Upside: European Tech & VC Weekly

Anthropic IPO, Nvidia Leadership Change, and Europe's Tech Sovereignty Push

45 min Listen to the original
NvidiaAnthropicAI InfrastructureEuropean TechIPOsVenture CapitalUK EconomyDrone Delivery

Summary

Speakers analyze Nvidia's CEO resignation and its impact on the AI ecosystem, noting Jensen Huang's $61.1 billion in investments and the company's role as an AI 'central bank.' They detail Anthropic's IPO, valuing it up to $1.4 trillion, while debating whether $26 billion in supplier purchases represents genuine confidence or circular financing. The episode highlights Anthropic overtaking OpenAI in Q3 revenue and its rapid migration to AMD chips, challenging Nvidia's moat. European tech sovereignty is discussed, with new procurement rules and local alternatives like Helsing's Clarity gaining traction. Finally, the hosts cover UK economic growth driven by AI, startup fundraising advice for 2027, and Zipline's drone delivery expansion.

Key takeaways

  1. 1
    Nvidia committed over $40 billion to venture deals by May, with $30 billion going to OpenAI, creating significant interdependencies.

    Unidentified speaker (speaker_0) · 3:06

  2. 2
    A $1.5 trillion valuation for Anthropic is reasonable if it grows at 21-23% annually to reach $4.5 trillion by 2036, similar to early hyperscaler trajectories.

    Unidentified speaker (speaker_2) · 13:59

Episode sections

  1. 1:22 – 11:30

    Nvidia's CEO and its moat: Jensen walks after 33 years and 60 direct reports, Mads applies for the job, and Anthropic moves Claude onto AMD in three days

    The hosts discuss the resignation of Nvidia's CEO Jensen Huang following allegations of an affair with a Mega Speed exec, noting his central role in the company's operations and ecosystem deals. They analyze potential successors and the governance implications, while highlighting Anthropic's rapid migration of its Claude model to AMD chips as a sign of shifting competitive dynamics.

  2. 11:30 – 23:53

    Anthropic's IPO goes live: up to $1.4tn, 20% for retail, $26bn from its own suppliers. Vote of confidence or circular money? Yes. Plus OpenAI discovers it's no longer the biggest kid in class

    The podcast details Anthropic's IPO launch, valuing the company up to $1.4 trillion with significant participation from major tech suppliers like AMD and Nvidia. The hosts debate whether this represents genuine confidence or circular financing, while noting that Anthropic's recent revenue growth has surpassed OpenAI's, challenging OpenAI's market leadership.

  3. 23:53 – 29:28

    The AI trade whiplash: CoreWeave's best day in 18 months, junk-priced bonds, and why your old H100s are suddenly worth more

    The speakers address the volatility in the AI sector, citing CoreWeave's stock surge despite high debt yields and concerns about neo-cloud survival. They also note that spot prices for older H100 GPUs have risen due to insatiable demand, complicating depreciation calculations for data center operators.

  4. 29:28 – 30:21

    Nvidia Chip Economics and Capacity Constraints

    Speaker 2 explains that cannibalizing power from old H100 chips to power new NVIDIA chips requires the H100 spot price to rise by 33% to $4/hour to reach parity. Below this price, it is more economical to rip out old H100s and install new generation NVIDIA chips due to superior performance per megawatt.

  5. 30:21 – 32:01

    Europe's Push for Tech Sovereignty and Local Alternatives

    Speaker 0 discusses European fears of US tech firms switching off digital services, citing Helsing's launch of 'Clarity' as a European alternative to Palantir and France's intelligence service dropping Palantir. Fractile raised $720m to build AI chips, with Peter Thiel as a key investor.

  6. 32:01 – 36:40

    EU Procurement Rules and Sovereignty Debates

    Speaker 2 outlines potential new EU procurement rules allowing governments to mandate European provenance for strategic interests, discussed by EU heads of government with Anthropic's CEO and Geoffrey Hinton. Speaker 2 expresses concern that this could lead to second-tier technology and retaliatory lockouts from the US market.

  7. 36:40 – 40:31

    UK Economic Growth and Bank of England Rate Hike

    Speaker 0 notes the UK economy grew 0.7% in Q3, the fastest since 2023, driven by AI and cloud. The Bank of England raised rates for the first time in three years, with inflation expected to hit 5% mid-next year. Speaker 2 argues the rate hike is counterproductive as most inflation is imported.

  8. 40:31 – 42:11

    Startup Playbook for 2027

    Speaker 1 advises founders to raise at least 24 months of runway, demonstrate a path to profitability rather than just growth, avoid planning exits around IPOs, and ensure AI usage is defensible for cost-cutting or productivity. Investors will be more selective due to higher rates and inflation.

  9. 42:11 – 44:04

    Zipline Deal of the Week

    Speaker 1 highlights Zipline raising $1.5bn at a $21.5bn valuation and signing a national deal with Walmart for 300 new drone delivery sites. Zipline completed over a million deliveries in Dallas, with median arrival times of 19 minutes compared to 30+ minutes for traditional deliveries.

  10. 44:04 – 45:10

    Week Ahead: Nvidia, Anthropic, and Slush

    Speaker 2 notes NVIDIA reports Q3 results on Tuesday under new interim CEO Kress. Anthropic IPO is expected to price on Wednesday and trade on Thursday. The Millburn Welfare Review conclusions will also be released. The industry is expected to attend the Slush conference.

Assets discussed

AssetKindStanceContext
NVIDIANVDAequityneutralShares of bounce demand all over the place at time of pod down 13%, $600 billion a week. Uh in May, NVIDIA was worth over five $5 trillion, just for some context.
AlphabetGOOGLequityneutralstill there, but AMD, Google's TPUs, um, and the custom chips suddenly start to look much more like serious alternatives. I think it's leveled up the competitive landscape a little bit.
MicrosoftMSFTequitymixedAMD, Broadcom, Google, Micron, Microsoft, NVIDIA, TSMC are all buying 26 bill of stock at the IPO price.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Nvidia shares bounced down 13% and lost $600 billion in market cap in a week following allegations involving a Mega Speed exec.

Unidentified speaker (speaker_0) ·1:48

How to verify: Nvidia stock price history and market cap data from financial news sources (e.g., Bloomberg, Reuters) for the relevant week. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

Nvidia was worth over $5 trillion in May.

Unidentified speaker (speaker_0) ·1:48

How to verify: Nvidia market capitalization data from financial databases (e.g., Yahoo Finance, Bloomberg) for May 2026. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

One in every eight people in Singapore is buying a GPU.

Unidentified speaker (speaker_2) ·2:18

How to verify: Nvidia revenue reports broken down by region and Singapore population data; likely a rhetorical exaggeration rather than a literal statistic. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Mega Speed grew 90-fold in a year to almost $3 billion.

Unidentified speaker (speaker_0) ·2:47

How to verify: Mega Speed financial disclosures or press releases regarding revenue growth. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Nvidia committed over $40 billion to venture deals by May, with $30 billion going to OpenAI.

Unidentified speaker (speaker_0) ·3:06

How to verify: Nvidia investor presentations or SEC filings detailing venture investments and partnerships. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

Anthropic moved its Claude model onto AMD chips in three days.

Unidentified speaker (speaker_1) ·6:43

How to verify: Anthropic or AMD press releases or technical blogs detailing the migration timeline. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Anthropic's IPO is valued up to $1.4 trillion, issuing 825 million shares between $70 and $78, raising around $64 billion.

Unidentified speaker (speaker_0) ·11:40

How to verify: Anthropic S-1 filing with the SEC or official IPO prospectus. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

AMD, Broadcom, Google, Micron, Microsoft, Nvidia, and TSMC are buying $26 billion of Anthropic stock at the IPO price.

Unidentified speaker (speaker_0) ·12:00

How to verify: Anthropic S-1 filing listing underwriters and strategic investors, or press releases from the companies. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

Anthropic's run rate revenue was about $81 billion at the end of October, up $10 billion in a month.

Unidentified speaker (speaker_0) ·12:20

How to verify: Anthropic S-1 filing or financial statements included in the IPO prospectus. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

H100 spot price would have to increase another thirty-three percent to four dollars an hour until there's parity with cannibalizing power for new chips.

Unidentified speaker (speaker_2) ·29:48

How to verify: Calculate break-even point for H100 vs new NVIDIA chip power cannibalization using NVIDIA's performance-per-watt data and current H100 spot market rates. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Three quarters of Europeans now worry that someone outside Europe could switch off their digital services.

Unidentified speaker (speaker_0) ·30:41

How to verify: Search for Eurobarometer or similar EU public opinion survey results from 2026-08-14 to 2026-11-14 regarding digital sovereignty concerns. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Helsing launched Clarity, a European answer to some aspects of Palantir.

Unidentified speaker (UNKNOWN) ·31:10

How to verify: Check Helsing press releases or news articles from 2026-08-14 to 2026-11-14 for the launch of 'Clarity' and its stated purpose. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

France's domestic intelligence service is already dropping Palantir.

Unidentified speaker (UNKNOWN) ·31:10

How to verify: Search for news reports from 2026-08-14 to 2026-11-14 regarding France's DGSE or similar agency terminating contracts with Palantir. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Fractile raised $720 million at a valuation of over five billion to build AI chips.

Unidentified speaker (UNKNOWN) ·31:40

How to verify: Check Fractile's press release or SEC Form D (if filed) or Crunchbase/PitchBook data for the $720m raise and $5bn+ valuation. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

Peter Thiel is a key investor in Fractile.

Unidentified speaker (UNKNOWN) ·31:40

How to verify: Check Fractile's investor list or press release for Peter Thiel's involvement. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

The UK economy grew 0.7% in the third quarter, the fastest since 2023.

Unidentified speaker (speaker_0) ·36:52

How to verify: Check UK Office for National Statistics (ONS) GDP data for Q3 2026. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Computer programming and consultancy grew 4.4% in a single quarter in the UK.

Unidentified speaker (speaker_0) ·37:12

How to verify: Check UK ONS industry-specific output data for Q3 2026. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

The Bank of England raised rates for the first time in three years.

Unidentified speaker (speaker_0) ·37:32

How to verify: Check Bank of England interest rate decisions for 2026. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Zipline raised one point five billion dollars at a twenty one point five billion dollar valuation.

Unidentified speaker (speaker_1) ·42:11

How to verify: Check Zipline's press release or SEC Form D for the $1.5bn raise and $21.5bn valuation. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.

Primary source

Do your own research

  • Verify Nvidia's Q3 earnings report and interim CEO Kress's statements on leadership transition.
  • Check Anthropic's S-1 filing for detailed revenue breakdowns and customer concentration risks.
  • Review EU procurement rule proposals regarding 'European provenance' mandates for strategic tech.
  • Analyze Helsing's Clarity product specifications and France's intelligence service contract termination details.
  • Examine UK Office for National Statistics data for Q3 GDP growth drivers and Bank of England rate hike implications.
  • Assess Zipline's Walmart contract terms and delivery volume metrics against traditional logistics benchmarks.
  • Investigate Fractile's $720m raise details and Peter Thiel's involvement in European chip manufacturing.
  • Audit Nvidia's venture capital portfolio disclosures to assess exposure to OpenAI and other major AI players.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): Hello and welcome to Upside, where every week we dig behind the headlines that are going to affect European venture. Today it is Mads, Pree, and my good self, and what are we talking
0:08
Unidentified speaker (UNKNOWN): about?
0:10
Unidentified speaker (UNKNOWN): We've
0:10
Unidentified speaker (speaker_0): got Jensen, he is now gone. How will markets respond? Unpacking the Anthropic IPO. Europe decouples from American big tech. What's going gonna to happen there?
0:22
Unidentified speaker (UNKNOWN): Plus,
0:22
Unidentified speaker (speaker_0): what's happening in the week ahead? Predictions and deal dean of the week.
0:28
Unidentified speaker (UNKNOWN): Boys
0:28
Unidentified speaker (speaker_0): and girls, how are you? Pri, what's cooking with you?
0:32
Unidentified speaker (speaker_1): It's been, I think, a big week in tech. A lot has been happening.
0:36
Unidentified speaker (speaker_0): It is not
0:37
Unidentified speaker (UNKNOWN): a slow news week.
0:39
Unidentified speaker (speaker_0): It is
0:39
Unidentified speaker (UNKNOWN): not
0:40
Unidentified speaker (speaker_0): I'm I I'm just anxious. We were talking off off camera. I'm just anxious that we're missing stuff as we're talking. We are recording slightly ominously Friday the 13th in the morning. So who knows what
0:50
Unidentified speaker (UNKNOWN): we're doing.
0:51
Unidentified speaker (speaker_0): the rest of the week will bring? Mads, what's happening with you, my good man?
0:55
Unidentified speaker (speaker_2): Much like yourselves. I mean, you know, what is going on, right? And you can't get a minute's piece to focus on the real exciting news about Anthropics IPO before you get hit left, right, and center with meltdown this and scandal that. Get out of my
1:10
Unidentified speaker (UNKNOWN): way.
1:10
Unidentified speaker (speaker_2): Let's get back to business.
1:11
Unidentified speaker (speaker_0): Who would have thought a sex scandal could bring down AI? Well, I mean, let's let's talk about it. Let's talk about you know, Jensen resigning. So thirty-three years he's been at the helm. Uh one one founder, total control. Allegedly, now Mad, you know about this. Sixty direct reports. Is that is that right?
1:28
Unidentified speaker (speaker_2): Yep.
1:28
Unidentified speaker (speaker_0): He's uh he's proper founder mode, isn't he? So he's resigned as Nvidia's C. I'm sure everybody listening to this will know exactly what's going on. New York Times recently, or yesterday I think reported another alleged affair. I have to say alleged, because I we don't really know what what the deal is, but he's gone. So um the first was a Mega Speed exec, whom US
1:48
Unidentified speaker (speaker_0): officials believed was a Chinese spy. China denies it. Shares of bounce demand all over the place at time of pod down 13%, $600 billion a week. Uh in May, NVIDIA was worth over five $5 trillion, just for some context. But it's a it's an export controls issue, I would, I would argue. So we've got to follow the chips in twenty five,
2:08
Unidentified speaker (speaker_0): twenty four billion dollars worth of NVIDIA re revenue is through Singapore. Now, Mads, what was your quote? Was it didn't you work out how many GPUs per per capita or something that was?
2:18
Unidentified speaker (speaker_2): Yeah, I I think at the time we sort of said, Look, it's it's one in every eight people in Singapore that are buying a GPU, which seemed a bit extravagant. Now, NVIDIA at the time also said, no, no, no, we're just building through the company uh the country and things are really going elsewhere and yada yada yada they've since changed their reporting practices. But it did look slightly odd that Singapore, which is an important
2:38
Unidentified speaker (speaker_2): country, but not exactly one of the biggest countries in the world, was the the second largest market for NVIDIA in the world after the United States.
2:47
Unidentified speaker (speaker_0): Yeah. And uh and no great surprise, um mega speed, which is um the other huang, the uh the alleged affair of Jensen. Uh, her company, Megaspeed, grew 90-fold in a year to almost 3 billion. NVIDIA has also, everyone will know this, but they've become the central bank for the
3:06
Unidentified speaker (speaker_0): AI, and I'm sure people will remember the Bloomberg circular financing charts, which were all very interesting. So 67 venture deals last year. over 40 billion committed by May, 30 of that going to OpenAI. Now the CFO is running the company at time of pod. Makes sense that the CFO, I guess, who's who's running all of the
3:26
Unidentified speaker (speaker_0): finance has some control. But Mads, first off to you, one man's indiscretions. What's it bringing down? How deep will the rabbit hole go? And I this, I know this is a global a issue. but if there is any kind of specific European angles, let's dig into those as well.
3:43
Unidentified speaker (speaker_2): Yeah, a hundred percent. So I'd say it matters a lot, both of you know, both because of who he was and what he did, but also because of how hard he will be to replace. I mean, he personally ran and built the company from ground up, as you said, 60 direct reports, everything ran through him. He did the deals, he did so many things. And so he'll be hard to replace. It really took somebody
4:03
Unidentified speaker (speaker_2): of his caliber to do all the things that were done. You think about the roles he had. He said the technical direct direction around chips, systems, and software. He ran the company operationally, he handled Washington, he dealt with all the China issues. Maybe rightly or wrongly,
4:17
Unidentified speaker (UNKNOWN): you could say,
4:20
Unidentified speaker (speaker_2): but there you go. And he also made all the ecosystem deals. And you know, you were asking about uh what does this mean for Europe? I mean, maybe that's something we can come back to, but so many of the deals were done here. So I think that the big question is why is it a big thing and why is it hard to replace him? Well, because of all the things he did, and because of, you know, there's the fact that there's so few people that can do that
4:40
Unidentified speaker (speaker_2): sort of thing. You talked about Colette Crest, the interim CEO that's come in, you know, was the CFO, obviously know uh knows the finances well, knows the company well, but she is, you know, she's trusted. by Wall Street, but she's got no technical background. And then there's a number of other people that are being speculated about. There's uh Jim Buck, who invented CUDA, uh who's provided a
5:00
Unidentified speaker (speaker_2): lot of technical direction. And and yes, he has testified before Congress before, but he's never run a big organization, he's never been a CEO. Now, some people are saying, I think Trump was one of them, that Tim Cook could ride to the rescue, and he's run a big company, but arguably he was. the guy who totally missed the AI way. So
5:17
Unidentified speaker (UNKNOWN): is that the guy you waving in
5:20
Unidentified speaker (speaker_0): order to
5:21
Unidentified speaker (speaker_2): run this? I think you know one of the kind of names people are uh mostly throwing around right now is is René Haas, who is the CEO of ARM. He did work at NVIDIA for uh more than half a decade. He's running a listed chip company, he knows the technology. There was a a sort of an attempted action. acquisition of arm by NVIDIA a few years
5:41
Unidentified speaker (speaker_2): back. So he will have intimate knowledge of the company, although still you know nothing to the level of what what what Jensen did. So I will say and I'll put my hat in the ring here now if they want to replace one Jensen with another Jensen, who surely is very qualified. Well I'll I'll stand up and
5:57
Unidentified speaker (speaker_0): there you go. Jensen for Jensen. Fair match. Uh Pri, do you do you have any notes on this one?
6:03
Unidentified speaker (UNKNOWN): Yeah,
6:03
Unidentified speaker (speaker_1): I think it's just sort of r sort of reiteracing um the point of governance um amongst amongst the sound or going scandal going on with NVIDIA. It's only appointed its first ever chair this week. And as you said, the interim CEO is the CFO. And it's quite crazy that a four trillion dollar company, I guess it's in fact be maybe scrambling for for the right replacement in a such
6:23
Unidentified speaker (speaker_1): situation like this. And in then an ideal scenario, you have someone to step in. But I think the bigger question here is for the AI trade is the mode that NVIDIA has. In the same week, I think Anthropic said it's moved, um, its model moved clawed onto AMD chips, and it's been able to build um a system that allows it to do this in
6:43
Unidentified speaker (speaker_1): a few days
6:45
Unidentified speaker (speaker_1): instead of many months. And if AI can rewrite the software that log developers into NVIDIA, then NVIDIA's advantage depends more than ever on on execution, just as it loses Jensen, just as it's loses its leader who who drove it. And I don't mean this is the end for NVIDIA, the the products and orders and the customers they're all
7:04
Unidentified speaker (speaker_1): still there, but AMD, Google's TPUs, um, and the custom chips suddenly start to look much more like serious alternatives. I think it's leveled up the competitive landscape a little bit.
7:14
Unidentified speaker (speaker_0): That's interesting. Uh Mads, wha how how how deep or wide will this go? Will this be a a fumble, a bounce, and and back to business? Or are we are we gonna see residuals off the back of this? How how where do you think this will land?
7:29
Unidentified speaker (UNKNOWN): There
7:29
Unidentified speaker (speaker_2): would be residuals in so many places. He did so many deals, he was so intimately involved in them. Sixty $61 one billion dollars invested in stakes is and 229 startups. He's invested in listed companies, you know, more than 42 billion on acquisitions and buyouts over the last couple of years, including Hucking Face. Let's not forget
7:49
Unidentified speaker (speaker_2): Jensen Wang was one of the main proponents of open source, open weight, and has been putting so much money behind that. You asked about Europe and the link back to Europe. Now, Nvidia's been one of the biggest backers of European AI, 18 startup deals in 2025, worth almost five billion euros. Um he's been investing in
8:08
Unidentified speaker (speaker_2): Mistral and NScale and Lovable and Black Forest Labs and Synthesia. Ten of those 18 were here in the UK. So he's been extremely important for our ecosystem. And in Paris last month he said he would you know, he's already put eight billion dollars into Europe in the past year, two billion more to come, he said at the time. And because he's been so involved in the
8:28
Unidentified speaker (speaker_2): deals, it's hard to see immediately who will easily replace him. Obviously, having built the company, he had a license to go and you know be a little bit uh you know washbachling in terms of how he took these things on. And, you know, could a sort of more of a bureaucrat that takes over, uh more of an operator uh do deals in the same way and be quite as risk-seeking as
8:48
Unidentified speaker (speaker_2): he has been. So my guess is probably. not, and that will have an impact. Now, pre over to your point, I think the good thing is capitalism works and there's plenty of backups in the wider ecosystem from a technology perspective. So we've, you know, the TPUs are there, you know, AMD is making chips. There's plenty of competition coming up. And so I don't think, you know,
9:08
Unidentified speaker (speaker_2): we're not going to be lacking for tech and the drive to provide new uh technology and and new innovation. for the ecosystem.
9:16
Unidentified speaker (UNKNOWN): Uh,
9:16
Unidentified speaker (speaker_2): but I do think from a deal perspective and capital flow perspective, there will be implications.
9:21
Unidentified speaker (UNKNOWN): Is
9:21
Unidentified speaker (speaker_0): a question for you both is this could this be a positive insofar as it doesn't allow the AI crazy train to go too fast and too hard? So we've seen a couple of not correct, we're not in correction territory yet, but well, actually obviously in NVIDIA we really are. But is it is there an argument to say that this
9:41
Unidentified speaker (speaker_0): is just going to temper some of the crazy? And actually the AI train is still f running hard, but it's not going to be at the extremes, which would create even more volatility. Uh,
9:53
Unidentified speaker (UNKNOWN): in the sense
9:53
Unidentified speaker (speaker_2): you already had you already had quite a lot of that from the the political situation. Yeah.
9:58
Unidentified speaker (UNKNOWN): Let's
9:58
Unidentified speaker (speaker_2): not forget the election we had. Let's not forget kind of the expectation or the assessment. that almost half of all data center projects could be under threat. I think that is probably as much tempering as one would ideally want. That's what I also see as the flip side of it is there's so much fundamental positivity in the economics of AI that it
10:18
Unidentified speaker (speaker_2): will take more than one person stepping down from a role to really derail it. So I think individual deals, there'll be companies that miss out, there'll be deals deal makers missing out. You know, if you are you know one of the companies that was in the orbit and had the backing of Jensen, that this could hurt you. But I think overall as an industry, the the AI wave is too big and
10:38
Unidentified speaker (speaker_2): too powerful to be slowed down or at least be stopped by this.
10:42
Unidentified speaker (speaker_0): We're gonna talk later a little bit about how the the surprise UK economic growth from the last quarter was partly AI driven. So I that feels right. to me. Pri, I don't know if you've got anything to add to the
10:53
Unidentified speaker (UNKNOWN): to
10:54
Unidentified speaker (speaker_0): the level of discomfort and dislocation that we're going to see from this.
10:58
Unidentified speaker (UNKNOWN): I
10:58
Unidentified speaker (speaker_1): think exactly that. I think my biggest concern is is with the level of circuit financing um that that has happened within within the industry and where we can see how much of um NVIDIA's capital is going into the hyperscalers So and into to the neo clouds and um and into basically every player in the in the AI landscape. I think my question is well, was the knock on impact going to be? And we're
11:18
Unidentified speaker (speaker_1): gonna we're probably gonna see this quite soon.
11:20
Unidentified speaker (UNKNOWN): Well,
11:20
Unidentified speaker (speaker_0): let's let's talk about Anthropic, because this is your happy place, Mad. So we we've got to talk about Anthropic's IPO, which launched on Tuesday, I think. So a few background facts, and then I'm gonna come to you for some questions. So we're looking, I think between one and one 1.4 point four trillion, probably more in the 1.3 1.3-ish ish bucket, up from just under a trillion from its May funding
11:40
Unidentified speaker (speaker_0): round. They're issuing 825 million shares, between 70 and 78 dollars, looking to raise around 64 bill, something like that. First trade on the 19th. That's like next week. That's coming up quick. Uh, they are trading as ANT, as Ant, and about hit this is quite interesting one. About 20% is looking to go to retail against a typical five
12:00
Unidentified speaker (speaker_0): to 10. ten. I don't know if there's anything in that, but we'll come to that. back to in a sec. AMD, Broadcom, Google, Micron, Microsoft, NVIDIA, TSMC are all buying 26 bill of stock at the IPO price. And the run rate, so revenue is still looking very, very fresh and fruity. Run rate uh about 81 bil at the end of October, up 10 bill in a month. Pri, what uh just give us a bit of
12:20
Unidentified speaker (speaker_0): context and maybe lean into what what you make of this and your your initial take on the anthropic IPO.
12:27
Unidentified speaker (speaker_1): Yes, I think the headline valuation that you've mentioned of 1.4 trillion is definitely a statement. But I think the detail I'd focus on is who's buying this. And as you said, it's AMD, BoardCon, Google, Micron, and Microsoft, NVIDIA, and TSMC are all buying 26 billion, as you mentioned. And I think my question here is: is that a bird bode of confidence or is that circular money?
12:47
Unidentified speaker (speaker_1): Um, Is it both? These companies sell anthropic. um as chips and cloud capacity, and Anthropic has committed hundreds of billions to buying from them. So when they buy shares, they're partly protecting their own order books. That's one take on it. The run rate revenue is around 81 billion and grew 10 billion in a single month, which is extraordinary.
13:07
Unidentified speaker (speaker_1): But we spend the last month talking about circular AI financing and in our in our previous conversations. And to me, this is this is an example of it. that. The other detail I like is that about 20% is going to retail investors, which is which is double the norm for a lot of IPOs. And for ordinary investors, this to me seems like an opportunity to own a piece of the AI
13:27
Unidentified speaker (speaker_1): boom. Also, if it wobbles, it's not just going to be larger institutions who who feel this.
13:32
Unidentified speaker (speaker_0): Mad, you you were quite comfortable at 2 trillion. Um, is it cheap?
13:39
Unidentified speaker (UNKNOWN): Look,
13:39
Unidentified speaker (speaker_2): I said there was a case to be made. I think you know you're sort of saying that the valuation had initially been taken down, expectations had been lowered. You know, you're talking about 1.4, maybe 1.3. It's not what I'm seeing. From what we can see, the bankers are saying, look, it's already been five to six times oversubscribed. I'm expecting it to be priced at the higher end of the range, five 1.5.
13:59
Unidentified speaker (speaker_2): Kind of the case I looked at is if you look at anthropic. and you compare them to the Titans of today, right? Kind of the hyperscalers, if you think of anthropic as something that could be a hyperscaler within a decade, call it sort of the four and a half trillion dollar valuation, uh if you're targeting that by 2036, well, all the
14:19
Unidentified speaker (speaker_2): company will have to do is to grow at 21 to 23% a year for a decade. And that's not a scary growth rate when you compare to where they came from. Now, of course, there are lots of assumptions in it. And you know, you could take all the vectors of attack. You can talk about commoditization and open source models and lack of compute and lack of power and all the other things. And yes, things will
14:39
Unidentified speaker (speaker_2): have to go right. And so so you you sort of juxtapose the two the two things. But it doesn't feel like a scary hurdle to clear. And you know, what we've seen in our own tests is yes, open source, open weight notionally is cheap. But once you stand up the inference and the data centers to run it, it's hard
14:59
Unidentified speaker (speaker_2): to compete with the value you get out of the Frontier Labs. Our own tests are saying that under Frontier, I mean really what you're finding from a price performance perspective, it's mainly OpenAI and anthropic with the odd open source thing thrown into the mix. So if you have a situation where entropy can you know attain the margins of a hyperscaler, which doesn't seem unreasonable
15:19
Unidentified speaker (speaker_2): and can grow low, low 20s, low 20s percent a year, then that justifies a $2 trillion valuation. And so, yes, in the light of that, uh 1.5 doesn't seem completely unreasonable. Now, obviously, this is not investment advice. People should do their own research and all that stuff. But I I I'm certainly not in the camp a that says this is a crazy number at all.
15:39
Unidentified speaker (speaker_2): But this is, I think what probably become the most consequential company, uh, certainly has the potential to become the most consequential company of our lifetime. And I think it's a very, very exciting IPO.
15:51
Unidentified speaker (speaker_0): The is there an argument that this is just inertia. So the amount of work going into open source, open away weight and on the competitive side, is there an argument? to say that it's just inertia and the abilities and the management of open source open
16:11
Unidentified speaker (speaker_0): weight models will actually start to erode that, even though a relatively modest 20, 25% growth rate, is there an argument that that's just a 12 month inertia?
16:21
Unidentified speaker (speaker_2): There is always an argument, but let's go back 15 years, 10, 10-15 15 years and look at what people said about the hyperscalers. At the start, people said, AWS, how can it ever be a business? It's much cheaper to buy your own service and stand them up. But we all figured out that probably wouldn't make sense. All AWS is doing is, you know, mainly taking open source models and wrapping them in their own management systems and
16:41
Unidentified speaker (speaker_2): then reselling them. Like they've managed to do that, grow at a formidable clip over more than a decade with expanding margins.
16:49
Unidentified speaker (UNKNOWN): Okay?
16:50
Unidentified speaker (speaker_2): Now, running things at that scale is oftentimes harder than people think. There's been lots of issues being raised here over the last year around some of the potential issues of open source models, especially if things are baked into them. If you can't kind of fully show the provenance of the data and the training that go in, what will the models do? Could they be calling home to
17:10
Unidentified speaker (speaker_2): their makers and their creators? And so I think you could just as easily make an arch. argument the other way that people will want something that's that's controlled. Now, what I do think is real is I think we'll see more sovereign AI. I think Mistral has got a formidable playbook. And so here's the question, right? Would if anthropic and OpenAI become not the world's
17:30
Unidentified speaker (speaker_2): AI, but just the AI of the United States? Could If that still justify this valuation? Possibly. The US is a massive market. The TAM is almost. you know, infinite, right? It's all the work you can you can attack. And so look, I'm not saying this is a slam dunk. I'm not saying this is cheap at 10x the price and you should just go put all your savings into this
17:50
Unidentified speaker (speaker_2): company. But I certainly think there is a case for why it's it's a very attractive and exciting investment at the types of valuations we're discussing.
17:59
Unidentified speaker (speaker_0): Yeah, fair. Pri, any last notes before we move on?
18:03
Unidentified speaker (UNKNOWN): I
18:03
Unidentified speaker (speaker_1): think um, should we talk about OpenAI?
18:05
Unidentified speaker (speaker_0): Yeah, do it. Uh we're kind of ignoring it. I guess. Well, we are talking about the anthropic IPO. So yeah, so we have you know actively ignored it. But yeah, please do.
18:13
Unidentified speaker (UNKNOWN): So
18:13
Unidentified speaker (speaker_1): I think um what we've seen is um their third quarter revenue stored 50% to 10 billion, and that's 10 billion and a quarter up nearly 50%, um, which would be a record for for almost any company in history. But the more telling detail here is the comparison. Andropic did nearly 17 billion in the same quarter. So the company that was the clear leader two
18:33
Unidentified speaker (speaker_1): years ago is now smaller on the revenue, at least looking at that quarter. And on the um IPO side, um, OpenAI originally wanted to raise $30 billion at a $1.4 trillion valuation. And we're talking about Anthropic, of course, cutting down its valuation from $2 billion down. OpenAI has now cut that. to $20 billion at
18:52
Unidentified speaker (speaker_1): $1.25 trillion valuation. And last last month its revenue was below what I did signal to investors. So I mean the story is that
19:02
Unidentified speaker (speaker_1): the AI market is no longer a one horse race. Enterprise customers are spreading their bets across across several of the labs, and investors are starting to also price that in, as we can see with Anthropic.
19:12
Unidentified speaker (speaker_0): But the challenge I have, and I don't know if you've got any thoughts on this either of you, is there's a quality of revenue challenge, and there's also a how they calculate revenue challenge. So a lot of the services, from what I can see, and Mads, you'll know more about this than I do, but it feels like there's not hinky accounting, I don't think that's fair, but
19:32
Unidentified speaker (speaker_0): taking your best month of usage rather than you know committed annual contract and then multiplying that by thirteen just feels like we're gonna perhaps see more of what happened to OpenAI where they had to they lopped off twenty bill of their of their calculations or wherever they landed. So is there, Mads, maybe to you, is there
19:52
Unidentified speaker (speaker_0): a quality of revenue issue and is there a calculation of revenue issue that's slightly cooking the books? bugs.
19:58
Unidentified speaker (speaker_2): I mean let let's let's be clear. There is a mixture of kind of people in startup land doing startup you know finance and startup economics and it's kind of you know at the early stages we run on uh rules of thumb and it it is not a very precise odd right because things move very quickly and so uh there's speculation there's CSA pre
20:18
Unidentified speaker (speaker_2): you were talking about them cutting back from two trillion dollar valuation I actually don't think uh that was like that there was just speculation right there was people like us excited in the industry that were talking about that the company didn't go out and say we're going to raise the two trillion, at least not from what I've seen. So I think there's so much speculation and hearsay that gets loaded on top of this. And then I think you're right. There's some level of, well, we were a
20:38
Unidentified speaker (speaker_2): startup five minutes ago and now we're a billion-dollar company, but we maybe still run things a little bit like startups did in terms of how we count revenue and what's really recurring, what's really contracting and committed and all that stuff. And so yes, I think that's fair. All of the ability. above. And you can cut it both ways. You can say, oh, but does that mean it's not high quality? But you can also say, does that mean the opportunity is
20:58
Unidentified speaker (speaker_2): so big that these things are growing so fast that we just have to think a little bit differently about it from the way we would if it's a stale old hundred, you know, hundred-year-old company that sort of grows you a nice seven percent a year and it's just all steady as she goes.
21:10
Unidentified speaker (speaker_0): Yeah, fair. I I I I'm I'm relatively comfortable with the wild west. nature of it. I just get a little bit anxious when I see fairly major
21:22
Unidentified speaker (speaker_0): accounting uh retractions. I go, okay, well, how are you calculating revenue? Where is the revenue?
21:28
Unidentified speaker (UNKNOWN): So the way the
21:29
Unidentified speaker (speaker_2): discussion that is at least one of them, there's there's probably several, but one of the discussions have been about what is really your kind of your revenue figure. And if you sell through a hypers. hyperscaler and that hyperscaler bills I don't know, you know, five bucks per million tokens, do you then count the five bucks per million tokens they're billing, or do you count the stuff
21:49
Unidentified speaker (speaker_2): that you get once they've taken their cut? So it's sort of the gross number or the other or the net number. I think if you look at it the other way around, then And I think what you know people at Anthropic might say or the bankers might say is look, actually the company's moving to towards becoming a hyperscaler. Uh after this IPO, they'll have two hundred billion dollars of cash in the bank
22:09
Unidentified speaker (speaker_2): and they have got all these data center leases and they will be selling tokens directly and they will be taking that gross revenue number and therefore it's a fair reflection of the business they're building. You could make an argument either way, I think.
22:23
Unidentified speaker (speaker_0): Uh well, and also we're looking at a nearly a trillion uh trillion in in debt on the books as well. So I scary big numbers for a company that's what four years old? I mean, this is
22:33
Unidentified speaker (UNKNOWN): I don't think I don't
22:34
Unidentified speaker (speaker_2): think it's fair to say there's a trillion in debt. I think you have you have com commitments
22:38
Unidentified speaker (UNKNOWN): or
22:38
Unidentified speaker (speaker_2): liabilities
22:39
Unidentified speaker (speaker_2): in the form of contracts that have been signed. But that's that's that's normal. As as we know, if any of those data centers, if the uh people that are standing them up, if the developers don't manage to deliver, if they are late, and probably can walk away from those contracts and can maybe even get liquidated damages on them. So you know, as we discussed last week, kind of yes, eight hundred
22:59
Unidentified speaker (speaker_2): billion dollars of committed contracts, but if they uh really truly get their hands on that much, the revenue they can deliver with that far exceed what they need to have to justify even a two $2 trillion dollar valuation. And so this could mean one of two things. Either that they expect that they will grow
23:19
Unidentified speaker (speaker_2): much more, much faster than even the optimistic scenarios are painting, or that they're simply just saying, look, we know that if one and two data center projects fall through, we'll commit to 800 bill, knowing we'll end up with probably only 400 bill, and the others won't cost us anything other than the optionality of making
23:39
Unidentified speaker (speaker_2): sure we have the hedge that that there is uh you know multiple suppliers for our capacity.
23:45
Unidentified speaker (speaker_0): Well, while we're on the topic, let's also talk about the AI trade whiplash. Because because of the midterms, because of the the Jensen scenario, we've seen a lot of AI, the Mag 7, the big tech, um bouncing all over the place. And last week, AI trade found. fell off a cliff.
24:04
Unidentified speaker (UNKNOWN): Uh
24:04
Unidentified speaker (speaker_0): this week it bounced, but is it a recovery? Uh core weave had its best day since April 25. So last year, revenue up 168%. And we've spoken a lot about Core Weave.
24:14
Unidentified speaker (UNKNOWN): So
24:14
Unidentified speaker (speaker_0): they've got a backlog of 132 billion in orders, but they also have 47 billion in debt or owed. Bond yields over 16%. That's pretty much junk bond pricing. Softbank lost 7 billion. 3.7 billion on power equipment it bought for US data centers before it changed its plans. Microsoft is switching off older chip
24:34
Unidentified speaker (speaker_0): capacity. So there is a question about w the depreciation of chips and what that looks like. Oracle says it's renewing old GPUs at a 20% premium. And we saw recently, I think even this week, the Fed looking at a hot US inflation rate and uh a talk of a December rate hike. So a bit all over the place, but there is a obvious AI trade
24:54
Unidentified speaker (speaker_0): whiplash, which makes complete sense. Madge, last week you called this a knee jerk. What how do you feel about it now?
25:01
Unidentified speaker (UNKNOWN): Well,
25:01
Unidentified speaker (speaker_2): it all has to be put in context of what's happening at NVIDIA this week, doesn't it? You know, we know that NVIDIA had been a backer of Core Weave and i for NVIDIA the neo clouds are very important because they want to have that as a hedge versus the hyperscalers that have been their big customers. And want to make sure they have a plurality of firms out there. Is the new management going to be as
25:21
Unidentified speaker (speaker_2): assertive in terms of backing that whole ecosystem? For NVIDIA, I hope, but I also suspect there will be a temptation to scale some of that back, as there always is with Wall Street, right? Instead of you know funding all of these different things, you can hold more of the cash on your
25:38
Unidentified speaker (UNKNOWN): own
25:38
Unidentified speaker (speaker_2): balance sheet. or maybe even distribute it back to your shareholders. Wall Street often likes to see that. So, you know, what do I feel about it? I feel that the underlying AI move is still as strong as ever. The desire and the need for capacity is insatiable. You know, we've heard from all corners, all the leaders of every AI lab, including Mr. Al here in Europe,
25:58
Unidentified speaker (speaker_2): are saying, look, if we had more GPUs, we could deliver more revenue. There is an absolutely almost limitless demand for intelligence and for capacity. At the margin, as rates go up, as the central bank of Jensen Wang steps away, that will lead to more failures. There's no doubt. Of course, it has
26:18
Unidentified speaker (speaker_2): to. The marginal project won't go ahead, the marginal company And he won't survive. But my suspicion is that the underlying trend is still so, you know, the secular. trend is still so strong that the kind of the center will hold and that the industry will pull through.
26:33
Unidentified speaker (speaker_0): Pree, go dig into the core weave story a little bit. I think it's quite interesting.
26:37
Unidentified speaker (UNKNOWN): Yeah,
26:37
Unidentified speaker (speaker_1): so yeah, core weave's definitely an interesting one. So three weeks ago, we saw core weave's bonds were trading at 75 cents on the dollar and yielding over 15%. And we, of course, all saw the share price plummet. And Gartner was saying uh that that most neo clouds. won't survive this sort of period. And this week, revenue, as you said, it was up, and guidance was raised. And
26:57
Unidentified speaker (speaker_1): I think the revenue is up to 3.7 billion. And it says its remaining 2020 2027 capacity is 10 times oversubscribed. And the shares jumped almost 25%. So demand amounts, as you say, clearly isn't the problem. I guess there's a challenge on was it was it the financing and the fixed the fixed. fix is asset back debt at seven to
27:17
Unidentified speaker (speaker_1): seven point five percent borrowing secured against GPUs and and the contracts that they have with their customers. And then there's a question to say is that is that solving the problem? Is that postponing the problem? I'd say maybe it's postponing the problem, but it's sensibly done and it's sensibly structured. It's much cheaper than um than the 15% yield, which buys them time and it's
27:37
Unidentified speaker (speaker_1): debt secured on chips that that technically lose value every year so only if customers it only works if customers keep paying at today's prices. I think the Texas news helps core weave. The governor elect says the the site already um permitted and under construction will continue to be developed for those data centers. So
27:57
Unidentified speaker (speaker_1): I think the market the market as we saw swings on sentiment which was um three weeks ago and then you see their best day in 18 months.
28:06
Unidentified speaker (UNKNOWN): I
28:06
Unidentified speaker (speaker_0): wonder if we'll see more businesses moving to Florida now, Texas has turned blue. I wonder if there'll be like an emotional reaction to that. Um Matt, have you got any notes or thoughts on the depreciation piece and you know, ripping out the old chips and putting in some new?
28:20
Unidentified speaker (speaker_2): It's a really good question. And it's a delicate dance. I mean, But we've tried to do some analysis on it. Pre Pretty interesting. You said that the value of the chips uh depreciate every year. Actually, the spot prices are going up.
28:33
Unidentified speaker (UNKNOWN): So
28:33
Unidentified speaker (speaker_2): you can get more out of renting out your old H100 now than you could in May. It's gone up uh 20%.
28:42
Unidentified speaker (speaker_2): And that's just because the demand is so insatiable. So even the older trips chips are
28:48
Unidentified speaker (speaker_2): you thinking about your valuation of as something, is it about right? now or about the future? Yes, of course, ultimately it's about the future. But right now there's just such a shortage of GPUs that everybody's paying more and more, even for the old stuff. And so the question becomes: if we have to think about the world as a power-constrained world, so we're power constrained, memory constrained, data center constrained,
29:08
Unidentified speaker (speaker_2): at what point does it make more sense to hold on to the old GPU and rent it out? And at what point does the new GPU become so much more performant that if you're power constrained, it makes more sense to rip out the old GPU and put in a new one, even if the old one worked? Because you can't get the power.
29:28
Unidentified speaker (speaker_2): And so you have to sort of almost cannibalize the power that was used to run the old chip to get the new chip in, even with the cost of that additional uh depreciation in capex. From what we can see, if there is, if you imagine that tomorrow there's not going to be any more power, okay, so any new capacity will have to come from cannibalizing existing chips, the
29:48
Unidentified speaker (speaker_2): H one hundred spot price would have to increase another thirty-three percent to four dollars an hour until there's parity. And anything below four dollars a an hour, it makes more sense to simply rip out the old H one hundreds. and put in the newest generation of NVIDIA chips because even though you have the Capix
30:08
Unidentified speaker (speaker_2): up front, you get so much more performance per gigawatt or per megawatt out of the new chips that it pays for itself.
30:16
Unidentified speaker (UNKNOWN): Okay,
30:16
Unidentified speaker (speaker_3): so it's a slightly different conversation then, and then that then it becomes a capacity issue.
30:20
Unidentified speaker (UNKNOWN): I
30:21
Unidentified speaker (speaker_0): want to talk about um how Europe effectively is rejecting US big tech and where this starts and stops and how meaningful this is actually going to be. We can feel the the motion of the ocean and and how people are looking to go sovereign. And I think that makes sense intellectually. But what's the reality? So
30:41
Unidentified speaker (speaker_0): some background for us. So three quarters of Europeans now worry that someone outside Europe could switch off their digital services. Obviously, they're looking at Elon and Starlink and making up their own mind as to what's going on there.
30:54
Unidentified speaker (speaker_0): I think that's a fair fear. Fewer than half believe that foreign tech firms play by European rules. Voters have made up their minds. Brussels is vacillating. EU countries are watering down various proposals. We've spoken about EU Inc. and data localization issues.
31:10
Unidentified speaker (UNKNOWN): This
31:10
Unidentified speaker (speaker_0): week, Helsing launched Clarity, a European answer to some aspects of power. Palantir. And we're going to see more local operators. France's domestic intelligence service is already dropping Palantir. And we saw a report, I think, in The Guardian this week about Palantir lobbying a UK health minister to get into GP records, which I fear
31:30
Unidentified speaker (speaker_0): and assume there will be some resistance to. Meanwhile, Britain's fractal, fractile, I don't know if you saw this, dear. I think this is great. So they've they raised 720 million at a value. valuation of over five billi
31:40
Unidentified speaker (UNKNOWN): to
31:40
Unidentified speaker (speaker_0): build AI chips. The semi-irony, or maybe comment from the room, is that Peter Teal is a key investor and he's a co-founder of Palantir. So, you know, we I don't know how you cut this, but everyone's got their foot in everyone else's door. But um, Mads, kick this one off. So maybe
32:01
Unidentified speaker (speaker_0): the big broad question should European government governments buy European when the US product may be better? I don't know if you want to start there.
32:08
Unidentified speaker (speaker_2): Yeah, I mean I think it depends on what better means. You know, first ask is you know how vital and strategic is the piece we're talking about? Now I I'm not a fan of excluding all US vendors from everything. Let's not forget we've got lots of European technology companies that want to sell into the States, and presumably we don't want to block them from access to what is still the most attractive
32:28
Unidentified speaker (speaker_2): market in technology. So I think what we should do is we should be smart about it. We should create a process that enables there for there to be alternatives. But just as we want US procurement teams to buy from European companies, we should be buying from US firms as well, when that makes sense. There's discussion today, the 27 EU heads of government, they're
32:48
Unidentified speaker (speaker_2): holding a discussion around AI and Mr. Arles Arthur Mensch is there together with Jeffrey Hinton. The idea they're discussing is that there will be a new set of procurement rules that enables or allows European governments to say, look, when we go out and buy stuff, we can put in a clause in the procurement and the
33:07
Unidentified speaker (speaker_2): tenders to mandate that things are or have European provenance. But there there's is an important sort of addition to that, which is that the proposal is that the EU will be able to say to uh national governments, if something is a vital security, a strategic interest, we can
33:27
Unidentified speaker (speaker_2): mandate that you mandate that what you buy is European. Um, so that's an interesting change to the rule book. And I'm um I'm a little conflicted on it because you can easily see where this could go and how it could be abused to push people to yes, buy locally, doesn't it sound good? But if that means that European citizens
33:47
Unidentified speaker (speaker_2): and consumers end up with second-tier technology and with less choice and you know, sort of attempts to lock out vendors from the US, uh, with then the predictable retaliation, meaning that European firms will be locked out of the US market, then I'm not sure it will take us to where. we we ultimately want to be.
34:05
Unidentified speaker (UNKNOWN): Pri,
34:05
Unidentified speaker (speaker_0): I've got I've got two questions for you. One is uh how how a big the big picture, how serious is all this? And then secondly, I want to I don't know if you've got any notes on the Helsing Clarity product specifically and what it's looking to replace outside of Palantir. So first up your general thoughts on on the whole topic and then maybe if you can dig into some
34:25
Unidentified speaker (speaker_0): of the Helsing housing bits.
34:27
Unidentified speaker (speaker_1): So I think from a point of um of sort of Europe getting serious about um ending tech dependence, voters are clearly ahead of their uh politicians on this. Brussels wants European capacity in AI models, um chips, quantum, satellites and cloud, while member states keep watering down the reforms and cutting the budget on this. And what struck me is the
34:47
Unidentified speaker (speaker_1): point about the public sector could be the catalyst for this, simply by being a better customer. Mad's mentioned European tech startups often had good products, but government's defaults were the big US providers, just out of habit, and of course, out of out of having sort of more history in the market. So the fastest way to build European alternatives isn't through another fund, but it's
35:07
Unidentified speaker (speaker_1): government procuring from smaller European companies and letting them form consortiums to help meet the scale and to give give these European companies their first big reference customer. So I think that that those are my thoughts on what needs to happen. On healthing, so health and clarity, which which is the new product that goes straight at
35:27
Unidentified speaker (speaker_1): Palantir, is military AI that pulls together drones, satellite, and then sensor data to help commanders decide on what to act on, um, with a human authorizing every stripe. Can it win on sovereignty alone? And no, I don't, I don't think it needs to. Sovereignty will get gathered the room as a supplier, but running
35:47
Unidentified speaker (speaker_1): the customer's own servers under European control matters a lot after this year. And France's domestic intelligence service, dropping Palantir or a homegrown Home grown tool, shows that buyers are starting to listen in Europe. When you look at defense cards, customers, ultimately, they'll buy what works in the field. And Palantir is
36:07
Unidentified speaker (speaker_1): deeply embedded in Pentagon and NATO workflows. Hesling haven't effectively disclosed a large customer that are using Clarity yet. So I think the test is whether it can match Palantir on capability and then win on trust and control. And I think the good news is that it fits the determinant treatment point from the last
36:27
Unidentified speaker (speaker_1): story, which is that if European governments are serious about tech sovereignty, defence is where they should be the best customer.
36:35
Unidentified speaker (UNKNOWN): Well
36:35
Unidentified speaker (speaker_3): now maybe that um Alex Carp
36:37
Unidentified speaker (UNKNOWN): has become a
36:37
Unidentified speaker (speaker_3): Democrat,
36:38
Unidentified speaker (speaker_0): maybe that'll make everything
36:39
Unidentified speaker (UNKNOWN): a bit
36:39
Unidentified speaker (speaker_3): slicker.
36:40
Unidentified speaker (speaker_0): I want to change tack again and talk about the UK economy because we saw some lovely surprising stats this quarter from the third quarter. The UK economy grew 0.7%.
36:52
Unidentified speaker (speaker_0): Doesn't sound huge, but it's the fastest since 2023. And the ONS says the main engine is AI and cloud. And I would assume some defense in there, but yeah, let's go with AI. I like that story. Uh, computer programming and consultancy grew 4.4% in a single quarter. That's great news. I love it. Bring on more of the AI, more of
37:12
Unidentified speaker (speaker_0): the efficiency, more of the good stuff. The Bank of England just raised rates for the first time in three years. They're looking to see inflation hit about 5% in the middle of next year, which is not great news because obviously we'll then see an ensuing rate hike. So, AI powering the growth figures, good news, but we've still got more than a
37:32
Unidentified speaker (speaker_0): million young people not in work education or training. We've talked a little bit before about uh entry-level jobs. Mads, start this one off. I don't know where you want to start. I wanted to talk a little bit about should the bank have increased rates, should they've gone a bit bit higher now, but I'm not sure. Start start wherever you want on this, but but let's keep it
37:52
Unidentified speaker (speaker_0): positive. The UK economy growing. Whoop, whoop.
37:55
Unidentified speaker (speaker_2): Yeah, it's a good thing. I uh you know, of course it's tempting to say, oh, great, it's AI productivity that's flowing through, and we're all making much more with much less. And isn't it wonderful? I do think that's maybe slightly premature. I do think what's happening is that you enterprises companies are investing in technology and in AI because of the potential. And what we're seeing is the benefit of that.
38:15
Unidentified speaker (speaker_2): So it's to spend, it's to spend on services to implement, it's to spend on the technology, more so
38:21
Unidentified speaker (UNKNOWN): than
38:21
Unidentified speaker (speaker_0): being part of the figures.
38:23
Unidentified speaker (speaker_2): Exactly. It's the investment, which is super important, right? Investment is critical. And you know, in many ways, the UK has been an underinvested country for a long time. And so we need to invest more. And this is that. This is investing in capability. And isn't it, as you say, a great thing? Now, what should the rate be? I mean, it so quickly the sentence is
38:43
Unidentified speaker (speaker_2): something that's quite wonkish. I think my read is that most of the UK inflation is imported. So what does that mean? It means that it's stuff we can't influence ourselves here. It's stuff that comes in through higher energy prices and other international prices. And so you know, r jacking up our rates locally to counter inflation that
39:03
Unidentified speaker (speaker_2): comes from abroad is probably a bit counterproductive. I saw somebody saying, look, you know, maybe you should do it preemptively because the the it's not that the job market is overheating right now, but there will be way wage and salary negotiations negotiations next year. And it's important that we jack up the rates now to show that we're serious about inflation and we're not going
39:23
Unidentified speaker (speaker_2): to accept it. But I also think the country desperately needs this growth. And it would seem counterproductive to sort of slam a rate break on right when the growth is This here. I think much more important is to fix government spending. There was the uh review from Alan Miller. Millburn out this week, the welfare review. We know that there are far too
39:43
Unidentified speaker (speaker_2): many young people that are not in work, not in training, not in education, and just on welfare instead. And it's sad in so many ways. It's sad for the individuals, it's sad for the country, and it's it's devastating for the economy, uh, which is is just can't afford it. And so if we can you know restructure that, reform that, bring those deficits down, I think that'll be much more important.
40:03
Unidentified speaker (speaker_2): than than rate hikes and also ultimately much better for everybody in the long run.
40:07
Unidentified speaker (speaker_3): I agree. I agree. I th I'd like to see some big bold movements around
40:12
Unidentified speaker (UNKNOWN): uh
40:12
Unidentified speaker (speaker_3): youth unemployment and training and work. I think there's there's an argument that that should be quite there should be some bold strikes in that area to get people motivated. Uh Pri, bring this back to us and bring this home. What does this mean for startups? What does this mean for our ecosystem? How do you see that playing forwards?
40:31
Unidentified speaker (speaker_1): I think it could change your playbook a little bit with rates of 4% and probably staying high and potentially increasing, and inflation heading back above 4% and the IPA, IPO window better barely open, money in 2027 is going to be more expensive and more selective. So I guess when it comes to start to startups raising, for first,
40:51
Unidentified speaker (speaker_1): raise more runway than you think you need, which I guess is always the rule, at least 24 months because the next round may take longer and come in at a lower price. So that's the founders. Second, they're going to need to share a path of profitability, not just growth. Investors can now earn 4% safely and maybe that's going up. So they need a real reason to take a risk on you. Third,
41:11
Unidentified speaker (speaker_1): um, I guess yeah, do you want the this is a question, this might be contentious, but don't plan your exit around an IPO given given the current um IPO market here. And and maybe, maybe you're planning for a strategic sale eventually. And then fourth, think about where AI genuinely is going to help your business. Um investors will be separating companies that will use AI
41:31
Unidentified speaker (speaker_1): to cut costs versus Is creating productivity from than those just just sort of adding adding AI into their business. So it really needs to be defensible. I think the upside is that strong capital efficient founders will will face less competition for uh for the investors they want and they'll be able to get them on better terms than than in a frothy market.
41:49
Unidentified speaker (speaker_3): It sounds very much post post ZERPY, post COVID. Um actually not post COVID, post ZERPY times when uh everyone starts to
41:56
Unidentified speaker (UNKNOWN): you know
41:57
Unidentified speaker (speaker_3): be more capital efficient, tighten the belt, belts, do smart things, go AI native. Bree, Pre, you've got a deal of the week. My deal of the week we've already spoken about, which is fractal. I thought that was just brilliant. I'm loving it and and all good juice for them. But Pree, what was your deal of the week?
42:11
Unidentified speaker (speaker_1): Yes, um, my deal of the week is Zipline. So it raised one point five billion dollars at a twenty one point five billion dollar valuation, nearly triple what it was worth in January, and signed Walmart to a national deal for three 300 hundred new drone delivery sites up from twenty. 21 today. This is physical AI at a real scale, which is what I love
42:31
Unidentified speaker (speaker_1): about it. And over a million deliveries in Dallas this year alone, with median orders arriving at around nine in and out in 19 minutes, against over 30 minutes for traditional deliveries benchmark. And it started out flying blood and medical supplies in Africa, solving a real problem for it before because. it became
42:51
Unidentified speaker (speaker_1): a less use for for sort of day-to-day deliveries. And it also fits the pattern that I think that we touched on last week, which was Walmart were putting another six billion into symbiotic warehouse robots. And now Zippline are delivering on the on the effects of the last mile. So the biggest retailer in the world is automating its entire supply chain from warehouse
43:11
Unidentified speaker (speaker_1): to to the doorstep. I think the risks here are are in regulation, noise complaints, and then potentially Amazon's own drones. But I guess this is what happens when physical AI finds a customer with enormous scale. And I'm quite excited about it.
43:27
Unidentified speaker (speaker_3): I've never bought into getting my loo roll delivered to my garden via drone. I put Point sight to site sight, point to point I get. I've always figured Last Mild would be some kind of driving robot. I just I I can't buy it in so many markets, but obviously America is a little bit looser and their spread. I mean the Texas or wherever they're launched is is
43:47
Unidentified speaker (speaker_3): huge. So maybe we'll see, we'll see. Maybe there'll be some geoterritory plays on that and some regulatory stuff for sure. This week coming, I I think our entire industry will be at slush. But what else, Mads? What else do we need to look out for in the diary?
44:04
Unidentified speaker (speaker_2): Well, on Tuesday, we have NVIDIA reporting their Q3 results. So it'll be the first time that new interim CEO, Collect Kress, reports directly, not as CFO, but as the leader of the company. Same day we've got the Millburn Belfare review coming up as we just spoke about. So it'll be interesting to see what conclusions he draws. Uh, And then we have
44:24
Unidentified speaker (speaker_2): Wednesday is the expected pricing of the entropy. Anthropic IPO with Thursday being the first day of trading. So some big swingers coming next week for sure.
44:33
Unidentified speaker (speaker_3): Awesome, exciting. Uh that is it. Anyone got anything left in their brains that we haven't spoken about? Obviously, we're looking at the Microsoft meltdown at time of pod. Microsoft is melting down. But any any other little tidbits or thoughts before we close off?
44:49
Unidentified speaker (UNKNOWN): Then we
44:50
Unidentified speaker (speaker_3): shall call. it a day. My lovely people, thank you, and thank you so much for watching or listening this far, and we'll catch you next week.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights