Amazon’s Big Week: AI Chips, USPS Drama, and a Grocery Gamble

According to CNBC, Amazon had a massive news week that could signal a turning point for its stagnant stock, which is up only 4.6% this year compared to the S&P 500’s 16.4%. At its AWS re:Invent 2025 conference, CEO Matt Garman unveiled the Trainium3 chip, boasting four times the performance of its predecessor, and revealed AWS is already working on Trainium4. Separately, a Washington Post report suggested Amazon might end its U.S. Postal Service contract when it expires in October 2026, a claim Amazon refuted. In AI investment news, startup Anthropic—which Amazon has invested $8 billion in—is reportedly eyeing a massive IPO in early 2026, though the company says it has no immediate plans. Finally, Amazon is testing an ultra-fast 30-minute grocery delivery service in Seattle and Philadelphia, with Prime member fees starting at $3.99 per order.

AWS Doubles Down on Its Own Silicon

Here’s the thing about the Trainium3 announcement: it’s a classic Amazon long-game play. Sure, it’s impressive tech that diversifies their reliance on Nvidia. But the real investor focus, as CNBC notes, is on simply getting more data center capacity online, which still means buying a ton of Nvidia GPUs to catch up in the immediate AI infrastructure race. Announcing Trainium4 already is a smart confidence move, signaling they’re in this for the long haul. It shows they’re building a complete stack, but the payoff is years away. For enterprise customers, more competition in chips is always good—it promises better pricing and specialized performance over time. But right now, does it move the needle against Microsoft’s perceived AI lead? Probably not immediately.

The USPS Rumor and Amazon’s Logistics Endgame

This rumor about ditching the USPS is fascinating. Amazon already delivers more packages in the U.S. than UPS or FedEx through its own Amazon Logistics network. So cutting out the Postal Service, especially for those lightweight, last-mile deliveries in tough areas, seems like the final step in total vertical control. It would absolutely give them more operational and financial leverage. But is it worth the potential political headache and the loss of that crucial rural delivery partner? Amazon’s quick refutation suggests they know it’s a sensitive topic. For third-party sellers and customers in remote areas, this is one to watch—it could eventually mean higher costs or even service changes if Amazon fully insources everything.

The Anthropic IPO Wild Card

An Anthropic IPO in 2026 would be a huge validation event for Amazon’s AI strategy. That $8 billion investment wasn’t just for equity; it locked Anthropic into using AWS as its primary cloud. A successful public offering would essentially be a massive advertisement for AWS’s AI capabilities. It would tell the market, “Look, we backed a winner, and they built it on our cloud.” That’s powerful social proof in the enterprise sales game. The potential financial payday is nice, but the strategic boost to AWS’s reputation as the place to build serious AI is arguably worth more.

Groceries in 30 Minutes and the Margin Game

Ultra-fast grocery delivery feels like a brutally tough business, with famously thin margins. So why is Amazon, already the #2 online grocer, testing it? It’s all about locking in the Prime customer for *everything*. If you get your bananas in 30 minutes from Amazon, why ever go to the store? This is about dominating daily essentials and driving more data on shopping habits. They mention using AI for inventory and forecasting—that’s key. The only way this works is by driving down the “cost to serve” through insane logistical efficiency. It’s a massive operational challenge, but if anyone has the warehouse network and tech to maybe, possibly, make it work over time, it’s them. For other retailers, this is just another front in a war they’re already losing.