Trade notes tell you what I did. This one tells you what I think. Strong opinions, loosely held, and definitely not advice — argue with me on eToro.

The AI trade didn’t die. The bill arrived.

The headline story of H1 was the great rotation inside AI: the hyperscalers — Microsoft, Alphabet, Amazon, Meta, Oracle — got hammered while memory makers went vertical. Microsoft suffered one of its sharpest June drawdowns in decades. Meanwhile Micron, SK Hynix, and Samsung climbed into the ranks of the most valuable companies on earth.

Here’s my critical read: this is not a bubble popping. It’s accounting reality catching up with a story. Depending on the estimate, hyperscaler capex is now tracking somewhere around $700–800 billion this year — roughly double the pace of two years ago — consuming the bulk of these companies’ operating cash flow, versus about a third of it in 2023. The market spent two years paying premium multiples for capital-light cash machines, and in 2026 it noticed they had quietly become some of the most capital-hungry businesses in history. The stocks didn’t get punished for AI failing; they got punished for AI costing exactly what building a new industrial base costs.

The memory mania is the mirror image. When the picks-and-shovels layer — chips, memory, power — is where all the money flows, you’re in the infrastructure phase of a buildout, not the payoff phase. That’s normal. It’s also where excess gets built: the KOSPI’s circuit-breaker-style selloff showed how narrow the exits get when leverage crowds into the same trade, and margin debt has been rising across the board. When the marginal buyer is leverage, I don’t short the mania — but I don’t chase it either.

So why did I buy Microsoft into that mess?

Because my process doesn’t care about narrative — it cares about quality at a fair price. A business earning elite returns on capital, with durable margins and a fortress balance sheet, fell to a forward P/E around 20 while my model says a business of that quality justifies a multiple in the low 30s. The market is pricing the capex bill as if the returns never arrive. Maybe it’s right — that’s why I sized it as a half position with the rest in reserve. But paying an average-company price for one of the best businesses ever built is the kind of bet I’ve made for fifteen years. The full reasoning is in my H1 review.

Gold: humbled, not broken

Gold had a wild half: a record near $5,600 in January, then a brutal unwind toward the low-$4,000s as war premium faded and rate-hike fears returned. That is not a broken bull case — it is what happens when insurance gets chased like a momentum stock.

My critical take on the gold-bug narrative: if you bought gold above $5,000 in January, you weren’t hedging — you were momentum trading with extra steps. Gold’s real story is quieter and, to me, more bullish long-term: central banks — China’s especially — keep accumulating as insurance against a dollar-based system they no longer fully trust. That structural bid is why the low-$4,000s have held. But the rate backdrop is no longer a clean tailwind for a non-yielding asset. My outlook: rangebound around $4,000–4,500 unless growth genuinely cracks. I treat gold as portfolio insurance, sized like insurance. When your hedge becomes your biggest winner, that’s not skill — that’s a warning about everything else.

The risk nobody’s pricing: the consumer

Everyone’s debating AI multiples while the actual economy sends quieter signals: the consumer is no longer an obvious tailwind, the savings rate sits near historic lows, and inflation is running hot again — with the AI buildout possibly adding a new pressure point through electricity, components, and device prices. The Fed is no longer a clean tailwind either: near-term hike odds eased after weak jobs data, but the market still sees a meaningful chance of tightening later this year. An expensive market priced for double-digit earnings growth, meeting a Fed that might tighten and a consumer that’s tiring, is a fragile combination. It’s why my breakout system spent most of H1 gated in paper mode, and why the money I do deploy is going into balance-sheet quality, bought on red days.

H2 in one line: respect the buildout, distrust the leverage riding on it, own quality when the crowd panics, and keep insurance boring.

This is opinion and education, not investment advice. I hold positions in $MSFT and other securities mentioned or related. Past performance is not an indication of future results.