Main asset classes: long-term behaviour analysis
The saying goes: "history does not repeat itself, but it rhymes." Serious investors would be well advised to understand how the main asset classes behaved not only over the past decade, but across the full historical record, under conditions and market cycles very different from this one. This analysis sets out that record for equities, bonds, cash, gold and residential property, by country, with data covering the last century.
Should investors add leverage to optimise their portfolio's risk and return profile?
Most investors think about borrowing roughly like this. My portfolio makes 10% a year. Money costs 5%. So if I borrow another half, I should end up nearer 12.5%. The only real question is whether I can live with the bigger swings. The sum is right. It describes one year, on average. It does not describe what happens to real money over ten years with a real lender. And it says nothing about risk: the extra volatility, the deeper worst drawdown, or how close you came to a margin call at the worst moment. This piece is about the gap between those two.
Mistral AI new funding round: what investors should know
Three billion euros at a €21 billion valuation, led by Samsung, with ASML and Nvidia following on. This article examines what probably matters most to an investor: how the sovereignty promise at the core of Mistral's moat has changed. The last few months brought a large partnership with Microsoft, the arrival of other companies' models on the platform starting with a Chinese one, and a Korean giant leading this round. I close with the valuation, and how it compares with Mistral's own past and with its peers.
Where a decade of stock returns actually landed
The last decade was a strong one for American equities, better than three-quarters of any ten-year stretch since 1928. But inside the index, the returns were startlingly concentrated. This is where a decade of returns actually landed for the thousand largest companies in the world, and why the average company and the typical one are not the same thing.
Lovable, the Fastest-Growing Company Ever?
A Swedish startup reached half a billion dollars of run-rate in nineteen months, on intelligence it does not own, and is now reported to be raising at $13.2 billion. The record and what the price implies, the business model and what it rents, whether the moat holds when the models belong to someone else, how it stands against its peers, and what return is left at $13.2 billion. An examination, not a recommendation.
The Mistral Mystery: How Much is the European AI Champion Worth?
Four funding rounds, three years, not one published revenue figure: Mistral is the rare frontier AI company that keeps its numbers to itself. In June 2026, reports put its next valuation near €20 billion. A discreet history, what the company sells and to whom, what €20 billion implies against its American and Chinese peers, and how much information an investment decision really needs.
Project Prometheus: from zero to $41 billion in 7 months
A Socratic examination of whether a seven-month-old company with no product and no disclosed revenue can be worth $41 billion. The piece treats the price not as a valuation but as an option premium, and works through it input by input: the physical-AI thesis, the bet on Bezos, the audacious plan to own industrial data by owning the companies, the Blue Origin record, the bubble question, and the size of the prize.
Is the SpaceX IPO Valuation Really That Stupid?
A Socratic examination of the largest IPO ever, and whether $1.75 trillion is a number to mock or to understand. The offering prices around $135 a share to raise roughly $75 billion, listing on Nasdaq as SPCX. What the price is a claim on, the Google and Anthropic compute deals, why a trailing revenue multiple misleads, the price-discovery and index-inclusion debate, the curse of the new listing, founder control, and the questions worth answering.
Is Anthropic worth $965 billion?
A Socratic examination of the question every potential investor in this round should ask. The Series H closed on 28 May 2026 at $965 billion post-money. Revenue and the $47 billion disclosed run-rate, the funding trajectory through Series H, the SpaceX compute deal, the void-shares notice, the DoD dispute, the AI-flation signals, the criticism of the reported metrics, and the four regimes for the next decade.