Let’s be honest — watching Amazon stock crawl sideways while the rest of the Magnificent Seven sprinted ahead has felt painful for a lot of us holding the name. Up just 4.6% on the year when the S&P 500 is up over 16%? That stings. But something shifted this week. A barrage of announcements hit the wires, and suddenly the narrative feels different. Maybe, just maybe, the market has been sleeping on a turnaround that’s already in motion.
The Week That Could Change Amazon’s Trajectory
I’ve followed this company for years, and I can’t remember the last time so many meaningful updates landed in the space of a few days. New custom AI silicon, whispers of a multi-billion-dollar IPO gains, ultra-fast grocery tests, and even quiet talks about cutting out the middleman in shipping. Each story on its own is interesting. Taken together, they start to paint a picture of a company quietly repositioning itself for the next leg up.
Trainium3 and the Long Game Against Nvidia
The star of the show was clearly AWS re:Invent 2025 in Las Vegas. The new Trainium3 chip stole headlines — four times the performance, four times the energy efficiency, and four times the memory bandwidth of its predecessor. They’re already teasing Trainium4. That’s the kind of cadence that makes hardware nerds (and investors) pay attention.
Look, nobody sane believes Amazon is going to displace Nvidia overnight. Nvidia owns the high-end GPU market for a reason. But Amazon doesn’t need to win the entire race. It just needs to reduce dependency enough to protect margins and offer customers a credible alternative when Nvidia capacity is sold out eighteen months in advance. Every Trainium chip that goes into an AWS data center instead of an H100 is pure profit protection.
“We’re not trying to beat Nvidia at their own game. We’re trying to give customers choice and make sure AWS remains the most cost-effective place to train massive models.”
— Paraphrased from AWS CEO Matt Garman’s re:Invent keynote
The other half of the AI capacity story is simpler: Amazon is spending whatever it takes to bring new data centers online. The company has openly admitted it’s behind on GPU supply. But the tone has shifted from apology to aggression. Billions are flowing into new regions, new power agreements, new everything. Capacity is the new oil, and Amazon is drilling like crazy.
Anthropic IPO: The $8 Billion Lottery Ticket
Buried beneath the chip news was a Financial Times report that Anthropic — the company behind Claude — is preparing what could be one of the largest tech IPOs ever, possibly in early 2026. Amazon has poured roughly $8 billion into Anthropic across multiple rounds, and in return became its primary cloud and training partner.
Do the math. Even a conservative $40–50 billion valuation at IPO (and many analysts think higher) means Amazon’s stake could be worth north of $15 billion on paper, maybe much more. That’s not pocket change. More importantly, a successful Anthropic debut validates Amazon’s “partner-first” AI strategy — let the best model builders focus on models while AWS focuses on scale and reliability. If Anthropic prints money, every other frontier lab starts asking why they’re still giving Google or Microsoft the lion’s share of their spend.
In my view, this is the most under-priced catalyst in the Amazon story right now. The market seems to have completely forgotten about it.
The USPS Question Nobody Saw Coming
Then came the Washington Post report (yes, ironic) suggesting Amazon is seriously considering letting its USPS contract expire in October 2026. Amazon quickly pushed back, but the denial felt more like “we’re not doing it tomorrow” than “we would never happen.”
Think about what full in-housing would mean. Amazon Logistics already moves more packages domestically than UPS or FedEx. The Postal Service mostly handles lightweight, low-value items from third-party sellers and last-mile in rural areas. Cutting out that middleman could shave billions off the cost-to-serve over time. Even a 2–3% improvement in net shipping expense drops almost straight to the operating line.
- More control over delivery quality and timing
- Rich first-party data on every mile traveled
- Potential to raise Prime fees or bundle new services
- Higher margins on the fastest-growing part of the business
Of course, there are risks — political backlash, union issues, Sunday delivery complications — but Amazon has shown it’s willing to invest decades ahead when the payoff is structural advantage. I wouldn’t bet against them figuring it out.
30-Minute Groceries: Attacking Walmart’s Moat
Quietly, Amazon rolled out tests of 30-minute delivery for fresh groceries and daily essentials in Seattle and Philadelphia. Prime members pay as little as $3.99 per order. Non-Prime? $13.99. That pricing gap is brutal — and deliberate.
Grocery has always been the white whale for e-commerce. Thin margins, cold-chain complexity, and Walmart’s iron grip on physical proximity. Yet Amazon keeps chipping away. They’re already the #2 online grocery player in the U.S., and the gap to Walmart is shrinking. Ultra-fast delivery flips the script from convenience to necessity. Once consumers experience milk and bananas on their doorstep in half an hour, the old weekly Walmart run starts feeling… archaic.
The margin story here is longer-term. Amazon doesn’t expect grocery to be highly profitable tomorrow. They expect to use AI-driven placement, robotic fulfillment centers, and sheer scale to grind costs down the same way they did with everything else. When that happens, the recurring revenue flywheel becomes terrifying.
Putting It All Together — Why 2026 Could Be Different
Here’s what ties the week’s news into a single investment thesis:
- AWS growth is reaccelerating — both from traditional cloud and new AI workloads
- Custom silicon + partner models protect margins even if Nvidia stays supply-constrained
- Logistics investments made over the last decade are reaching inflection on cost
- Grocery penetration is quietly becoming a real business line
- Balance sheet catalysts (Anthropic, potential Rivian stake sale, etc.) could unlock billions
The stock trades at roughly 36× 2026 earnings estimates right now — not exactly expensive for a company growing revenue 11–13% and potentially expanding operating margins back toward 15% over the next few years. Compare that to Meta at 27× or Alphabet at 23×, both of which face tougher regulatory headwinds.
Perhaps the most interesting aspect? Sentiment is still terrible. Most growth managers I talk to are underweight or outright avoided Amazon this year because “they’re late to AI.” That usually marks the exact moment when a great company becomes interesting again.
I’m not saying the stock rips 50% next year. But the risk/reward feels materially better today than it did six months ago. The pieces are moving. Capacity is coming. Costs are bending. And the market hasn’t fully priced any of it yet.
Sometimes the best investments are the ones everyone else gave up on.
Disclosure: The author and his clients are long Amazon shares at the time of writing.
