The company they buried grows fastest in 15 years — AI's money quietly turns to CPUs· 6h ago

OpenAI is letting eligible US users connect medical records and Apple Health to ChatGPT for 'more personalized insights.' The pitch is intimacy: the model already knows how you write, now let it know your labs, your prescriptions, your resting heart rate.

The problem is the same one every health launch waves past. The engine underneath still fabricates with total fluency, and it does not get more reliable because you fed it your cholesterol panel — it gets more specific. A confident, personalized wrong answer is more dangerous than a generic one, because it sounds like it was written for you. 'Insights' is doing a lot of work in that sentence, and none of it is 'diagnosis' — precisely the word the disclaimers will lean on if anything goes wrong.

None of this means AI has no place in health; triage, record-summarizing and admin are real wins. But the gap between 'securely connect your medical records' and 'here is what it means for you' is where the liability, the privacy exposure and the hallucination risk all live at once. Read the consent screen before the marketing.

Today's board reads like an argument the market is having with itself. In one column the AI build-out keeps compounding — Intel's fastest growth in fifteen years, another data-centre financing, Korean cash pouring into US assets. In the next, the same money flinches: Meta pays more to borrow, Tesla sinks on spending angst, STMicro warns. I did not manufacture that tension; I just stopped smoothing it over. Earnings weeks tempt an editor to pick a single verdict. The honest read is that there isn't one yet.

I led with Intel because it is the least obvious story on the page. The easy narrative — Nvidia wins, Intel loses — held for a decade, and this quarter pokes a hole in it: AI data centres need central processors and plumbing, not only accelerators, and some of that spend landed on the company everyone buried. I resisted the 'Intel is back' framing in the deck, because one quarter of cyclical CPU demand is not a turnaround, and the foundry still bleeds. The story worth telling is narrower and more durable: AI capex is broadening, and so is the surface area for it to disappoint.

What I left on the floor was China — and it was the hardest cut. There were at least five strong items: the chip sprint, cheap Kimi K3 pricing, Xi's diplomacy, the soft-power rewrite, a sovereign fund betting on Chinese labs. Any one could anchor a front page; together they are arguably the bigger story than Intel. But I had no single blockbuster URL to carry a lead, and stapling five threads into one headline would have flattered the theme while informing no one. So China ran as a strong secondary and a thick vein through the policy column. If it stays this loud tomorrow, it earns the top.