AI infrastructure investment scale, profitability challenges, and market skepticism discussed with Azeem Azhar
Summary
Host Robert and guest Azeem Azhar examine the massive scale of AI capital expenditure, projecting $750 billion in global data center investment for 2026. They analyze specific company metrics, noting Anthropic's $850 billion in forward commitments and rapid revenue growth, while highlighting that most enterprises see minimal bottom-line impact from AI integration. The discussion contrasts the financial pressures on companies like OpenAI and CoreWeave with historical analogies to the early automotive industry. Speakers also address cybersecurity risks from AI-generated code and the market's reaction to tech firms shifting toward capital-intensive utility models.
Episode sections
- 0:01 – 3:07
Intro and UK Energy Context
Host Robert introduces guest Azeem Azhar and the topic of AI infrastructure investment risks. Greg Jackson briefly discusses UK electricity costs and grid bureaucracy as barriers to AI competitiveness.
- 3:07 – 4:30
Scale of AI Infrastructure Investment
Azeem Azhar details the massive scale of capital expenditure in AI data centers, noting a rise from under $100 billion five years ago to $750 billion projected for 2026, with hyper-scalers like Amazon, Google, and Microsoft leading.
- 4:30 – 10:03
Anthropic's Financial Commitments and Revenue
Azhar cites Anthropic's regulatory filing showing $850 billion in forward commitments for computing capacity. He notes Anthropic's revenue is growing rapidly, potentially reaching $85-$100 billion annualized, driven by new product launches.
- 10:03 – 17:05
Product Obsolescence and Customer Adoption
Azhar explains that 80% of Anthropic's sales come from products launched in the last six months, indicating rapid obsolescence. He highlights that most companies struggle to integrate AI end-to-end, with only 6-7% seeing significant bottom-line impact.
- 17:05 – 21:22
OpenAI's Challenges and Market Skepticism
The discussion contrasts OpenAI's 'scattergun' approach with Anthropic's focus. Azhar notes bond and equity investors are becoming skeptical, demanding more evidence of profitability, while SoftBank's share price halved due to its OpenAI exposure.
- 21:22 – 24:00
Safety vs. Financial Imperatives
Host and guest discuss the tension between the financial need for rapid, powerful product releases and the safety risks of AI self-improvement. Azhar argues that physical constraints like data centers and chips will slow recursive self-improvement.
- 24:00 – 34:11
Cybersecurity Risks and Corporate Power
Azhar warns that AI-generated code vulnerabilities are making the internet insecure, calling it the 'Internet's Wuhan moment.' He argues that powerful private entities like Anthropic and OpenAI hold more power than governments, creating a political challenge.
- 34:11 – 35:06
Credit Risk and Refinancing in AI Infrastructure
Speakers discuss the fragility of the global economy and the lack of transparency in AI financing, citing the Bank for International Settlements. They analyze CoreWeave's debt issues, noting its shares and bonds have fallen to junk status with effective interest rates of 15-16%, while highlighting that major refinancing pressures do not hit until late 2028.
- 35:06 – 40:33
Historical Analogies for the AI Boom
Azeem Azhar rejects the 2008 housing bubble analogy, arguing that current demand is driven by large, creditworthy corporations rather than subprime borrowers. He proposes the early 20th-century automotive industry as a better model, where weak firms failed but the industry grew due to sustained consumer demand, leading to consolidation among strong players like GM and Ford.
- 40:33 – 44:41
Oracle's Strategic Shift and Market Reaction
The discussion covers Oracle's scaling back of data center investments and the market's negative reaction to tech companies becoming capital-intensive utilities. Azhar explains that investors are re-evaluating companies like Google and Oracle as they shift from high-margin software models to lower-margin infrastructure businesses, causing share price volatility despite growing AI revenue.
- 44:41 – 49:00
Anthropic's IPO and Market Transparency
Azhar notes Anthropic's strong balance sheet with $130 billion in cash, reducing immediate bankruptcy risk. He emphasizes that Anthropic's upcoming IPO is a significant event for index funds and pension schemes, as it will bring transparency to the company's finances and allow analysts to scrutinize its business model, which was previously opaque as a private entity.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| AmazonAMZN | equity | neutral | THE BIG COMPANIES WHO INVEST IN DATA CENTERS THAT WE CALL THEM THE HYPERSCALERS, YOU'LL KNOW THEM AS AMAZON, GOOGLE, AND MICROSOFT. |
| MicrosoftMSFT | equity | neutral | to Anthropic, it's to OpenAI, it's to other companies, and sometimes it's directly to the infrastructure companies like Microsoft and Google, |
| AlphabetGOOGL | equity | neutral | to Anthropic, it's to OpenAI, it's to other companies, and sometimes it's directly to the infrastructure companies like Microsoft and Google, |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
Global AI data center investment is projected to be around $750 billion in 2026, up from less than $100 billion five years ago.
Unidentified speaker (speaker_3) ·3:30
How to verify: Check industry reports from firms like Synergy Research Group or IDC on global data center capex trends for 2026-2026. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Anthropic has $850 billion in forward commitments or forward plans for computing capacity over 10-12 years.
Unidentified speaker (UNKNOWN) ·4:55
How to verify: Review Anthropic's S-1 or S-1/A filing with the SEC EDGAR database for details on long-term lease commitments and capital expenditures. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic's annualized revenue is between $85 billion and $100 billion based on likely December revenues.
Unidentified speaker (speaker_0) ·6:56
How to verify: Calculate annualized revenue from Anthropic's latest quarterly revenue disclosures in its SEC filing or press releases. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary source80% of Anthropic's sales come from products launched in the six months prior to the filing.
Unidentified speaker (UNKNOWN) ·11:08
How to verify: Analyze product launch dates and revenue attribution in Anthropic's SEC filing or investor presentations. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceOpenAI is estimated to burn $660 billion before becoming profitable.
Unidentified speaker (UNKNOWN) ·20:06
How to verify: Review OpenAI's financial projections or disclosures in its regulatory filings or official statements regarding cash burn and path to profitability. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceOpenAI has almost a billion users, but only 5% pay them money.
Unidentified speaker (speaker_3) ·20:25
How to verify: Check OpenAI's user statistics and subscription conversion rates from official reports or SEC filings. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceSoftBank's share price has halved over the last few months.
Unidentified speaker (UNKNOWN) ·18:46
How to verify: Check SoftBank Group Corp (9984.T) stock price history on financial data providers like Bloomberg or Yahoo Finance for the period leading up to November 2026. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
The UK has a 700 gigawatt queue for electricity connections, while the entire electricity system is only 50 gigawatts.
Unidentified speaker (speaker_2) ·1:28
How to verify: Review National Grid ESO or Ofgem reports on the UK electricity connection queue and system capacity. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
UK electricity is about four times more expensive than in the US and China.
Unidentified speaker (speaker_2) ·1:15
How to verify: Compare industrial electricity prices in the UK, US, and China using data from the IEA or national energy regulators. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Between 6% and 7% of companies think they are getting a 6% bottom line impact from AI.
Unidentified speaker (speaker_3) ·14:31
How to verify: Identify and review the specific survey cited by Azeem Azhar (e.g., from Gartner, McKinsey, or similar) regarding AI ROI perceptions. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
The Bank for International Settlements is worried that there is not enough transparency around the financing of AI investments.
Unidentified speaker (speaker_0) ·34:32
How to verify: Review public statements, reports, or speeches by Bank for International Settlements (BIS) officials regarding AI financial risks and transparency between 2026-08-07 and 2026-11-07. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
CoreWeave's bonds fell to junk bond status with an effective interest rate of 15-16 percent.
Unidentified speaker (speaker_0) ·36:24
How to verify: Check CoreWeave's bond trading data and credit ratings from financial data providers (e.g., Bloomberg, Reuters) or SEC filings for bond indentures and interest rate disclosures. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceCoreWeave has to refinance tens of billions of dollars of debt, with refinancing hitting late 2028.
Unidentified speaker (speaker_3) ·37:23
How to verify: Review CoreWeave's debt maturity schedule in SEC filings (if public) or press releases/financial reports detailing debt maturities and refinancing plans. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceIn the early 20th-century US automotive industry, hundreds of companies failed and consolidated into three big winners: General Motors, Ford, and Chrysler.
Unidentified speaker (speaker_3) ·38:22
How to verify: Consult historical business records, automotive industry histories, or SEC filings of GM, Ford, and Chrysler for historical market share and consolidation data from the early 1900s to 2026. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceOver 100 years of US public companies, 60 companies are responsible for about 60% of all shareholder returns out of 25,000 companies.
Unidentified speaker (speaker_3) ·39:44
How to verify: Verify against academic studies on stock market returns, such as those by Elroy Dimson, Paul Marsh, and Mike Staunton (Credit Suisse Global Investment Returns Yearbook), or S&P Dow Jones Indices long-term return data. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Google's utility business (selling computing) grew 100% year on year, but the company lost 11% of its share price in a day because search advertising stopped growing.
Unidentified speaker (speaker_3) ·41:35
How to verify: Check Alphabet's quarterly earnings reports for Cloud/AI revenue growth and search advertising revenue trends, and cross-reference with Google's stock price history for the specific day mentioned. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Amazon stated its computing infrastructure business could be a $1 trillion revenue business within five or six years or seven years.
Unidentified speaker (speaker_3) ·42:52
How to verify: Review Amazon's investor presentations, earnings call transcripts, or official press releases from AWS leadership regarding long-term revenue projections. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Anthropic has $130 billion in cash.
Unidentified speaker (speaker_3) ·44:43
How to verify: Check Anthropic's financial disclosures, press releases regarding funding rounds, or SEC filings (if available) for cash and cash equivalents. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic is going to enter the index, making it one of the bigger companies on the market.
Unidentified speaker (speaker_3) ·46:27
How to verify: Review announcements from index providers (e.g., S&P Dow Jones, MSCI) regarding the inclusion of Anthropic in major indices following its IPO. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Anthropic will start to get proper transparency after its IPO, allowing analysts to ask penetrating questions.
Unidentified speaker (speaker_0) ·47:09
How to verify: Monitor post-IPO analyst reports, SEC filings (10-K, 10-Q), and earnings call transcripts for Anthropic to assess the level of financial transparency and analyst coverage. For recent events, verify 2026-08-07 through 2026-11-07. Use another period only when the speaker explicitly names it.
Primary sourceDo your own research
- Verify the $750 billion 2026 AI data center investment projection against recent earnings reports from Amazon, Google, and Microsoft.
- Check Anthropic's regulatory filings for the $850 billion forward commitment details and current revenue figures.
- Review surveys on enterprise AI adoption rates to confirm the 6-7% bottom-line impact statistic.
- Analyze CoreWeave's bond prospectus and credit rating reports for the 15-16% effective interest rate and refinancing timeline.
- Examine the Bank for International Settlements (BIS) reports on transparency in AI investment financing.
- Compare current AI infrastructure margin profiles with historical automotive industry consolidation data.
- Assess Google's recent earnings calls for details on search advertising stagnation and AI revenue growth rates.
Full transcriptexpand
- 0:01
- Unidentified speaker (speaker_0): Are we close to a global financial crisis and is the nexus of weakness, irrational exuberance about the prospects for artificial intelligence, reckless financing of all that infrastructure, the data centers, the compute? Everywhere you look, you can see
- 0:20
- Unidentified speaker (speaker_0): investor NETWORK COMPANIES anxiety. INC, NETWORK But is COMPANIES this just INC, the NETWORK sort COMPANIES of necessary INC, adjustment?
- 0:45
- Unidentified speaker (speaker_0): IS, Azeem Azhar. He runs a consultancy, a firm of analysts called Exponential View, and literally nobody is more steeped in the data than Azeem Azhar.
- 0:58
- Unidentified speaker (speaker_1): We're delighted to say that this year the rest is money is powered by Octopus Energy. And that means we get to welcome back to the show the founder and CEO of Octopus Energy, Greg Jackson. With all the talk of data centres and AI, how can the UK have the competitive advantage?
- 1:15
- Unidentified speaker (speaker_2): Yeah I mean the reality is the UK has got electricity that's about four times more expensive than the US and China and so the first thing we've got to do is get our electricity prices down. The second thing is
- 1:28
- Unidentified speaker (speaker_2): they talk about a queue to get an electricity connection. These numbers may sound confusing but basically BASIC of the UK has got a 700 gigawatt queue and our entire electricity systems only 50 gigawatts. What we need is reform of the bureaucracy that manages the queue so important projects get done quickly, not just data centres but houses, electricity generators, grid
- 1:48
- Unidentified speaker (speaker_2): scale batteries to store our renewables and of course we need to get our electricity costs down.
- 1:55
- Unidentified speaker (speaker_1): Thank you very much Greg. You'll be hearing more from Greg on The Rest is Money this year but for now let's get on with the episode.
- 2:21
- Unidentified speaker (UNKNOWN): BUSINESS COMPANIES INC, BUSINESS COMPANIES INC, BUSINESS COMPANIES INC,
- 2:45
- Unidentified speaker (speaker_3): OLLI, LANEIGE, and, yeah, you're going to need a bigger bag. Shop at Sephora. Limitations apply. Must be a beauty insider. See terms at sephora.ca for complete details.
- 3:07
- Unidentified speaker (speaker_0): Azeem, great to see you as always. Talk us through where we are in terms of understanding the scale of investment in data centres, in compute, in the infrastructure of artificial intelligence.
- 3:18
- Unidentified speaker (speaker_3): It's much bigger than when I was last on your show, Robert. We're going to look at around $750 billion in data centres in 2026, the
- 3:30
- Unidentified speaker (speaker_3): vast majority of that going into the specialist A.I. data CENTERS. centres for AI.
- 3:35
- Unidentified speaker (speaker_3): If we go back about five or six years, that number globally was less. THAN 100 BILLION. SO WE HAVE SEEN A LARGE INCREASE. AND THE BIG COMPANIES WHO INVEST IN DATA CENTERS THAT WE CALL THEM THE HYPERSCALERS, YOU'LL KNOW THEM AS AMAZON, GOOGLE, AND MICROSOFT. THEY HAVE THESE OTHER BUSINESSES OTHER THAN THE CONSUMER ONES.
- 3:54
- Unidentified speaker (speaker_3): AND NEW COMPANIES IN THAT SECTOR ARE INDICATING THAT THEY WILL CONTINUE TO INVEST IN THIS INFRASTRUCTURE
- 4:01
- Unidentified speaker (speaker_3): APPROACHING A TRILLION OR MORE IN 2027. 2021.
- 4:03
- Unidentified speaker (speaker_3): More than that in 2028.
- 4:05
- Unidentified speaker (speaker_3): And we're
- 4:06
- Unidentified speaker (UNKNOWN): talking
- 4:06
- Unidentified speaker (speaker_0): dollars here.
- 4:06
- Unidentified speaker (UNKNOWN): We're talking dollars.
- 4:07
- Unidentified speaker (speaker_0): And we're really just talking about the West, aren't we? We're not talking about China in these numbers.
- 4:11
- Unidentified speaker (speaker_3): We're not talking about China. The Chinese numbers are much smaller from a sixth to a quarter of that size as it goes out.
- 4:20
- Unidentified speaker (speaker_3): There's just much, much more money in American and Western
- 4:24
- Unidentified speaker (speaker_3): technology companies than there is in the Chinese companies.
- 4:27
- Unidentified speaker (speaker_0): How much of that forward plan is actually committed money?
- 4:30
- Unidentified speaker (speaker_3): Well, there's a little bit of a disagreement, I suppose, in the market about what really
- 4:36
- Unidentified speaker (speaker_3): is committed.
- 4:38
- Unidentified speaker (speaker_3): I think if we look at some of the well-known AI labs, Anthropic is one of them. It makes Claude. We know from its recent regulatory filing to go public that there's several hundred billion in dollars there. about $850 billion of commitments or forward plans. So
- 4:55
- Unidentified speaker (UNKNOWN): $850
- 4:56
- Unidentified speaker (speaker_0): billion,
- 4:57
- Unidentified speaker (speaker_0): one company
- 4:59
- Unidentified speaker (UNKNOWN): of
- 4:59
- Unidentified speaker (speaker_0): committed forward commitments. I mean,
- 5:00
- Unidentified speaker (UNKNOWN): I should
- 5:01
- Unidentified speaker (speaker_0): remind people that the GDP of the UK is about £3 trillion.
- 5:05
- Unidentified speaker (speaker_0): So that is, I don't know, what we call about a quarter of UK GDP, something like that.
- 5:11
- Unidentified speaker (speaker_3): It is. And to be clear about what Anthropic's commitment is, it's not to... building data centres. It is to renting the computing capacity of those data centres over
- 5:22
- Unidentified speaker (speaker_3): between 10 and 12 years. So you really have to smear that 850 billion out over, say, a decade.
- 5:28
- Unidentified speaker (speaker_0): And we should remind people, this is a business that didn't exist three years ago.
- 5:32
- Unidentified speaker (speaker_3): It didn't exist three years NETWORK ago. COMPANIES No. But it INC, has NETWORK grown COMPANIES very, very INC, NETWORK quickly. COMPANIES INC,
- 5:57
- Unidentified speaker (UNKNOWN): NETWORK COMPANIES INC, NETWORK COMPANIES INC, NETWORK
- 5:58
- Unidentified speaker (speaker_0): COMPANIES INC,
- 6:21
- Unidentified speaker (speaker_3): That was way too high. That was a crazy number to reach. And Anthropic actually blew their own business plan. I mean, there's this great scene in The Italian Job where
- 6:30
- Unidentified speaker (speaker_3): Michael Caine character asks one of his other characters to
- 6:34
- Unidentified speaker (speaker_3): open a van and the character blows the van up and Caine goes, I only asked you to blow the doors off. And in a way, that's what's happened with Anthropic. They'll deliver Over $53 billion of revenue in a year where most people thought seven was going to be an extraordinary asset.
- 6:51
- Unidentified speaker (speaker_0): And they've just done a filing literally in the last 24 hours where I think one could say the
- 6:56
- Unidentified speaker (speaker_0): annual run rate is even higher than that, isn't
- 6:59
- Unidentified speaker (speaker_3): it? If you multiply out their likely December revenues, you'll get to a number between $85 and $100 billion, which means that they are growing and so on, very, very quickly. That marketing document, which is a marketing document for They're their filing fighting their public. listing, just shows a lot of really interesting
- 7:19
- Unidentified speaker (speaker_3): details about the business. And in a sense, it's the encapsulation of the whole question about AI. On the one hand, you've got the revenues. Is demand real? Is it sustainable? Is it broad?
- 7:32
- Unidentified speaker (speaker_3): Because it's revenues NETWORK ultimately COMPANIES INC, that NETWORK COMPANIES pay INC, for NETWORK all of COMPANIES this. And INC, on the other side, you've got these $700, $800 billion of
- 7:57
- Unidentified speaker (speaker_0): EW, NUCOR COMPANIES INC, EW, NUCOR COMPANIES INC, EW, NUCOR COMPANIES INC, EW, NUCOR
- 8:21
- Unidentified speaker (speaker_0): Decent Return on This Investment, Pay Their Debts, and Thrive. So, where are you at the moment in your assessment of whether the revenues are growing fast enough and can be sustained?
- 8:38
- Unidentified speaker (speaker_3): Right now, as we talk in November 2026, the
- 8:43
- Unidentified speaker (speaker_3): sales that are coming into the whole of the generative AI ecosystem, as we call it, so it's
- 8:49
- Unidentified speaker (speaker_3): to Anthropic, it's to OpenAI, it's to other companies, and sometimes it's directly to the infrastructure companies like Microsoft and Google,
- 8:57
- Unidentified speaker (speaker_3): are pretty significant. They're around about $230 billion for the last 12 months. But if you annualize October, you're at a little bit above NETWORK $350 COMPANIES INC. billion. So that tells you that revenues are still growing very, very quickly. And so this isn't just a story about Anthropic or
- 9:17
- Unidentified speaker (speaker_3): OpenAI, it's a story about essentially companies in the US, 75%, and the rest of the world, 25%, spending money on AI.
- 9:28
- Unidentified speaker (speaker_3): And the two questions that I think are really pertinent here is, how long will growth rates stay 可以保持著這麼高成ưởng? reasonably B, high? Are my They companies obviously still won't experimenting? stay D, at this growth rate for years. the second is,
- 9:38
- Unidentified speaker (speaker_3): are companies
- 9:40
- Unidentified speaker (speaker_3): Could they take a step away from this technology in a year or two's time or-
- 9:44
- Unidentified speaker (speaker_0): David Willis
- 9:44
- Unidentified speaker (speaker_3): Customers, right. Our customers, but you know, ordinary high street businesses, main street businesses, as we call them in the US, are they going to stick with this in this level of investment over the next five years? Those are two questions I think that you have to keep looking at.
- 9:59
- Unidentified speaker (speaker_0): But, Matt Faircloth okay, But let's break it down into- I think we should try and look at it now. Yes,
- 10:03
- Unidentified speaker (UNKNOWN): let's look at
- 10:03
- Unidentified speaker (speaker_0): it now. As it were. And one of the things that I was struck by, and I think you commented on this,
- 10:11
- Unidentified speaker (speaker_0): was that if you look
- 10:13
- Unidentified speaker (speaker_0): at Anthropix filing for its, you know, what in this country would be called a flotation
- 10:20
- Unidentified speaker (speaker_0): globally as and an IPO, an initial public offering, it's quite clear that revenue growth
- 10:27
- Unidentified speaker (speaker_0): is very dependent on the launch of new products.
- 10:32
- Unidentified speaker (speaker_0): Are new products still, over the next two to three years, going to be the big driver of the kind of growth that is necessary?
- 10:43
- Unidentified speaker (speaker_0): And if so,
- 10:44
- Unidentified speaker (speaker_0): how worried should we be that the pace of development of new products may well slow down, not least because of safety concerns?
- 10:54
- Unidentified speaker (speaker_3): If we go back to that document for the flotation, we saw that 80% of Anthropix sales come from products they launched from the beginning of April over the next six months, and that is really extraordinary. That's going to be $40 billion by
- 11:08
- Unidentified speaker (UNKNOWN): the end of And the year.
- 11:09
- Unidentified speaker (speaker_0): That is just astonishing.
- 11:11
- Unidentified speaker (speaker_3): It's astonishing. And it tells you that the products they developed last year, or even the year before, decay in their-
- 11:17
- Unidentified speaker (UNKNOWN): Are basically
- 11:17
- Unidentified speaker (speaker_0): obsolete now.
- 11:18
- Unidentified speaker (speaker_3): Quite quickly. Yeah, basically obsolete. And so, I
- 11:22
- Unidentified speaker (speaker_3): think it tells one positive story,
- 11:24
- Unidentified speaker (speaker_3): which is the investment that they make in new products pays off, and it pays off quite quickly.
- 11:30
- Unidentified speaker (speaker_3): The negative story is that you may need to continue to spend quite a lot, as you allude to. Now, Now you asked a really pertinent question. which is, will they be dependent on that level of innovation? And that comes back now to the customer, the ordinary business
- 11:49
- Unidentified speaker (speaker_3): in the UK, France, the US, which is, can they actually
- 11:54
- Unidentified speaker (speaker_3): absorb and pay for these really advanced products? Or is at some point the product simply too good and too powerful for a typical insurance company to make use of? And I think that, again, that's a pertinent question. And one indication that we get that it's going to take a little bit of work to get there
- 12:14
- Unidentified speaker (speaker_3): is that companies like Anthropic and OpenAI have, together with private equity businesses, launched professional services firms. And the goal of those firms is to take the most advanced models into companies, and help them...
- 12:28
- Unidentified speaker (UNKNOWN): Teach them how to use it.
- 12:29
- Unidentified speaker (speaker_3): Teach them how to use it, because on their own, they haven't shown broadly that they can do that.
- 12:34
- Unidentified speaker (speaker_0): I want to come back to the issue of the risks around revenue growth, but just since we are talking about the application of AI within In businesses,
- 12:48
- Unidentified speaker (speaker_0): you're immersed in this question
- 12:51
- Unidentified speaker (speaker_0): of, you know, how fast and how broad
- 12:54
- Unidentified speaker (speaker_0): these products will be disseminated.
- 12:59
- Unidentified speaker (speaker_0): Give us an example
- 13:01
- Unidentified speaker (speaker_0): of
- 13:02
- Unidentified speaker (speaker_0): what kind of productivity enhancing application is now being applied in the sort of best case uses, but why many businesses might I be anxious about think speedy rollout.
- 13:15
- Unidentified speaker (speaker_3): businesses aren't so much anxious about speedy rollout but they're just struggling to do it well because this is not as simple as replacing one printer with a better printer. It really requires thinking about how you operate and I'll just give you an example of a
- 13:32
- Unidentified speaker (speaker_3): problem and we ran into. If you make one part of your business move quickly because you're using AI and it's moving five times quicker than the next bit, so say manufacturing is moving faster than distribution, then all you do is you create a
- 13:47
- Unidentified speaker (speaker_3): bottleneck at the loading dock because the trucks can't take the goods away. And so the companies that have figured out how to do this really well, and it's not many of them, have redesigned the way they work so that it becomes an end-to-end loop that means that there's no bottleneck in the process. And that's more than just
- 14:08
- Unidentified speaker (speaker_3): improving one part of the business. That means you have to really think about what happens when you approve a new contract? What happens when that approved contract goes to manufacturing? What happens when your manufacturing is able to produce it more quickly? Can the warehouse team get it out the door more quickly? And that's much, much more complicated than swapping out a laser printer. And that's why I
- 14:27
- Unidentified speaker (speaker_3): think when we see data from surveys,
- 14:31
- Unidentified speaker (speaker_3): typically they show that between 6% and 7% of companies
- 14:35
- Unidentified speaker (speaker_3): think they're getting a 6% bottom line impact from AI. And the vast majority of companies are in that 1% to 2% who really cares NETWORK COMPANIES level right INC, NETWORK now. COMPANIES And I INC, think NETWORK the COMPANIES challenge INC, is going to be, can that 94%
- 15:09
- Unidentified speaker (speaker_0): NETWORK COMPANIES INC, NETWORK COMPANIES
- 15:11
- Unidentified speaker (speaker_0): INC, NETWORK COMPANIES INC,
- 15:33
- Unidentified speaker (speaker_3): We hear these announcements, I think we have to put ourselves in the minds of those companies. If you want to cut costs, cut employment costs, you can tell a good news story to the market by saying, we're bringing our costs down because we've mastered this amazingly powerful and difficult technology, which will pay us rewards continuously for the next several years.
- 15:53
- Unidentified speaker (speaker_3): So I think we need to also discount what we see on face value from these announcements and recognize I there'll be lots of other motivations think at play.
- 16:02
- Unidentified speaker (speaker_3): my bigger concern about employment over the next few years is actually the quiet, slow degradation of the market, making it that much harder for a new graduate to get their first foot on the ladder because there are fewer
- 16:19
- Unidentified speaker (speaker_3): junior jobs being hired. And we've seen professional National services firms cut their hiring. And then the other aspect is when someone leaves, does a company advertise for a replacement to fill that role or not? And all of that, Robert, happens very, very quietly, very perniciously. It really impacts people, but it doesn't show up in a headline
- 16:39
- Unidentified speaker (speaker_3): figure that makes the news. And I suspect that that may well be in effect, that we will feel more over the and next the OEMs in the three UK. or four years and that we have to look out And we've for.
- 16:48
- Unidentified speaker (speaker_0): got another question here
- 16:49
- Unidentified speaker (UNKNOWN): which
- 16:50
- Unidentified speaker (speaker_0): is, are you expecting to see a significant rise in the next three or four years and that we have to look out for? Now, we've focused
- 16:58
- Unidentified speaker (speaker_0): on a company that, in many respects, seems to be doing pretty well.
- 17:05
- Unidentified speaker (speaker_0): But OpenAI, it's a big challenger.
- 17:08
- Unidentified speaker (speaker_0): We There had originally are thought quite it widespread was also concerns going to IPO that this year. all It's is not. not tickety-boo, all is not great at OpenAI. What is actually going on there?
- 17:20
- Unidentified speaker (speaker_3): Well, OpenAI has a different way of running to Anthropic. Anthropic has been very, very focused. OpenAI has
- 17:29
- Unidentified speaker (speaker_3): been a little bit more scattergun, and that means that in a way, they didn't realize where this huge opportunity was going to come in 2026, which was all about making
- 17:39
- Unidentified speaker (speaker_3): AI tools for software development. But OpenAI also exists in a broader context, which is the capital markets and the financial markets, which you know very well. And what we've seen over the last four months is, first,
- 17:54
- Unidentified speaker (speaker_3): bond investors, the people who write the loans to TRI, companies, BUSINESS to data centers, getting much, much more skeptical and much more demanding about the loans that they will write. And in the last few weeks, we've seen equity investors absolutely flee the tech sector. And while you've seen the NASDAQ come down several percentage points, it's not
- 18:14
- Unidentified speaker (speaker_3): by any means in the last five years the worst dip it's had.
- 18:17
- Unidentified speaker (speaker_0): No, but it's in what people would call correction territory,
- 18:20
- Unidentified speaker (UNKNOWN): it's not bear
- 18:20
- Unidentified speaker (speaker_0): market.
- 18:21
- Unidentified speaker (speaker_3): And some companies have dropped 50% or 60% in
- 18:24
- Unidentified speaker (UNKNOWN): the
- 18:24
- Unidentified speaker (speaker_3): AI business from their June peaks. So, that is also significant.
- 18:27
- Unidentified speaker (speaker_0): It is significant. I mean, look, SoftBank is one of the biggest tech investors in the world. It is intimately associated with OpenAI. It's got this big investment. I think it's got something like 13% of OpenAI. And its share price has halved I think over the last few months. that's
- 18:46
- Unidentified speaker (UNKNOWN): And right a
- 18:47
- Unidentified speaker (speaker_0): lot of it is connected to its investment in open OpenAI. ai So, what is going on
- 18:51
- Unidentified speaker (speaker_3): so i think what is what's going on is that you know the market is being asked to
- 18:58
- Unidentified speaker (speaker_3): value things for which there's only two or three years worth of data and that data is in the in that when those companies are really young they're quite indisciplined and the bond investors NETWORK COMPANIES are the INC, ones NETWORK who are COMPANIES most INC, steel NETWORK COMPANIES headed INC, about needing to find data and they started to lose
- 19:33
- Unidentified speaker (speaker_3): I think, which is extraordinary numbers.
- 19:36
- Unidentified speaker (speaker_3): But that is the context of a market that has over the last three months,
- 19:41
- Unidentified speaker (speaker_3): both on the bond side and the equity side, become much more sceptical and demanded much more evidence.
- 19:47
- Unidentified speaker (UNKNOWN): I
- 19:47
- Unidentified speaker (speaker_0): mean, on this podcast, Sebastian Malaby, who did that very compelling biography of Dennis Hassabis, said he thought there was a significant risk that OpenAI would go bust.
- 19:59
- Unidentified speaker (speaker_2): I actually think it's going to go bust. Their own estimates have been that they're going to... burn $660 billion before they start
- 20:06
- Unidentified speaker (UNKNOWN): making a profit. Sorry,
- 20:06
- Unidentified speaker (speaker_0): 600, let's just say that number
- 20:08
- Unidentified speaker (UNKNOWN): again. Yeah,
- 20:09
- Unidentified speaker (speaker_2): it's a good point. $660
- 20:11
- Unidentified speaker (speaker_2): billion with a B. I mean, this is insanely big. They're
- 20:14
- Unidentified speaker (UNKNOWN): going to
- 20:15
- Unidentified speaker (speaker_0): lose that amount of money.
- 20:16
- Unidentified speaker (speaker_2): That's how much they're going to burn. Yeah, lose that amount of money.
- 20:19
- Unidentified speaker (UNKNOWN): Negative cash flow. Negative
- 20:20
- Unidentified speaker (speaker_2): cash flow. They've got almost a billion users, but only 5% pay them money. the
- 20:25
- Unidentified speaker (speaker_3): No, I don't think it's right. I think OpenAI has got more than enough cash.
- 20:29
- Unidentified speaker (speaker_3): It also has more importantly than cash, more importantly than imagination, a business that will do in the order of $35 to $40 billion of revenue this year with a path to doing that profitably. And it has lots and lots of customers. So there will be a price at which it'll be attractive for an investor to come in and
- 20:49
- Unidentified speaker (speaker_3): buy a chunk of that company or put capital into that company. Companies will go bust if they don't have a customer franchise or if they don't have sustainable economics. In other words, if they're selling
- 21:04
- Unidentified speaker (speaker_3): five pound notes for three pounds, that is not a business that is sustainable. But we've seen from Anthropic, and we've seen as OpenEye's got more disciplined, that it can become profitable.
- 21:15
- Unidentified speaker (speaker_3): and it has got customers who stick with it and continue to spend with it.
- 21:22
- Unidentified speaker (speaker_0): One of the great paradoxes of
- 21:25
- Unidentified speaker (speaker_0): where we are in AI is,
- 21:28
- Unidentified speaker (speaker_0): in a sense, the better, the more powerful the product, the greater, and you know, this is what the companies themselves can see, the greater, we might call the societal
- 21:39
- Unidentified speaker (speaker_0): risks.
- 21:41
- Unidentified speaker (speaker_0): And, you know, we're not going to, you know, rehearse this at great length again, but, you know, we've seen pretty much every AI leader talking about the meaningful risk of
- 21:50
- Unidentified speaker (speaker_0): AI wiping us all out. Now, we are seeing in AI development, as you say, these better and better products. And one of the things that's quite striking at The the moment is
- 21:59
- Unidentified speaker (speaker_0): the data on how AI is moving.
- 22:02
- Unidentified speaker (speaker_0): moving to this
- 22:04
- Unidentified speaker (speaker_0): position of, you know, what they call self-recursion, self-improving.
- 22:09
- Unidentified speaker (speaker_0): And
- 22:11
- Unidentified speaker (speaker_0): there is quite a lot of data with the latest products. So much of what is making the products better is AI improving
- 22:18
- Unidentified speaker (speaker_0): itself
- 22:19
- Unidentified speaker (speaker_0): with sort of human oversight, but not human control.
- 22:24
- Unidentified speaker (speaker_0): And so this plays against this. growing concern that, in
- 22:30
- Unidentified speaker (speaker_0): a sense, this is not a million miles from a sort of nuclear arms race, and the world is being put at risk by the speed of development. So I really wanted to look at that tension
- 22:40
- Unidentified speaker (speaker_0): between a
- 22:42
- Unidentified speaker (speaker_0): company like Anthropix saying safety really NETWORK COMPANY matters
- 22:47
- Unidentified speaker (speaker_0): but then what
- 22:49
- Unidentified speaker (speaker_0): we discussed at the start of the PROGRAM IS THAT
- 22:52
- Unidentified speaker (speaker_0): IN ORDER TO MEET ITS FINANCIAL TARGETS AND ACTUALLY BASICALLY IN ORDER TO PAY ITS DEBTS
- 22:59
- Unidentified speaker (speaker_0): AND
- 23:00
- Unidentified speaker (speaker_0): GIVE INVESTORS WHAT THEY EXPECT,
- 23:03
- Unidentified speaker (speaker_0): IT HAS TO ESSENTIALLY KEEP COMING UP WITH
- 23:09
- Unidentified speaker (speaker_0): BETTER AND BETTER PRODUCTS AND ARGUABLY EVEN AT AN ACCELERATED RATE. AND SO THERE IS SORT of this tension,
- 23:17
- Unidentified speaker (speaker_0): and I find it quite a worrying tension, between the financial markets imperative, better and better products, and the safety of the world imperative, which is, let's keep proper control over this development.
- 23:32
- Unidentified speaker (speaker_0): And I just wondered how you think this is gonna going to
- 23:36
- Unidentified speaker (speaker_0): play out, because it is that tension is what causes me. most anxiety
- 23:40
- Unidentified speaker (speaker_3): you know we are you've put it beautifully uh we're in uncharted territory uh at this moment and often radical technologies whether it's the printing press or the steam engine put us into uncharted territory and so that means we have to think NIC, quite NIC, NIC, quite NIC, NIC, clearly about what is going
- 24:00
- Unidentified speaker (speaker_3): on and how we as
- 24:21
- Unidentified speaker (speaker_3): So good at software coding now that they are finding millions, if not billions of vulnerabilities
- 24:28
- Unidentified speaker (speaker_3): on the internet. Yeah. All the refrigerators have failed in supermarkets in Europe.
- 24:32
- Unidentified speaker (UNKNOWN): Yeah. And we saw
- 24:33
- Unidentified speaker (speaker_2): that incredible
- 24:34
- Unidentified speaker (speaker_2): PHP debacle only a few days ago.
- 24:37
- Unidentified speaker (UNKNOWN): Right.
- 24:38
- Unidentified speaker (speaker_3): So that technology now
- 24:40
- Unidentified speaker (speaker_3): is out. If it hadn't come out this year, it would have been available next year or the year after. And you have to think NETWORK about COMPANIES How INC, do NETWORK you COMPANIES protect INC, NETWORK and COMPANIES defend INC, against that? And right now, the very, very best.
- 25:09
- Unidentified speaker (speaker_3): and so on. It all depends on these proprietary models. So that's one complicating factor.
- 25:13
- Unidentified speaker (speaker_3): Then the second complicating factor is the observation that you make, which is if they are getting towards this idea that AI will make AI better and that AI will make the next AI better, it's called recursive self-improvement,
- 25:27
- Unidentified speaker (speaker_3): if
- 25:28
- Unidentified speaker (speaker_3): how do you actually control that system and we don't have the science to control it? Now I am
- 25:33
- Unidentified speaker (speaker_3): skeptical about the argument that these companies are, even though they claim in it their documents. on the verge of a recursive self-improvement that accelerates off to infinity. I think things may speed up, but the reality of trying to get a data center, trying to get the data, trying to find the chips, are physical
- 25:53
- Unidentified speaker (speaker_3): realities that start to slow things down. But nevertheless, things and so on will and so forth. get faster. Things will
- 25:59
- Unidentified speaker (speaker_3): And the place we will have to go to for that is going to be not the market, it will have to be the state, whether it is a
- 26:07
- Unidentified speaker (speaker_3): coalition, whether it is the governor level in the US,
- 26:12
- Unidentified speaker (speaker_3): that will be the way in which we will interrogate and inspect all of this.
- 26:16
- Unidentified speaker (speaker_0): Okay. But I'm just going to ask you, given
- 26:21
- Unidentified speaker (speaker_0): that NETWORK COMPANIES you have INC, known NETWORK this community, COMPANIES INC, NETWORK this COMPANIES world INC, for years, right, you know,
- 26:45
- Unidentified speaker (speaker_0): and you're you're faced with you know it's a couple of days before the thing is priced and you're faced with this dilemma you've got this you know you've got this new product
- 26:57
- Unidentified speaker (speaker_0): um and you know to to use the italian job analogy on one level you know if you release it it's going to blow the doors off um
- 27:07
- Unidentified speaker (speaker_0): uh investors uh and customers are going to get super excited but at the back of your brain you're thinking do I actually understand what the safety implications of this? Have I tested it enough? In the end, do you make the moral choice that you don't release it?
- 27:26
- Unidentified speaker (speaker_0): Or do you take a risk that the rest of us would rather you didn't?
- 27:33
- Unidentified speaker (speaker_3): N.V.S.I.S.T.A.R.I.O.S. HEAD. BUT I THINK
- 27:37
- Unidentified speaker (speaker_3): IT'S ALSO REALLY ALLURING FOR US TO WORRY ABOUT EXISTENTIAL RISKS BECAUSE THEY'RE SO EXCITING, THEY'RE SO ONE-OFF IN A SENSE. THAT'S WHY WE GO AND SEE MOVIES THAT ARE THRILLERS.
- 27:50
- Unidentified speaker (speaker_3): I THINK IT DISTRACTS US FROM
- 27:53
- Unidentified speaker (speaker_3): MUCH MORE PRESENT AND IMPORTANT QUESTIONS, ONE OF WHICH WE TALKED ABOUT WAS SECURING OUR DIGITAL INFRASTRUCTURE. at the second. Well,
- 28:02
- Unidentified speaker (UNKNOWN): OK,
- 28:02
- Unidentified speaker (speaker_0): but on that, but do you think that the
- 28:05
- Unidentified speaker (speaker_0): giants of this industry, I mean, I mean, let's let's not talk about it in terms of existential risk. Let us just talk about it as this tension between security and safety and the rate of product release, which I just think is a very real and worrying tension. And we have an American government that, you know, is
- 28:25
- Unidentified speaker (speaker_0): led by a president. who simply, you know, sees all of this in a very, you know, simplistic America versus China framework, which is why he's not very interested in imposing regulation on these businesses. And therefore, we're still in a position where it is largely about self-regulation and and
- 28:45
- Unidentified speaker (speaker_0): Self self-policing. Policing. And I just you know I just want to know you know how uneasy you are about it and how uneasy we should be about
- 28:53
- Unidentified speaker (UNKNOWN): it. Well
- 28:53
- Unidentified speaker (speaker_3): I mean I wrote about this a few weeks ago about how you know the, we've crossed a Rubicon
- 29:00
- Unidentified speaker (speaker_3): of Capability in these models and it's going to make the Internet insecure. I called it the Internet's Wuhan moment, this moment where- Yeah,
- 29:08
- Unidentified speaker (UNKNOWN): it was a brilliant piece, I wrote it.
- 29:10
- Unidentified speaker (speaker_3): And my view, my conclusion from that is that there's
- 29:13
- Unidentified speaker (speaker_3): two years of very, very hard work for cyber security folk to secure the internet, to use the fact that the closed models are somewhat better than the models the attackers have, that in that two years, there will be bad outcomes as we may
- 29:28
- Unidentified speaker (UNKNOWN): well be seeing today. As you
- 29:29
- Unidentified speaker (speaker_0): say, we're living through all these fridges.
- 29:31
- Unidentified speaker (UNKNOWN): We've had a British government telling us to stock up on baked
- 29:32
- Unidentified speaker (speaker_0): beans, You know, we've had a British government telling us to stock up on baked beans and water, which,
- 29:41
- Unidentified speaker (speaker_0): you know, most of us assume is because, you know, is because they think there will be a serious power outage as a result of an
- 29:49
- Unidentified speaker (UNKNOWN): AI
- 29:49
- Unidentified speaker (speaker_0): attack for two or three days. And who knows how long it would take them to restore it. But that appears to be...
- 29:54
- Unidentified speaker (speaker_3): I'm sure that's what's in their mind. I've not spoken
- 29:56
- Unidentified speaker (UNKNOWN): to them about it, but that makes And
- 29:57
- Unidentified speaker (speaker_0): sense. I'm Like you, not that is making, what I think is
- 29:59
- Unidentified speaker (UNKNOWN): going you on.
- 30:00
- Unidentified speaker (speaker_3): know, Robert, I'm just, I'm describing the weather
- 30:03
- Unidentified speaker (UNKNOWN): as
- 30:03
- Unidentified speaker (speaker_3): opposed to saying whether the weather is good or bad. I think this is
- 30:05
- Unidentified speaker (UNKNOWN): what
- 30:05
- Unidentified speaker (speaker_3): will happen over
- 30:06
- Unidentified speaker (UNKNOWN): the
- 30:06
- Unidentified speaker (speaker_3): next two years. But I want to come back to the other thing that
- 30:09
- Unidentified speaker (UNKNOWN): we
- 30:09
- Unidentified speaker (speaker_3): lose track of here, which is, I think a lot of
- 30:13
- Unidentified speaker (speaker_3): what we need to consider is this simple idea of power. You know, we have vested power in governments that we elect, and they are meant to keep private sector interests within check, within the kind of rules that we think are important. And there's always that little battle. They nick a bit here and the government gets a bit back
- 30:33
- Unidentified speaker (speaker_3): there.
- 30:35
- Unidentified speaker (speaker_3): With AI, we're looking at new
- 30:37
- Unidentified speaker (speaker_3): classes of power. Because right now, the most powerful machine intelligence anywhere ITS is not possessed or controlled by any government. It's possessed or controlled by
- 30:49
- Unidentified speaker (speaker_3): Anthropic. AND THEN AFTER THAT OpenAI. AND THE ABILITY TO USE THAT IN MANY, MANY DIFFERENT WAYS IN PRIVATE, WE SAW THAT SIX WEEKS AGO, EIGHT WEEKS AGO, BOTH COMPANIES MADE ENORMOUS BREAKTHROUGHS IN UNSOLVED MATHS PROBLEMS. BUT IT WAS DONE PRIVATELY.
- 31:07
- Unidentified speaker (speaker_3): AND WHAT ELSE COULD THEY BE DOING PRIVATELY THAT WE DON'T KNOW ABOUT? AND I DON'T THINK WE HAVE A MODEL OF POLITICS THAT ALLOWS US TO MAKE SENSE OF THESE companies yet. And so when I think about these existential risks, I think, let's not lose sight
- 31:21
- Unidentified speaker (speaker_3): of figuring out how we deal with these new, really powerful corporate entities
- 31:25
- Unidentified speaker (speaker_3): in our world, and who's going to, and how are we going to take any kind of control, societal control over them.
- 31:32
- Unidentified speaker (speaker_0): I mean, these are very important points. Actually, one of the things I was struck by over the last And few days,
- 31:37
- Unidentified speaker (speaker_0): I guess sort of slightly shocked and upset by is pretty much the only global leader who seems to me to be saying quite sensible things about essentially how,
- 31:52
- Unidentified speaker (speaker_0): you know, whether it is the state or supranational,
- 31:56
- Unidentified speaker (speaker_0): you know, cooperating bodies
- 31:59
- Unidentified speaker (speaker_0): need to get more of a grasp about what's going on inside the global economy. these extraordinarily powerful businesses, is the Pope.
- 32:06
- Unidentified speaker (speaker_3): Well, he's been brilliant. He has been fantastic. I think Mark Carney in Canada has also been good. And in the UK, we have an extraordinary minister called Kanishka Narayan.
- 32:16
- Unidentified speaker (UNKNOWN): Yeah, he's
- 32:16
- Unidentified speaker (speaker_0): great.
- 32:17
- Unidentified speaker (UNKNOWN): He's been on the
- 32:18
- Unidentified speaker (speaker_0): podcast. I think he's pretty impressive.
- 32:21
- Unidentified speaker (speaker_3): And He knows his book, and he's saying extremely sensible things. But you're right, It's going to take time and, of course, time is not really of the essence right now.
- 32:30
- Unidentified speaker (speaker_0): So the good news, Azeem, is that malevolent AI super brains haven't wiped us out yet. So there's tons more that we need to talk about in the second part after a quick break.
- 32:41
- Unidentified speaker (speaker_2): Okay. When I sell my business, I want the best tax I and want investment advice. I want to help my kids and I want to give back to the community. Ooh, then it's... the vacation of a lifetime. I wonder if my head of office has a forever setting.
- 32:57
- Unidentified speaker (speaker_1): An IG Private Wealth Advisor creates the clarity you need with plans that harmonize your business, your family, and your dreams. Get financial advice that puts you at the center. Find your advisor at IGPrivateWealth.com.
- 33:11
- Unidentified speaker (speaker_1): This is big.
- 33:14
- Unidentified speaker (speaker_1): The Sephora Savings Event is back. And now through November 9th, everything's on sale. Everything you really want. Like one size Patrick Ta, Karastas, Kayali, Laneige, and yeah, you're going to need a bigger bag. Shop at Sephora. Limitations apply. Must be a beauty insider. See terms at sephora.ca and for complete details.
- 33:34
- Unidentified speaker (UNKNOWN): www.cab.ca for complete details.
- 33:37
- Unidentified speaker (speaker_1): Your favorite Starbucks flavor is coming home. Introducing new Starbucks Cinnamon Dolce by Nespresso for Virtuo.
- 33:46
- Unidentified speaker (speaker_1): Ready with the touch of a button on your Nespresso Virtuo machine, these pods are the perfect way to start your day with the Starbucks café quality you love. And they're super easy to customize Liveает, however you like.
- 33:58
- Unidentified speaker (speaker_1): Cover new Cinnamon Dolce and the full range of Starbucks by Nespresso.
- 34:11
- Unidentified speaker (speaker_0): Meanwhile,
- 34:13
- Unidentified speaker (speaker_0): the global economy is fragile at the moment.
- 34:16
- Unidentified speaker (speaker_0): know, ill-do with a massive markets shock, wherever it starts from A.R.M., and, you know, an AI tech market shock, we could definitely live. WITHOUT THE BANK FOR INTERNATIONAL SETTLEMENTS WHICH IS
- 34:32
- Unidentified speaker (speaker_0): THE SORT OF OVERRIDING POLICEMAN OF WHERE POCKETS OF RISKS LIE WITHIN THE GLOBAL FINANCIAL SYSTEM IS WORRIED THAT THERE IS NOT ENOUGH
- 34:42
- Unidentified speaker (speaker_0): TRANSPARENCY AROUND WTAI, THE FINANCING OF MANY OF THESE ENORMOUS INVESTMENTS. HOW MUCH CAN WE SEE OF THE FINANCIAL RISKS?
- 35:06
- Unidentified speaker (speaker_3): The Global Financial Crisis, and the Housing NETWORK Bubble. COMPANIES INC. What happened in the housing bubble was that
- 35:12
- Unidentified speaker (speaker_3): the financial markets, using very very complex structures made mortgages widely available to people who otherwise wouldn't get mortgages. And they couldn't afford to pay for them. And when the rates reset, we went into that crisis.
- 35:28
- Unidentified speaker (speaker_3): The difference here is that
- 35:30
- Unidentified speaker (speaker_3): not unconnected companies, the high The street big businesses, the The banks that we use, the big firms we read about in the press, they're the ones that who are really now in 2026 driving the spending. So the demand to pay
- 35:47
- Unidentified speaker (speaker_3): off the depreciation and the loans on these data centres
- 35:52
- Unidentified speaker (speaker_3): seems to be there and seems to be growing in a way that may be sustainable. And that takes out quite a lot of the risk from something like this. Now, that isn't to say that
- 36:04
- Unidentified speaker (speaker_3): it will be a smooth ride.
- 36:06
- Unidentified speaker (speaker_0): So why would a
- 36:09
- Unidentified speaker (speaker_0): business with massive exposure to the
- 36:13
- Unidentified speaker (speaker_0): financing of all this data and computer development like Corweave, why would its shares fall so much? And indeed its bonds, it's ITS the debt that it's ISSUE issued
- 36:23
- Unidentified speaker (speaker_0): fall to
- 36:24
- Unidentified speaker (speaker_0): junk bond status, whether the so-called... THE SO-CALLED YIELD OF THE EFFECTIVE INTEREST RATE INC IS NOW 15-16 PERCENT.
- 36:30
- Unidentified speaker (speaker_3): COREWEAVE IS A GREAT EXAMPLE. YOU HAVE AN ECOSYSTEM OF INFRASTRUCTURE COMPANIES, SOME OF WHICH ARE REALLY DEEP AND RESILIENT, MICROSOFT, AND SOME OF WHICH ARE YOUNGER AND LESS RESILIENT AND HAVE HAD TO GO TO SPICY PLACES TO RAISE THEIR MONEY. So Corweave doesn't have the same
- 36:50
- Unidentified speaker (speaker_3): track record. The fact that their share prices have fallen is really neither here nor there. Share prices go up and down. But the bond skepticism is relevant. But Robert, you know...
- 37:00
- Unidentified speaker (speaker_0): And indeed, you know, the credit default swaps market, which we all had
- 37:04
- Unidentified speaker (UNKNOWN): to learn
- 37:05
- Unidentified speaker (speaker_0): about in 2007-8, which is basically when you insure against a business going bust. And the CDS price for Corweave is implying quite a high probability of the business collapsing over the next few years.
- 37:17
- Unidentified speaker (speaker_3): Over the next few years. But Corweave has to refinance an enormous amount of debt. And I
- 37:23
- Unidentified speaker (speaker_3): forget the numbers, but it's in the tens of billions of dollars.
- 37:27
- Unidentified speaker (speaker_3): But that refinancing doesn't start to really hit until late 2028. So you've got two NETWORK years COMPANIES to go. Now, INC, NETWORK COMPANIES I'd INC, NETWORK COMPANIES spend some time, talk to a few and
- 37:38
- Unidentified speaker (speaker_3): listen to.
- 37:57
- Unidentified speaker (speaker_3): NETWORK COMPANIES
- 37:58
- Unidentified speaker (speaker_3): INC, NETWORK COMPANIES INC, NETWORK COMPANIES
- 38:22
- Unidentified speaker (speaker_3): I think a better example is the automotive industry at the turn of the 20th century, particularly in the US.
- 38:29
- Unidentified speaker (speaker_3): Because what happened in that industry is that there was no
- 38:33
- Unidentified speaker (speaker_3): bubble that burst, no crisis, not until the depression in 1929.
- 38:38
- Unidentified speaker (speaker_3): But hundreds of companies went to the wall. The smaller firms failed, and there was consolidation. They were bought up by the three big winners, General Motors and Ford, and the last one whose name suddenly escapes me, but I'm sure there was a third, the three big Chrysler perhaps, right? And that model, I think, is quite a good one to look
- 38:58
- Unidentified speaker (speaker_3): at because why did that automotive market not go the way of the railroads or the dot-coms? It's because Americans wanted to buy cars and so demand kept pulling the industry forward even as the weaker players failed. And I think that's a good analogy for where WE ARE TODAY WHICH IS CUSTOMER DEMAND IS PULLING
- 39:18
- Unidentified speaker (speaker_3): THE INDUSTRY FORWARD. THERE ARE SOME PLAYERS WHO ARE VERY STRONG AND THERE'S A LONG TALE OF COMPANIES THAT ARE WEAKER SOME OF WHICH
- 39:25
- Unidentified speaker (speaker_3): MAY FAIL.
- 39:25
- Unidentified speaker (speaker_0): YOU KNOW WE'VE ALL GOT TO SORT OF SLIGHTLY HOPE THAT
- 39:29
- Unidentified speaker (speaker_0): THAT IS THE APPROPRIATE ANALOGY. THERE IS ANOTHER ANALOGY WHICH IS THE RAILWAY In the BOOM WHERE early days YOU KNOW of the railway boom where IN ESSENCE WE WERE LEFT WITH AN INCREDIBLY IMPORTANT infrastructure in the railways, but an enormous number of people lost an enormous amount of money.
- 39:44
- Unidentified speaker (speaker_3): If you look at 100 years of public companies in the US, 60 of them are responsible for about 60% of all the shareholder returns out of 25,000 So, companies in a century. So, in general, public companies in the US markets lose money over the long term for their shareholders. holders. So, that's not
- 40:03
- Unidentified speaker (speaker_3): surprising. I think the thing that I would be more worried about would be whether the demand from
- 40:11
- Unidentified speaker (speaker_3): customers, which today seems very strong and growing very quickly, and it seems that not many businesses have really got their teeth into AI, whether All of that demand
- 40:22
- Unidentified speaker (speaker_3): grows sustainably and quickly enough over the next four or five years.
- 40:27
- Unidentified speaker (speaker_0): Cool. We still, by most people's standards and certainly by British standards, quite a big business.
- 40:33
- Unidentified speaker (speaker_0): But as you say, in the overall scale of the AI
- 40:37
- Unidentified speaker (speaker_0): digital ecosystem, not that huge, but Oracle is a very big company
- 40:40
- Unidentified speaker (UNKNOWN): and there
- 40:41
- Unidentified speaker (speaker_0): has been contagion to Oracle, which is another big Thank you. investor in data centers. And in fact, Oracle has had to put out a statement saying it's massively scaling back its investment in data service.
- 40:52
- Unidentified speaker (speaker_0): You're a big investor in data centres.
- 40:54
- Unidentified speaker (speaker_0): in data centres. Such has been the concern. Why are you less what you don't seem to be that concerned about what's
- 40:58
- Unidentified speaker (UNKNOWN): going on
- 40:58
- Unidentified speaker (speaker_3): at Oracle?
- 40:59
- Unidentified speaker (speaker_3): I mean, I think there's there's a general question about a shift like this in technology that may change what these businesses are like. For the last 50 years, software businesses like Oracle have not had to spend very much on equipment, on capital equipment. They've been very high margin. It's all been about software margins.
- 41:19
- Unidentified speaker (speaker_3): And now, in the last four years, they've suddenly decided they have to have the biggest factories in the world with lots and lots of equipment in the ground. I think for investors, that's a shock. It's a bit like, you know, you're invested in a software business and you're getting a utility. And we saw in the last week or two,
- 41:35
- Unidentified speaker (speaker_3): Google, whose utility business, which is selling computing, has grown 100% year on year, and the company lost 11% of its share price in a day because the search advertising business that we're all familiar with stopped growing.
- 41:50
- Unidentified speaker (speaker_3): And that's what investors had invested
- 41:51
- Unidentified speaker (UNKNOWN): in. So would you be a
- 41:52
- Unidentified speaker (speaker_0): buyer of Alphabet Google
- 41:53
- Unidentified speaker (UNKNOWN): or
- 41:54
- Unidentified speaker (speaker_3): not? Well, I can't give financial
- 41:55
- Unidentified speaker (UNKNOWN): advice. No, no, but what I mean
- 41:56
- Unidentified speaker (speaker_0): is, if you look at that, do you think the market, all I mean by that is, do you think the market got it wrong?
- 42:01
- Unidentified speaker (speaker_3): I think that the market is in the process of figuring out what type of company Google is. And I think the difficult thing here, Robert, is that these companies, once they build these AI factories,
- 42:12
- Unidentified speaker (speaker_3): go from that ethereal software style business to something that looks more like a utility over the course of the next 10 years. And it may mean that their margins come down and they look like less exciting businesses than they did. Then the question is, why on earth did their executives do that in the first place, make
- 42:32
- Unidentified speaker (speaker_3): that choice? And I think they made the choice because they didn't have a choice. They didn't have a choice to stand on the sidelines and not participate in the next technology platform shift. And it may well be that even though they will build huge businesses, Amazon has said their computing infrastructure business could be a $1 trillion revenue
- 42:52
- Unidentified speaker (speaker_3): business within five or six years or seven years.
- 42:56
- Unidentified speaker (speaker_3): That's enormous, but it's not going to have the same margins as a software business does. And it may just be that we'll look back on this in 15 years and say, oh, up until 2020, the tech industry bathed in very high margins and great profits and very low levels of capital expenditure. Then AI came along and they had to
- 43:15
- Unidentified speaker (speaker_3): build all these enormous buildings all around the world and they got into a very different business. And I think that's what the market is trying to figure out right now.
- 43:23
- Unidentified speaker (speaker_0): I mean, it's so much of this. We are
- 43:27
- Unidentified speaker (speaker_0): learning as we go. You think we're still in a sort of learning phase now?
- 43:31
- Unidentified speaker (speaker_3): I think most of the answers to your questions, I could honestly say, I don't know. I don't know right now. I will have to see the data. I'll have to listen to people. We don't have a template. What we do know is
- 43:42
- Unidentified speaker (speaker_3): AI revenues are growing much faster than people thought. We also know that AI capital expenditure is growing much faster than people thought.
- 43:52
- Unidentified speaker (speaker_3): Let's look back at what we know and what we didn't know. We didn't know a lot
- 43:56
- Unidentified speaker (speaker_3): is the truth. So, I think we have to feel our way. and That uncertainty, I think, shows up in the market quite reasonably saying, we're going to demand higher standards, which is what it's done over the last four or five months.
- 44:08
- Unidentified speaker (speaker_0): Yeah. I mean, the important question, and you've hit upon it there, is
- 44:14
- Unidentified speaker (speaker_0): whether
- 44:15
- Unidentified speaker (speaker_0): the conventional means to impose discipline,
- 44:21
- Unidentified speaker (speaker_0): whether it's the market, whether it's regulators, whether it's
- 44:25
- Unidentified speaker (speaker_0): governments, uh are
- 44:27
- Unidentified speaker (speaker_0): up to the job at a time when change, technological change in particular, is just so rapid.
- 44:36
- Unidentified speaker (speaker_3): I think it's a really, really good question. My
- 44:41
- Unidentified speaker (speaker_3): sense would be
- 44:43
- Unidentified speaker (speaker_3): that Anthropic has an enormous NETWORK COMPANIES INC, amount NETWORK of COMPANIES money INC, in NETWORK its COMPANIES bank. It's got INC, $130 billion in cash.
- 44:49
- Unidentified speaker (speaker_3): they will raise perhaps a
- 45:09
- Unidentified speaker (speaker_3): NETWORK COMPANIES INC, NETWORK COMPANIES INC, NETWORK COMPANIES INC,
- 45:33
- Unidentified speaker (speaker_3): NETWORK
- 45:34
- Unidentified speaker (speaker_3): COMPANIES INC, NETWORK COMPANIES INC, NETWORK COMPANIES INC,
- 45:57
- Unidentified speaker (speaker_3): And then finally,
- 46:01
- Unidentified speaker (speaker_3): because I mean,
- 46:06
- Unidentified speaker (speaker_0): I
- 46:08
- Unidentified speaker (speaker_0): could continue this conversation for several days, but because we've got this big event
- 46:14
- Unidentified speaker (speaker_0): coming up, the
- 46:16
- Unidentified speaker (speaker_0): IPO, the listing
- 46:19
- Unidentified speaker (speaker_0): in America of Antropic. Anthropic,
- 46:22
- Unidentified speaker (speaker_0): how big a moment
- 46:24
- Unidentified speaker (speaker_0): does this feel to you?
- 46:27
- Unidentified speaker (speaker_3): Well, it's going to be a big moment for everyone in America who has a 401k pension scheme and for anyone in the UK who's sensibly bought index funds to put in their pension schemes because Anthropic is going to enter the index. It's going to be one of the bigger companies on the market. So suddenly you are tied to its fortunes NETWORK if COMPANIES you weren't INC, already. NETWORK COMPANIES So I think INC, that NETWORK that's
- 46:47
- Unidentified speaker (speaker_3): COMPANIES really INC, important. I can't quite
- 47:09
- Unidentified speaker (speaker_0): ATLAST
- 47:09
- Unidentified speaker (speaker_3): will start to get proper transparency. And at last,
- 47:13
- Unidentified speaker (speaker_3): all of the analysts on Wall Street and the city will be able to look at that in the way they look at it and ask the penetrating questions that you can't ask of a private company.
- 47:23
- Unidentified speaker (speaker_0): Right. Well, let's hope that that transparency happens before the end of the world. But as always, Azeem, it's been an
- 47:31
- Unidentified speaker (speaker_0): absolute joy speaking to you. Thank you. NU, And we'll see you again soon, I hope. I hope. And that's it from me.
- 47:39
- Unidentified speaker (speaker_0): Goodbye.
- 47:57
- Unidentified speaker (UNKNOWN): Everything's on sale. Everything you really want. Like one size Patrick Ta, Karastas, Kayali,
- 48:03
- Unidentified speaker (speaker_1): Laneige, and yeah, you're going to need a bigger bag. Shop at Sephora. Limitations apply. Must be a beauty insider. See terms at sephora.ca for complete details.
- 48:15
- Unidentified speaker (speaker_2): should you get a dog a great dane a teacup chihuahua a labradoodle golden doodle A a pool noodle maybe just a cat whatever you choose it's up to you just like it IT'S UP TO YOU TO ASK YOUR DOCTOR ABOUT WEGOVI.
- 48:30
- Unidentified speaker (speaker_1): IN 2002, RYAN WEDDING WAS AN OLYMPIC SNOWBOARDER AT THE TOP OF HIS GAME.
- 48:36
- Unidentified speaker (speaker_2): RYAN'S STYLE IS REALLY GUNNED AND GOLD.
- 48:38
- Unidentified speaker (speaker_1): AND IN LESS THAN A MINUTE, IT SLIPPED AWAY. BUT HE CONTINUED TO CHASE THE HIGH. RYAN JAMES WEDDING, HE CONTROLS ONE OF THE MOST VIOLENT DRUG TRAFFICKING ORGANIZATIONS IN THIS WORLD. I'M MICHELLE SHEPHERD AND FROM CBC PODCASTS AND THE TORONTO STAR, THE MOST WANTED OLYMPIAN.
- 48:56
- Unidentified speaker (speaker_2): THE MOST WANTED OLYMPIAN, AVAILABLE NOW ON SPOTIFY.
More from The Rest Is Money
332. How to make a million in... Crypto
Episode discussion: Anthony Scaramucci explains his shift from Bitcoin critic to investor, citing decentralization and institutional adoption.
Guest: Anthony Scaramucci (Part 1)
Speakers’ claims: Anthony Scaramucci discusses his evolution from a Bitcoin skeptic to an investor, attributing his change of heart to the realization that banks are essentially centralized spreadsheets while Bitcoin offers a decentralized alternative. He outlines Bitcoin's mechanics, including the halving cycle and fixed supply, contrasting it with government-issued currency and gold. Scaramucci addresses risks such as quantum computing and regulatory uncertainty, while distinguishing utility tokens from meme coins. He also highlights the growing institutional adoption of Bitcoin and stablecoins, arguing that superior technology will eventually overcome incumbent resistance.
331. Is Huang’s sacking the end of the AI boom?
Episode discussion: Speakers discuss Nvidia's leadership crisis, market concentration risks, and AI infrastructure sustainability.
Speakers’ claims: Speakers analyze the implications of Jensen Huang's resignation from Nvidia, noting an initial 14% stock drop that settled to 5%. They highlight Nvidia's systemic risk due to its ~8% weight in the S&P 500 and deep integration into pension funds. The discussion covers Nvidia's historical pivot to AI in 2012, its political engagements, and concerns about circular financing in the AI sector. Speakers also compare Nvidia's potential stagnation to post-Gates Microsoft and post-Jobs Apple, while noting geopolitical impacts on China and competitors like Huawei.