Ben Thompson argues in Stratechery that faster logistics enabled Amazon’s perishable grocery delivery

Ben Thompson argues in Stratechery that faster logistics enabled Amazon’s perishable grocery delivery

Ben Thompson’s Stratechery Update was published Tuesday, November 4, 2025. Thompson reports that Amazon CEO Andy Jassy described power as a possible bottleneck to adding cloud capacity, while saying Amazon was bringing more capacity online. Thompson hopes power generation will be a durable investment from the AI boom; he does not present that outcome as established. He says OpenAI’s AWS agreement provides capacity now and diversifies its suppliers, while asking what the deal suggests about demand beyond OpenAI. Thompson’s grocery argument is that Amazon’s improved overall logistics—not a specialized network built around Whole Foods—was the key prerequisite for perishable grocery delivery. He says his earlier explanation of Whole Foods’ role was wrong: Amazon made its broader delivery network fast enough for groceries to travel with other goods. Thompson says Amazon appears to be succeeding in perishable grocery delivery, while Jassy says the company will continue experimenting with physical formats. Thompson’s conclusion is that making the network fast enough for groceries to fit alongside other deliveries, rather than building a specialized grocery network, enabled this approach. Whether the shift in grocery-shopping habits Jassy describes will last remains unknown.

Publication: e53a813d-90bf-959c-0248-ff244e7b216a

attribution — pass: This is a single Stratechery Update by Ben Thompson, not a listing. The reading attributes the analysis to Thompson. It keeps Jassy's capacity and grocery remarks and the OpenAI–AWS deal framing as reported statements. It does not present them as established facts.

I think my explanation as to how Whole Foods would enable that was completely wrong

summary — pass: The reading covers the piece's three parts: AWS capacity and power, the OpenAI deal, and groceries. It leads the title with Thompson's grocery self-correction and keeps his admission that he was wrong. It omits some of Thompson's force: his remark that Jassy does not see chips as a limiter, his view that the deal is 'pretty small' in OpenAI terms, and his use of 'AI bubble' rather than 'boom'. These are softenings, not distortions. AWS is not expanded, but the context is understandable to a general reader.

I do think it’s notable that Jassy doesn’t consider access to chips a limiter

date — pass: evidence.originalPublished is 2025-11-04T11:00Z, which matches the dateline 'Tuesday, November 4, 2025' shown in the reading. The evidence kind is publisher-page, so no archive disclosure is needed. The trailing links to related Updates are not used. Forecasts, such as doubling capacity by 2027 and 2,300 cities, are not presented as outcomes. There is no hindsight framing.

Tuesday, November 4, 2025

support — pass: Each claim matches its excerpt. Jassy's 'maybe' about power is preserved as 'possible'. Thompson's power-generation point is framed as a hope. The diversification of OpenAI's supply base, the logistics argument and Jassy's remark about physical formats are all quoted accurately. The question and limitation accurately state that the lasting change in shopping habits is unknown. No unsupported benefits or costs are added.

And I also believe that this many year’s tradition of the weekly stock up grocery stock up is changing.

media — pass: No media was selected, and no verified media candidates were supplied. A text-only reading is appropriate.