5:27Is that what you read? Yeah, yeah.
- 5:29And actually that we're, we're pushing AI data centres to the
- 5:32front of the queue. So it's not 15 years for a data
- 5:35centre. They're actually getting bumped
- 5:36to the the head of the queue. But it's about 15 years to, you
- 5:40know, wait list to get to get on the National Grid for something,
- 5:43you know, for some, for some businesses, which is just
- 5:46shocking. I, I guess as well, we should
- 5:48never underestimate the government's ability or, or any
- 5:51institution's ability to undershoot on forecasts when
- 5:55there is a political reason to do so.
- 5:57When you see public sector projects frequently and let's,
- 6:02you know, let's not mention HS two though, I just did that.
- 6:05When, when there is a very strong political imperative to
- 6:08have the forecast push in a certain direction, then that
- 6:10will very much affect the assumptions that the forecast is
- 6:13based on. This is a huge difference as you
- 6:16said, and it, it sort of tends to suggest either, and of course
- 6:22we should begin by saying this is changing all the time,
- 6:25changing so fast. It's a very difficult thing to
- 6:27get your head around and predict for anybody, you know, never
- 6:29mind a relatively thoughtfully paced civil service or whoever's
- 6:34making this particular prediction.
- 6:35But it suggests either either a, an organisation that doesn't
- 6:38fully understand what they're dealing with or something, some
- 6:43something like in just a complete operational misreading
- 6:47of of what the situation is. So yeah, I'd agree on that.
- 6:50I think that it's it, isn't it? A smaller gap than that could be
- 6:55put down to some kind of forecasting error that this
- 6:58suggests something more fundamental is at play.
- 7:00But what what about the the different positions being
- 7:04adopted by Starmer and Miliband? What?
- 7:06What would you say about that? Yes.
- 7:08So that's an interesting 1, isn't it, that, you know, AI has
- 7:13a, has a really complicated, it's going to have a really
- 7:15complicated impact on, on countries and, you know, and,
- 7:19and the world in general. And, and a very big impact is
- 7:24going to be on our net 0 obligations and, and the drive
- 7:27to get our carbon emissions down, while at the same time
- 7:31being an attractive place to have higher, you know, energy
- 7:35intensive info, you know, infrastructure, whether that is
- 7:39data centres or even, you know, the ability to manufacture
- 7:42steel. And, and so I think that we've
- 7:45got ourselves in a, in a real issue where, you know, long term
- 7:49planning has failed to, to really prioritise the building
- 7:53of an energy infrastructure that looks, you know, that has a mix
- 7:57that is, you know, maybe includes more nuclear, includes
- 8:01more wind, more, more solar power, you know, solar power.
- 8:05And, and really looks to look at, to see how can we actually
- 8:08make sure these, you know, these are fully adopted.
- 8:11I mean, it, it beggars belief that we're still putting up new
- 8:14houses and we're not enforcing that, you know, new house build,
- 8:18should, you know, where appropriate, should have solar
- 8:21panels, battery storage and all the sorts of things that
- 8:25distribute some of that load around the, the, the, the
- 8:28network anyway. And so I, I, I do sort of feel
- 8:32that we've got a bit of a backward looking approach here.
- 8:36But I think that, you know, when you look at the, the predictions
- 8:39versus, you know, the, you know, sort of the forecast versus
- 8:43what's actually going on, that feels to me that we didn't want
- 8:47to really admit that this was going to have such an impact on
- 8:50the grid. One, because that would mean
- 8:52we'd need to invest more in the grid and secondly, would put out
- 8:55those net 0 obligations. So I think it's been a little
- 8:57bit of a case of, you know, looking between the fingers and
- 9:00hoping that it doesn't get as bad as it potentially is.
- 9:03But if we've got a big wait for, for, for, you know, connections
- 9:06to the grid, the challenge is, is that even if our grid isn't
- 9:10perfect in terms of net zero, if it's better than the
- 9:14alternatives, then that's got to be a good thing.
- 9:17And so I, I, I do feel that there is a, a real challenge
- 9:21there as to, as to how we perceive the rights and wrongs
- 9:26of energy intensive industry. I think yes.
- 9:29So I think as well that so on, on this point specifically about
- 9:33what is apparently a bit of a rift in policy between Starmer
- 9:36and Miliband, governments always have this and any organisation
- 9:40again, and particularly governments, they have various
- 9:43absolutely critical priorities they have to tend to.
- 9:45And when one of them is the economy and the other is the
- 9:48environment, this is something that leads inevitably leads to
- 9:52friction. That said, I think we're in a
- 9:55peculiar position at the moment and that we have a somewhat
- 9:57wounded Prime Minister who's position, very position, is in
- 10:01doubt. And someone in Ed Miliband, you
- 10:03have someone who's kind of lifelong devotee of net zero,
- 10:07He's very strong, supportive of that.
- 10:10It's difficult to see how you square this circle, I think, but
- 10:15from the research I've read, are you familiar with Andy Masley?
- 10:18Have you heard that? Have you come across that name?
- 10:21I believe so, yeah. So he's done quite extensive
- 10:24research into the environmental impact of data centres and he's
- 10:29been extremely thorough. And he, I think I've mentioned
- 10:31this before, but he got into he, he took issue with some figures
- 10:35quoted by a journalist called Karen Howe in her or whole Howe
- 10:39Howe in her book Empire of AI. And she, to her credit,
- 10:45acknowledged it that she made a mistake that essentially she'd
- 10:48overestimated, overstated that the water use of of many data
- 10:52centres by a factor of something like 10.
- 10:54So a massive amount, unfortunately, so that the, the
- 10:59original narrative had already become very widespread.
- 11:01And it's, it's proved very stubborn.
- 11:03And I think people are inherently suspicious of AI
- 11:06anyway. And a lot of these big
- 11:07companies, and they'll be inclined to believe that the the
- 11:09worst possible narrative about them.
- 11:12Yeah, I, I agree and I, I, I saw that.
- 11:15I mean, Empire of AI is a, is a fascinating and quite scary book
- 11:20to read. I mean, I think it's brilliant,
- 11:21a real eye opener. Yes, there are challenges around
- 11:24water usage. Yes, there are challenges around
- 11:27electricity usage. I think some of the shocking
- 11:29things is that when the grid isn't prepared, organisations
- 11:33will go and find out whatever they can to generate power.
- 11:37And I believe that for some of Musk's data centre, Elon Musk's
- 11:41data centres in the US, he was actually may even be be using
- 11:45jet turbines to, to turn, you know, gasoline, as they would
- 11:49call it, into, into energy. And so, you know, that's
- 11:52incredibly inefficient. And so there are some real
- 11:55concerns about how do we do this.
- 11:57And we know that across the US organisations are, you know,
- 12:00turning back on, are looking to turn back on nuclear power
- 12:04plants. I believe Two Mile Island is is
- 12:06coming back online. You know, there are other
- 12:09nuclear power plants where the entirety of the of the power is
- 12:12being bought by, you know, AI data centres.
- 12:15And so I think that there is a future where, you know, we can't
- 12:18just look at data centres, you know, of their own, you know, on
- 12:21their own. We also actually have to look at
- 12:23do we need to Co locate power production as well?
- 12:26And that that's maybe a question for, you know, for bigger minds
- 12:29than ours. But I do think that there's an
- 12:31issue there. Yeah, as a quick final point,
- 12:33this this whole thing about the public opinion, I think it's
- 12:36going to become increasingly important and would imagine
- 12:38hostility towards data centres is going to increase as it
- 12:42becomes more more political issue.
- 12:45I was surprised at the level of support in this.
- 12:49It was not the most rigorous of studies, but saying that support
- 12:52sits at around 69% was surprised it's that high.
- 12:55Would think people would instinctively feel opposed to
- 12:59these things given the bad reputation they've they've
- 13:02generated so far, but what was perhaps as informative is saying
- 13:0627% have never heard of them at all.
- 13:09Imagine for for most people this will remain quite esoteric issue
- 13:13and building any kind of national consensus around this
- 13:16is going to be very difficult. Yeah, it's one of those
- 13:21interesting ones where you see, you know, a lot of people are
- 13:23for it, but then you see a 13 point drop in support when
- 13:26people are asked about whether they want them on their
- 13:28doorstep. So whether that's wind farms,
- 13:31roads, housing, data centres. And we've seen in the US there
- 13:35is a huge backlash about the creation of data centres.
- 13:38And I think it's twofold. 1 is that there is this perception of
- 13:42the data centres stealing the, you know, pushing up the price
- 13:45of electricity, which it can do, you know, making water more
- 13:50scarce. And, and so those are real
- 13:53concerns for local, you know, developments.
- 13:56But I think that then there's then also what we've been seeing
- 14:00recently in the news where there is a general, there is a, there
- 14:03is the start of a pushback against AI in general with a
- 14:06number of, you know, these commencement speeches that they
- 14:09have in, in the US where a number of people are being booed
- 14:12when they start mentioning artificial intelligence because
- 14:15graduates are feeling at risk. You know, they've invested a lot
- 14:19of money in themselves and education.
- 14:21And then are they going to be replaced by an AI?
- 14:24And so I think that there is going to be this very delicate
- 14:26line to walk between. How do we how do we navigate
- 14:30between, you know, what was the, the old, you know, the old way
- 14:33of doing things pre AI and, and the, you know, the new world AI
- 14:38order, if you will. And, you know, hopefully it
- 14:40doesn't turn into something that is as draconian as a New World
- 14:43Order. Indeed, on that we can certainly
- 14:45agree. So moving on to our second story
- 14:49this week. So this concerns a big funding
- 14:54round and a couple of names that we'll be very familiar with, I
- 14:57imagine. Matt, over to you for that one.
- 14:59Yeah. So this is this is really
- 15:00interesting, isn't it? And actually it speaks to sort
- 15:03of potentially trying to negate some of that negativity.
- 15:07So you and Blair, Tony Blair's son has founded a an up sort of
- 15:13an upskilling platform called Multiverse.
- 15:15And just this last week the organisation has raised $70
- 15:18million around 55,000,000 lbs, putting it as a $2.1 billion
- 15:25valuation. That's it's first funding round
- 15:28since 2022. Schroeder's Capital lead the
- 15:31round, with existing investors including General Catalyst,
- 15:34Lightspeed and Index Ventures participating.
- 15:38The company has been repositioning itself from an Ed
- 15:41tech apprenticeship business into what it calls Europe's AI
- 15:44adoption platform, pairing the announcement with its Berlin
- 15:48acquisition of Stack Fuel in January and 50% year on year
- 15:51revenue growth for the third consecutive year.
- 15:54As part of the equity round, all employees have been offered
- 15:57equity in the business. Now, bearing in mind the you
- 16:01know the you know who the the founders father is, the
- 16:05political layer does become a little bit more interesting.
- 16:08Rachel Reeves, the Chancellor of the Exchequer, has publicly
- 16:11endorsed the deal, framing Multi First as proof of the
- 16:15government's ambition for the UK to achieve the fastest AI
- 16:19adoption rate in the G7. Schroeder's capital is using the
- 16:22same language about workforce capability as the productivity
- 16:26unlock, but underneath the optics is a narrow, narrower
- 16:30commercial reality. A four year valuation gap, a 24%
- 16:35markup during the largest AI capital cycle in history, losses
- 16:40of 63.3 million in the year to March 2025 and a single cash
- 16:45positive quarter that lands the week of the raise.
- 16:48The Department for Education figures show multiverses
- 16:51programme completion rate at about 52.6%.
- 16:54And that, and that's, that's not bad, but that's against a sector
- 16:58average of 65.4%. So it's clear that there's some
- 17:01work to do in making sure that those programmes, you know, are
- 17:06engaging, that they have good trainers on board.
- 17:08What makes this an AI story rather than they take one is, is
- 17:11the thesis Multiverse is selling the inverse of the Klana trade.
- 17:15You know, Klana itself decided that it was going to replace
- 17:19most of its development team with, with AI bots.
- 17:21And, you know, Klana has since walked back some of those.
- 17:24You know, some of it's more aggressive AI replacement
- 17:27claims. But where Klarna and a growing
- 17:30cohort of CFOs treat AI as headcount reduction, Multiverse
- 17:34is selling employers on AI as workforce expansion, where
- 17:39deployment value is determined by how how well existing staff
- 17:43can run the tools. The $70 million bet is a bet
- 17:47that enterprise buyers will write more cheques for the
- 17:49second view than the first over the next 12 to 36 months.
- 17:53And that feels like there's a real gap between AI tool
- 17:56procurement and workforce capability at enterprise scale.
- 17:59And we've seen this and you and I worked with this, where just
- 18:03because somebody's bought a load of Copilot licence doesn't mean
- 18:06to say that people are using them effectively, that people
- 18:09are really experimenting. And, and that's something that's
- 18:11really, you know, I think it is a big unlock once people start
- 18:15to realise how they can use these tools, You know,
- 18:18Schroeder's is not paying $2.1 billion for an apprenticeship
- 18:21business. It's, it's paying it for this,
- 18:23the capability unlock. And I think that's really
- 18:26interesting. What do you make of this?
- 18:28Very eye-catching. That is a very high valuation,
- 18:32$2.1 billion and it would certainly evaluation that's you
- 18:36think well beyond the average achieved by a regular education
- 18:40business, quote unquote regular education business.
- 18:44I have heard, I've heard good things about Multiverse and it
- 18:47seems to be a positive business story.
- 18:50I think you would need to pay attention to the kind of clients
- 18:53that they're bringing in and the clients have reported companies
- 18:56like Addison Lee, Babcock, Capita, you and I are, are very
- 19:01strong believers in, in the importance promoting proper AI
- 19:05literacy. And if Multiverse can help
- 19:07achieve that, then that is fantastic.
- 19:10What do you think, Matt? Yeah.
- 19:12So I think it's, it's a, it's a bit of a double edged sword
- 19:14really. You know, you start, start
- 19:15looking at some of the numbers, you know, $400 million markup
- 19:19with a cash positive quarter and 50% year on year growth across 3
- 19:22consecutive years. You know, starts to look what
- 19:26disciplined company building looks like.
- 19:28You know, and in a sector where you know, most UK growth
- 19:31companies have taken down rounds, everybody that all the
- 19:35existing investors, you know, participating around and that
- 19:38shows real confidence. But the benchmark for a UKI
- 19:41champion is whether you can build a compounding revenue
- 19:44bait, you know, in business inside enterprise AI, you know,
- 19:47multiverse is now the clearest UK example of that path.
- 19:50So it it it looks positive from at least that lens.
- 19:54It does, it does. I think this sort of feeds into
- 19:56my earlier point a little bit though that a 24% uplift over 4
- 20:02years. So in certain sectors that would
- 20:04be quite attractive in if it was a.
- 20:07Proper AI company an actual tech company that would be considered
- 20:11quite disappointing some of the valuations these companies are
- 20:15going for are perhaps excessive, but you would be looking at
- 20:19annual growth well in excess of 24%.
- 20:22I think that so part of this funding round was something of a
- 20:26move away from apprenticeships and towards corporate training.
- 20:30So Multiverse perhaps needed that kind of rebrand to clear
- 20:34this round. And it's worth also bearing in
- 20:36mind that the they've made significant losses last year in
- 20:39excess of 60 million and then the headcount actually slightly
- 20:42fell. You can't read too much into
- 20:44this. This happens with growing
- 20:46companies. But the IT it does suggest
- 20:49certainly that the rebranding has led to something of a change
- 20:53in how its growth metric is being defined.
- 20:56Yeah. And it's interesting, I've been
- 20:57doing a lot of work with with the US company called Neuro
- 21:01Collective and you know they've sort of looking at the adoption
- 21:05engineering level. So rather than, you know,
- 21:08training the workforce, looking at what does adoption
- 21:11engineering really look like across the organisation?
- 21:13So sort of taking that to the level above.
- 21:15And when you start looking at valuation and the, you know, the
- 21:18conversations that people are having, it does look like multi
- 21:22verses being valued more as an Ed tech company rather than as a
- 21:26real true play AI company. And I think that's that's a
- 21:30really interesting one. And the other thing that's that
- 21:33strikes me here is that, you know, the work that you do, some
- 21:37of the work that I do, you know, in training and you know,
- 21:40building confidence in teams to be able to, to use AI, that
- 21:45there's not much of A Moat there.
- 21:46There's not much to protect that part of a business.
- 21:49So the real, you know, the real indicators of success here are
- 21:54going to be stickiness of customers and what is the
- 21:57revenue unlock that the people that go through the programme
- 22:00really have. And that's where I think that
- 22:01it's going to become a a really interesting time for these
- 22:04organisations. That is an interesting measure
- 22:08actually that the the value unlock in the people that
- 22:10actually go through its courses. And I think any, any enduring
- 22:14corporate clients would want to see evidence of that.
- 22:16I mean by all accounts, I mean you and Blair has been a
- 22:19tremendous success for this. If we're going back to the
- 22:20valuation is punchy and you'd have to think that as you, as
- 22:25you sort of alluded to in in your opening story, it certainly
- 22:29won't hurt the is is connection to the family name and the kind
- 22:34of connections that can bring. Which also raises something of a
- 22:38question over Rachel Reeves's personal endorsement.
- 22:41It's perhaps not it's not that common for the chancellor to
- 22:46endorse a specific private fundraise by name.
- 22:49Nonetheless, this is exactly the kind of business they would want
- 22:52to be backing. So is so is, is multiverse
- 22:56something that we should be considering as a national
- 22:57champion in that regard? I think that remains to be seen.
- 23:00But a national champion strategy requires the champion to sit on
- 23:04the frontier. And I'm not sure that Multiverse
- 23:07is that company. It is doing what it does very
- 23:09well. These are astute investors and,
- 23:12and they're putting a considerable amount of money
- 23:15where the mouths are. But at the end of the day, to to
- 23:18sustain that kind of valuation, it will really need to show that
- 23:21they're considerably more than a workforce training business just
- 23:25with the UK enterprise book. Yeah, I agree.
- 23:27And I think one of the, one of the interesting challenges
- 23:31around this is that, yes, it looks like this could be sort
- 23:34of, you know, active UK AI industrial policy with, you
- 23:38know, with Rachel Reeves cheerleading for this.
- 23:41I don't feel that that's necessarily anything to do with,
- 23:44you know, who the founder is and their relation to, to, to the
- 23:47Labour Party and former leaders of the party.
- 23:50In fact, actually, it seems very often that the Labour Party
- 23:52seems to want to distance themselves from the past.
- 23:55I, I think one of the things that it does look like is
- 23:58potentially a chancellor looking to tie her name and her, her
- 24:04fiscal policy to any big success stories that can be, can be had
- 24:08because potentially they're, if they are few and far between.
- 24:11And so I, I wonder whether this is whether this is really
- 24:15looking at here's some good news.
- 24:16It's in an area that we want to be able to shout about.
- 24:19We're going to have to do it. So I wonder exactly, you know,
- 24:22how how happy she was actually having to to do this because,
- 24:26you know, realistically we should be having far more
- 24:30conversations when you look at just how much of the AI
- 24:34knowledge base was initiated in the UK and you know, and
- 24:38industrial strategy really should be supporting that, I
- 24:41believe far more than it is. Yes, indeed.
- 24:44OK. And with that, we'll move on to
- 24:45our final story of the week. Far more positive, I'd say,
- 24:48well, well, equally positive. London-based Isomorphic secures
- 24:52a 2 billion funding round A london-based AI company has
- 24:56closed one of the largest private funding rounds in
- 24:58European history and the British government has a seat at the
- 25:01table. Isomorphic Labs announced a 2.1
- 25:04billion Series B on May 12th, led by Thrive Capital, with
- 25:08backing from Alphabet, Singapore's Temasek, Abu Dhabi's
- 25:12MGX and the UK Sovereign AI Fund.
- 25:15The fund declined to disclose the size of its cheque, but its
- 25:19published guidance puts typical investments at between 1 and
- 25:2210,000,000 lbs. Isomorphic was spun out of
- 25:24Google DeepMind in 2021, built on Alpha Fold, the AI system
- 25:29that cracked one of biology's hardest problems, predicting how
- 25:32proteins fold. Isomorphic's engine applies that
- 25:35foundation to the full drug design process and already has
- 25:38commercial partnerships with Novartis, Eli Lilly and Johnson
- 25:41and Johnson. No human patient has yet been
- 25:44dosed. CEO Demis Hassabis originally
- 25:47promised clinical trials by the end of a target he revised at
- 25:52Davos in January to the end of 2026.
- 25:56Hassabis's dual role as head of Isomorphic and CEO of Google
- 25:59DeepMind points to something important.
- 26:02Isomorphic is not an independent British company.
- 26:05Alphabet is the majority shareholder.
- 26:08The research originated here, but the ownership sits in
- 26:11California. This returns us to a point we've
- 26:14raised again and again on his the US has a genuine claim to
- 26:18world class AI research and Isomorphic has absorbed talent
- 26:22from the wreckage of earlier British AI drug discovery
- 26:25companies like and benevolent AI.
- 26:29That's. XCNT, thank you and benevolent
- 26:33AI, but scientific advantage and commercial retention are
- 26:36different things. The Sovereign AI Fund's
- 26:38involvement is a signal of intent deployed into a company
- 26:41majority owned by one of America's largest corporations.
- 26:45So the question is, who captures the value when it succeeds?
- 26:48So what do you think, Matt? Is this a genuinely British AI
- 26:52success story? Well, I think it it is the
- 26:55research is unambiguously British.
- 26:57You know, Alpha Fold originated in London.
- 27:00You know, Hassabis is a Londoner and the company was built here
- 27:03from the ground up. So there's, and there's clearly
- 27:05a lot of the UK based on here. You know, Isomorphic has also
- 27:10absorbed talent from failed British AI drug discovery
- 27:13companies, as you said, Accenture and benevolent AI,
- 27:16turning the wreckage of earlier attempts into something
- 27:18hopefully more durable. And and the fact that Alphabet
- 27:21provides the capital backing doesn't make that, you know,
- 27:24that success any less British and the science any less
- 27:27British. You know, DeepMind has remained
- 27:29in London for a decade under Google ownership, but it does
- 27:33potentially change who benefits from, you know, the, you know,
- 27:38the increasing value of that that organisation.
- 27:41So the question, yeah, is it so genuinely British?
- 27:45I think your answer is kind of a is kind of actually answered in
- 27:50a somewhat different way and you might have to go to something
- 27:52more fundamental like what does it mean to be a British company
- 27:56now in terms of the research and the expertise.
- 27:59Yet I think the answer is an emphatic yes.
- 28:02In terms of ownership structure, I think it's is clearly no and
- 28:06and that is also supported by The Who the investors are the
- 28:11big, the bigger investors certainly although the
- 28:14involvement of the Sovereign AI fund is interesting.
- 28:17I'll come on to that in a second.
- 28:20We have discussed this on numerous occasions.
- 28:22How important is it that the company is actually primarily in
- 28:26the hands of British investors? I think we'd probably agree and
- 28:30we kind of concluded that it is important around, for example,
- 28:34decisions taken on where things are placed, where future funding
- 28:37rounds are. If an IPO took place, where
- 28:39would it be? But we can't ignore that the
- 28:42physical presence of these companies and the same goes for
- 28:46the growing presence of Open AI and Anthropic has got to be
- 28:49considered a positive, a net positive for the UK.
- 28:53It doesn't mean that the companies are British owned in
- 28:56in in the conventional sense. What it does mean is that it's
- 28:58giving rise to these really exciting spin offs and
- 29:01innovations and that in in the absence of having overall
- 29:05control of the company, that is still a massive benefit.
- 29:09I, I agree and, and look, there's, there isn't the capital
- 29:12in the UK that there is in the US You know, the economy is,
- 29:15are, are, you know, are considerably different in size.
- 29:17And so it's not surprising that the capital structure sits
- 29:20predominantly with with the US, but I think that the sovereign
- 29:24AI funds involvement, it is interesting, I think it does
- 29:28establish A precedent. The UK government is now a Co
- 29:31investor. I, I think more than anything
- 29:34else, this is pretty much strategic signalling it being at
- 29:37the at the table, even as a majority participant gives the
- 29:40government a relationship and potentially an influence it
- 29:42wouldn't otherwise have. I think it also gives it insight
- 29:46into what is happening. And I think that's really,
- 29:49really important. And I think that's a real value
- 29:51here add because understanding what's really going on within
- 29:55these organisations and how things can you can and will
- 29:58progress, I think is really, really interesting.
- 30:01You know, the fund is only a few months old, so judging it by the
- 30:05size of its first few cheques probably misses the point.
- 30:08But this does feel like it is a great it's a great thing for
- 30:12isomorphic in the fact that it is a really big deal in terms of
- 30:17the PR part of this when it comes to talking to really big
- 30:21organisations and really, you know, really does sort of talk
- 30:25to that the ability to then potentially open help open doors
- 30:28in other countries. I think as well to to follow on
- 30:32from that, I think well clearly in an ideal world, we would love
- 30:37for this one to be significantly bigger so that the scale of like
- 30:41a Singapore Temasek or something like that.
- 30:42But we are where we are to imagine that its investments are
- 30:48irrelevant and its activities are irrelevant because the the
- 30:50investment sizes are relatively small.
- 30:52I agree is also missing the point access experience and
- 30:56generally moving in a certain direction of backing potentially
- 31:00riskier initiatives. But on that point you just
- 31:04wonder. And I get, yes, it's early days,
- 31:06they've only just started. They probably wouldn't have
- 31:08wanted to go in with the first couple of investments on
- 31:11something completely unheard of. But the tougher decisions for a
- 31:15fund like this will arise when they are faced with an
- 31:19investment decision for a company who no one's heard of,
- 31:22with no famous founders or or backers, but with a potentially
- 31:26with with a potentially fundamentally influential new
- 31:29technology. That's when I think the amounts
- 31:33of money that they're putting in will become more relevant.
- 31:35But also their real appetite for risk and their and their
- 31:39willingness to back risky ventures will be far more
- 31:41apparent than it is with something like this, which is a
- 31:45fantastic company, isomorphic or as it looks like it has the
- 31:48potential to be if it succeeds. But you know, it's that old
- 31:52adage about no one got fired for hiring Microsoft.
- 31:56These are at this stage, these are probably as close to
- 32:00surefire bets within this sector as you can get.
- 32:03So we shall see When, when, when, when we're looking at
- 32:06investments in much lesser known companies.
- 32:09Yeah, indeed. I mean $2.1 billion is, is a,
- 32:12you know, right, does represent some of the most rigged rigorous
- 32:15institutional due diligence in the world.
- 32:17You know, these investors sort of thrive Capital, Alphabet,
- 32:20Temasek, as you say, you know, they've they've reviewed the
- 32:23data and committed at scale, you know, and they do have some
- 32:26impressive commercial partnerships, you know,
- 32:29Novartis, Eli Lee, J&J, you know, which is real world
- 32:33validation that the technology is producing value, not just
- 32:38promise. And when I say producing value,
- 32:40it's at least proving the that that there is some value in the
- 32:46the approach to being able to then look at what is the
- 32:49druggable space that we can look at and how do we approach
- 32:52identifying targets and then, you know, being able to drug
- 32:56them. You mentioned slip clinical
- 32:57timelines and I think that's that's a really interesting one.
- 33:01That's not going to be something within isomorphic power that's
- 33:04really reflects the complexity of drug development, not a
- 33:08failure of the underlying technology.
- 33:10It can also partly be down to how much, you know, how, what
- 33:14are the drug, what are the diseases that those
- 33:17organisations are targeting. So very often the review of
- 33:21these can often be challenging and actually recruitment into
- 33:24drug trials can take a long time.
- 33:26So while I think that we can be, you know, somewhat sceptical and
- 33:31is that a slip clinical timeline, I think repeated slips
- 33:34would be a matter of concern. Repeat, you know, and if and if
- 33:38potential drugs that are that were talked about then sort of
- 33:42get taught, you know stop being talked about and moved into
- 33:44other ways. That can be a challenge.
- 33:46But the real issue is, is looking at what are the
- 33:49milestones and how is the structure of the deal set up
- 33:53with those organisations, because that can cause real cash
- 33:55flow issues. And I think that that's where
- 33:57some of the, you know, some of their predecessors like
- 34:00Excellentia and Benevolent fell down because actually they had a
- 34:03huge amount of tied up in the delivery of molecules reaching
- 34:08certain parts of, you know, certain stages of the drug
- 34:11development pipeline. And when delays happened or
- 34:14things didn't happen, that put a crunch on on cash flow.
- 34:17Yes, indeed. And so on that note, Matt, I
- 34:20think that's a good point at which to sign off.
- 34:23Thanks, as ever, for talking through the week's issues with
- 34:25me. Yeah, likewise, Neil.
- 34:27It's been fun all. Right, we'll see you next week.
- 34:30Bye for now. And that's it for this week's
- 34:47episode of AI in the UK. If you found it useful, you can
- 34:52subscribe wherever you get your podcasts and we'd really
- 34:55appreciate a great rating or review.
- 34:58It genuinely helps others find the show.
- 35:01You can also follow Neil's Footman and Matt Wilkinson on
- 35:05LinkedIn for links, notes and occasional extra content.
- 35:10Thanks for listening and we'll be back next week.
- 35:13Helping you make sense of AI as it lands here in the UK.
- 35:15AI in the UK. AI in the UK.
- 35:15AI in the UK.