davidkwilliams

Nothing has settled yet. Based on 0 decided calls and 0% coverage, data through 2026-07-22, there is no verified track record to weigh — every one of the 200 sampled posts sits in 'pending.' The AI-infrastructure commentary is recent and dense, but none of it has resolved into a scored win or loss. Treat the ANTE 'Building' band literally: the ledger is open, the ink still wet. Judge the craft and voice below, not a record that does not yet exist.

Williams runs a thematic, top-down school: the AI-infrastructure supercycle, traced from GPUs to memory to photonics to NeoCloud. Posts open with a hook — 'The market should pay attention' — then stack a real why: earnings beats, analyst targets, Oracle's backlog, Dell's server growth. The references are genuine. But the format is narrative, not instruction: you get a thesis and a map of the supply chain, rarely an entry, target, or exit. It educates conviction; it does not hand you a trade ticket.

Here the silence is loud. Zero of the sampled posts state a stop, and zero state a timeframe — the pack's discipline fields are both empty. Because nothing has been adjudicated, there is no realized loss to inspect and no way to see whether this author cuts or averages down. The commentary format sidesteps drawdown entirely: you are shown a thesis, never a position with a documented exit. For capital, that is the core caution — enthusiasm arrives fully formed, but the risk plan does not.

Two people live in this corpus. The 2026 voice is a breathless AI-infrastructure evangelist — sirens, charts, 'THIS IS INSANE,' every bottleneck a revelation. Dig back and you find a Forbes leadership columnist and author, years of gratitude quotes and 'never quit' aphorisms, plus a stretch of sharp political posting. The tone is earnest, occasionally hyperbolic, never sneering. No Discord, newsletter, or masterclass funnel surfaces in the pack. With 711,906 followers, the reach is real; the promotional apparatus, at least here, is not.

As an explainer, Williams is useful. A newcomer to AI infrastructure will genuinely learn the map — why memory matters, why power and land gate compute, why custom silicon threatens the incumbents. He cites others generously: HSBC, BofA, Goldman, Tom Lee's roadmap. What he does not teach is method — position sizing, when to exit, how to be wrong. It is theme education, not craft education. And with nothing yet adjudicated, there is no track record of owning a losing call to model humility from.

You'll like this if you want a fluent, high-energy tour of the AI-infrastructure trade — someone to explain why NeoCloud, memory, and photonics matter, whose earnings and analyst call-outs give you threads to pull yourself. You'll like it if you supply your own entries and risk. Skip it if you need if-then plans with stated stops and horizons, or if you want a verified, decided track record before you weight anyone's conviction — because right now, honestly, there isn't one to point to.