The Capital Cycle is the official podcast that Marathon Asset Management (the London firm founded in 1986) launched in 2024, hosted by financial historian Edward Chancellor, who interviews Marathon's investors about each Global Investment Review letter — applying the firm's long-term, contrarian "capital cycle" supply-side approach.
This report warns that AI infrastructure investment (e.g., data centers) relies on financial engineering like options and IOUs, not real cash, echoing past bubbles. The author is cautious on AI. Meanwhile, European mid-cap firms with real cash flow (Intertek, DCC, easyJet) are being bought out with cash by private equity, suggesting they're undervalued. Key holdings: Iren (raised $19B via convertible bonds, but JPMorgan questions the circularity), Nvidia (got stock options from Iren, but the deal may be shaky), easyJet (private equity offered cash at a big premium to its stock price).
One-sentence summary of the author’s current market view: AI infrastructure investment currently relies on financial engineering rather than real cash, posing bubble risks; European small- and mid-cap companies are being taken private with cash, highlighting value opportunities. [Cautious/bearish on AI, optimistic on European value]
The report opens by highlighting a core contradiction: the current AI infrastructure investment boom is not driven by real cash but relies on financial engineering tools such as options and vendor notes. The author cites historical precedents from past technological revolutions (railroads in the 1840s, the TMT bubble in the 1990s), noting that "today's AI infrastructure buildout depends less on hard-earned cash than on promises, options and vendor paper..." The realization of these instruments hinges on an uncertain future. Meanwhile, European small- and mid-cap companies, which generate real cash flow and return capital to shareholders, are being taken private with cash. The author argues that investors are drawn to a "cashless AI society and future promises," but the ultimate question is "where real future returns are most likely to be generated."
Using Iren as an example, the report dissects the circular nature and high risk of the "cashless" financing model in AI investments. Iren, an Australian AI cloud infrastructure company, raised $19 billion through convertible bonds and chip-backed debt, signing a five-year, $3.4 billion contract with Nvidia. As part of the deal, Nvidia obtained the right to acquire $2.1 billion worth of Iren shares at $70 per share. The author points out that what is being exchanged here is primarily not cash but an option; not a settled transaction but a five-year framework; not realized revenue but an unearned contract backlog. Despite Iren's stock rising 27% on the announcement day, JPMorgan questioned the circularity of the deal, as Nvidia will use the capacity for its own internal needs. Iren still plans to invest $30 billion in AI infrastructure over the next 12 months. Its CEO claims "each unit of supply generates multiple units of demand," but the author believes this requires converting the contract backlog into utilized capacity, then into revenue, and finally into returns that significantly exceed costs—all before the hardware becomes obsolete. Citing a previous article, the report emphasizes that ultimately, a profitable business model capable of generating sufficient cash flow is needed, and growth alone does not guarantee this.
The report shifts focus to Europe, listing several small- and mid-cap companies being acquired with cash by private equity, demonstrating that their intrinsic value is severely undervalued by the public market. These companies include:
The core investment implication of the report is a significant "asymmetry" in the current market. On one side, AI companies with negative free cash flow (e.g., Iren) command high valuations based on future promises and circular transactions; on the other, European small- and mid-cap companies generating real cash flow are being acquired at low prices with cash due to a lack of investor interest. The author concludes that the AI frenzy must eventually be settled in cash (not promises), at which point its value will face a test. For investors, focusing on the age-old question of "where is the cash" (especially cash beyond insiders' bank accounts) is particularly important in the current environment. (Institutional bias note: As a value investor, the author's portfolio likely concentrates on undervalued European small- and mid-cap companies, hence the critical stance toward the AI boom and the tendency to emphasize the rationale for their holdings.)
| Ticker | Direction | Author's One-Sentence View | Key Data |
|---|---|---|---|
| Iren | Hold/Watch | Criticizes its "cashless" financing model, arguing that circular transactions and high valuations are unsustainable | Raised $19 billion, signed a $3.4 billion contract with Nvidia; stock rose 27% on announcement day |
| Nvidia | Hold/Watch | As the counterparty, obtained the right to buy Iren shares at $70, but the transaction is questioned for being circular | Five-year $3.4 billion contract, received $2.1 billion in stock options |
| Intertek | Add/Privatization | Believes its fair value exceeds £60, acquired by EQT for £60 in cash, a 67.5% premium | After stock fell to £35.82, acquisition price was £60 |
| DCC | Add/Privatization | Believes its free cash flow yield exceeds 8%, Marathon publicly opposes the takeover bid from KKR and Energy Capital Partners | Takeover bid at £65.25 in cash plus dividends |
| Irish Continental | Reduce/Privatization | Author rejects the CEO's €8 takeover bid, arguing "unwanted does not mean forever unwanted," but shareholders reluctantly accepted | CEO offered €8 in cash, valuation below 2018 peak |
| easyJet | Add/Privatization | Received cash takeover bids from Castlelake and Apollo successively; Apollo offered £7.15, an 81% premium | After stock fell to £3.94, Apollo offered £7.15 |