Amazon shares have been on fire lately. Now we know why
Amazon jumped nearly 14% last week — almost four times the gain in the S & P 500 — after the U.S. and Iran agreed to a two-week ceasefire. But Evercore ISI thinks artificial intelligence giant Anthropic is the reason the stock is going higher. Analyst Mark Mahaney in a Monday note said the potential use of the hyperscaler’s Trainium chip to train Anthropic’s new Mythos model is what is driving Amazon’s performance. In fact, Amazon stock jumped nearly 20% between March 27 — the day after news of the Mythos model first leaked — and Friday’s close. Mythos rattled the software market all last week after Anthropic said it was previewing the model to some users after the AI system was able to discover previously unknown bugs in cybersecurity systems. Concerns arose about what would happen to the infrastructure of the internet if hackers, or other people with malicious intent, use the model. While Anthropic hasn’t officially said it trained Mythos using Amazon’s Trainium chips, Evercore’s technology, media and telecommunications analyst Kevin Rippey suggested in a report that was the case. AMZN mountain 2026-03-27 Amazon shares since March 27 “Of course, we knew (or at least didn’t debate much) that Opus,” another Anthropic AI model, “was trained on Trainium, but I’d contend that investors were not prepared for something like Mythos to be possible on the n-1 generation Trainium Hardware,” Rippey wrote. Amazon Web Services CEO Matt Garman in an interview with CNBC’s Kate Rooney last week said as much when discussing Anthropic’s models. “The models, all their models that are out today, are trained on Trainium,” he said. The use of Trainium chips to develop Mythos represents a “proof of product” for Amazon’s chip offerings, demonstrating they’re stronger than investors previously thought, Mahaney wrote. “If Trainium 2 is powerful enough to help train Mythos, then Amazon’s Trainium offering may be much more robust than the market realized, in addition to having 30-40% better price performance than GPU-based instances,” Mahaney wrote, referring to graphics processing unit chips. Combined with other catalysts at Amazon — including growth in its retail and advertising divisions — the use of Trainium chips in AI could represent a similar moment as investors faced in Google parent Alphabet last September. After a favorable antitrust ruling at the time, investors began seeing Alphabet as an AI winner rather than as a loser owing to concern over its search business. Alphabet soared more than 60% over the next five months. Mahaney reiterated Evercore’s outperform rating on Amazon shares and a 12-month price target of $285, nearly 20% above Friday’s close.