The Watson Weekly is sponsored by Avalara - the agentic AI platform automating global tax and compliance for leading eCommerce brands.
TL;DR
The Big Idea: Amazon is spending $200B in capex in 2026, free cash flow dropped from $38B to $11B, and Jassy wants you to feel good about it. The letter is a permission slip — for investors to absorb the pain, for operators to study the playbook, and for competitors to understand that Amazon is not blinking.
This Week in the Watson Weekly eCommerce Digest - April 13: Amazon USPS Deal, New BNPL Data, Eko Product Catalog Data & Revolve Returns.
The Real Agentic Commerce Infrastructure Problem Is a Warehouse in Bentonville— Everyone is fighting over who processes the payment. Nobody is asking if the product description is accurate enough to trigger one.

TOP NEWS OF THE WEEK
Andy Jassy's Letter to Shareholders: What Amazon Is Actually Telling the Market
The annual Amazon investor letter opens with a career autobiography. That is not sentiment. That is framing. Jassy is building a narrative architecture before he touches a single number: progress is not linear; squiggly lines are the point; and the companies that panic during inflection points are the ones that lose. He is immunizing shareholders against a drop in FCF before it reaches the income statement.
Now look at the numbers. Revenue grew 12% to $717B. Operating income grew 17% to $80B. Free cash flow fell 71% — from $38B to $11B — driven almost entirely by a $50.7B year-over-year increase in property and equipment purchases. That is the story in one line. Amazon is converting operating profit into infrastructure at a rate that makes short-term cash flow unreadable as a performance signal. Jassy is telling you to ignore FCF as a metric right now, and he has the receipts to back it up.
Where The Letter Earns Its Length
AWS AI revenue is at a $15B run rate three years into the wave. Jassy compares that to AWS's $58M run rate three years after its commercial launch. That is a 260x difference in starting velocity. The comparison is not apples-to-apples — AWS launched in a different capital environment, without the customer base AWS now has — but the directional claim holds. This wave is structurally larger, and Amazon is better positioned for it than it was for the last one.
The Chips Argument Deserves More Attention
Amazon's annual revenue run rate for its chips business — Graviton, Trainium, and Nitro — is now over $20B, growing at triple-digit rates year over year. If Trainium were sold externally like other leading chip companies sell theirs, Jassy suggests the implied run rate would be closer to $50B. (Hold that thought for later, BTW — this is not the last you have heard of this idea.)
Amazon is building an input-cost advantage that makes AWS structurally cheaper to operate than any competitor running on NVIDIA at list price. That is a moat argument, not a product announcement.
What The Letter Does Not Say
Tariffs. The word does not appear. Amazon operates a global logistics and retail business, and the letter's discussion of international expansion and rural delivery networks describes a company building into uncertainty without naming it. That silence is either discipline or optimism. Neither reading is entirely comfortable.
The grocery number is the sleeper line. Amazon's grocery business has grown to over $150B in gross sales in 2025, making it the second-largest grocer in the U.S. Especially coming from someone who has been critical of Amazon’s grocery strategy, that is saying something. In the end, they are doing it their way, which to me feels like the right path. Amazon imitating others never turns out too well.
THE BOTTOM LINE
The market should read this letter as a capital allocation thesis, not an earnings recap. Jassy is not defending the past year. He is filing his flight plan.

THIS WEEK ON THE WATSON WEEKLY ECOMMERCE DIGEST

THE TRUTH ABOUT LAST WEEK: The Real Agentic Commerce Infrastructure Problem Is a Warehouse in Bentonville
Everyone is arguing about payment rails. Eko is photographing refrigerators.
That is not a criticism. It is the point.
While Adyen publishes white papers on mandate-based authorization and Shopify enables storefronts on AI surfaces, a former interactive video company quietly addresses the real bottleneck in agentic commerce: product data quality. Garbage in, garbage out. An AI agent cannot confidently recommend a pillow it cannot accurately describe.
Eko's model is straightforward. Shoot everything from every angle. Own the file. Lease it back. Do not let the brand make the pillow look fluffier than it is. Walmart has put over $300M behind this thesis since 2018.
The payments industry is debating who controls the transaction layer when an AI agent makes a purchase. The more immediate question is whether the agent has enough accurate information to buy the right thing in the first place.
WHY IT MATTERS
Structured product data is not sexy infrastructure. It is also not optional. Eko figured that out before most of the industry finished naming their protocols. The capture factory is not a footnote to the agentic commerce story. It is the premise.


WATSON EVENTS & WEBINARS
B2B Webinar sponsored by Avalara, Elastic Path, and Data Realm on April 23, 2026, at 12:30 PM ET: “When your customer is on a roof, your UX problems are their problems.” Register to attend
Missed any of the Watson Webinars? From recaps to earnings and more - Watch the webinars
Highlights and sizzle from our latest NRF 2026 Watson Weekend Live! event on January 11, 2026, presented by Radial: Rewatch it!





