Amazons 25B AI Bet|Chips, Drugs, and an Apple Shake-Up

Amazon's AI Gamble

Amazon just committed up to $25 billion more into Anthropic — bringing its total stake to $33 billion in a single AI startup. But the headline number isn't the story. The story is what Anthropic agreed to give back: more than $100 billion in cloud spending on Amazon Web Services over the next decade, plus exclusive access to Amazon's in-house Trainium chips for training its Claude models.

That arrangement is strategically decisive for Amazon. The company has spent years and enormous capital building custom silicon — Trainium for AI training, Graviton for general compute — as an alternative to Nvidia's GPUs. Every hyperscaler knows that whoever controls the chip layer controls the margin. By locking Anthropic into Trainium at scale, Amazon isn't just funding an AI company. It's validating its own chip roadmap in front of every enterprise customer watching from the sidelines.

The market read it that way immediately. Astera Labs jumped on news that Amazon's Anthropic deal would require massive Trainium3 buildout, since Astera's connectivity chips sit inside exactly that infrastructure. Marvell followed for the same reason. Analysts at RBC noted both companies are directly exposed to the Amazon custom silicon ramp.

Jefferies flagged the profitability question that follows. Anthropic and OpenAI combined could account for a significant portion of the roughly 15 gigawatts of data center capacity Amazon plans to build through 2027. That implies Amazon's capital expenditure — already projected near $200 billion for 2026 — stays elevated or rises. The short-term drag on free cash flow is real. The bet is that Trainium's cost efficiency versus third-party chips eventually compresses that gap.

Amazon CEO Andy Jassy said Trainium offers "high performance at significantly lower cost," and that framing matters. If that cost advantage proves out at scale, the deal isn't just a revenue commitment from Anthropic — it becomes a live demonstration of AWS's chip economics to every future enterprise customer.

Apple's Power Transfer

While the chip ecosystem was digesting Amazon's AI move, Apple dropped a quieter shock of its own. Tim Cook announced he will step down as CEO on September 1st, handing the company to John Ternus — Apple's head of hardware engineering, the architect of the M-series chip, and the executive who oversaw the MacBook Neo launch that sold out within days.

Apple stock dipped roughly 1% on the news. That reaction is actually the more interesting signal. Cook has been CEO since 2011, when Apple was valued at $350 billion. He hands it to Ternus at $4 trillion. The question isn't whether Cook succeeded — he clearly did. The question is whether Ternus is the right CEO for the next leg, which is defined almost entirely by AI.

Wall Street's answer was swift and nearly unanimous: buy. Wedbush, Evercore, Citi, and Bank of America all maintained buy ratings with price targets ranging from $315 to $350. Morgan Stanley's Erik Woodring drew an explicit parallel to August 2011, when Jobs handed to Cook and Apple stock dragged for months before rallying 57% in the following year. Woodring called the current setup similar.

But the bull case is more specific than historical precedent. Bank of America's Wamsi Mohan argued that Ternus, having led the M5 silicon development, is uniquely positioned to pitch what Apple's hardware-centric AI strategy actually means: inference running locally on device rather than in cloud data centers. That implies better privacy, lower latency, and critically — independence from the cloud giants Apple competes with. Ternus knows that chip better than anyone alive.

The skeptics have a point too. Analyst Anurag Rana of Bloomberg Intelligence noted the transition signals "continuity rather than strategic change" — an awkward message when Apple needs to prove it has closed the AI gap with Google and Microsoft. Services growth has also disappointed under Cook, and the Vision Pro was a commercial miss. Ternus's credibility is in hardware. Whether he can drive the software and services pivot that investors have been waiting years for remains genuinely open.

One read that circulated widely on Tuesday: Apple doesn't announce a CEO change at 5pm on a Monday if it expects to miss earnings. The guidance that accompanies its upcoming quarterly report will be the first test of Ternus's era — and the first signal of whether this transition is a vote of confidence or a careful exit.

Amazon Enters Healthcare

Amazon didn't stop at chips and AI. On the same day it announced the Anthropic expansion, Amazon One Medical launched a nationwide GLP-1 management program — offering same-day access to weight-loss drug prescriptions, physician oversight, and lab tracking through its primary care network.

The move immediately hit Hims & Hers, which dropped more than 4% on the session. Hims had been on a five-day rally after a deal with Novo Nordisk resolved a legal dispute over compounded weight-loss drugs. Amazon's entry directly undercuts that agreement's value. If Amazon can deliver GLP-1 access through One Medical at scale, the differentiation Hims was selling — fast, convenient, integrated care — becomes Amazon's product description too.

Eli Lilly and Novo Nordisk also fell, though for a compounding reason. CVS separately opted out of a Medicare obesity drug coverage model on the same day, signaling that payer resistance to GLP-1 cost structures isn't softening. The drugs cost upward of $1,000 per month, and the coverage debate has dragged on for years. Amazon entering as a distribution channel doesn't solve the cost problem — it amplifies the pricing pressure on the drug makers themselves by inserting a powerful intermediary with its own negotiating leverage.

What Amazon is building here fits a recognizable pattern. It assembled One Medical through a $3.9 billion acquisition in 2022, spent two years integrating it quietly, and is now deploying it as a distribution layer for high-demand, high-margin drugs at the exact moment GLP-1 demand is accelerating. The Hims selloff reflects that Amazon's playbook — low friction access, logistics depth, and pricing pressure on incumbents — is now pointing directly at telehealth.

The weight of evidence points toward Amazon gradually compressing margins across both the telehealth and pharmaceutical distribution layers, but this only holds if regulators allow Amazon's medical data integration to continue at its current pace and if GLP-1 manufacturers don't cut direct-to-consumer deals that bypass it. If Lilly or Novo moves aggressively to build their own direct channels — as Lilly has already begun with LillyDirect — Amazon's intermediary position becomes more contested. Watch Hims's Q1 subscriber numbers on May 11 and Lilly's commentary on direct channel investment in its next earnings call. Those two data points will tell whether Amazon's entry is a compression event or a displacement event for the sector.