Amazon (AMZN) Stock Dips Even as Trainium Chips Cut Costs by 35%

TLDR

• Amazon stock fell around 2% on Tuesday despite positive news on its Trainium AI chip lineup.
• Trainium is gaining developer interest, largely because Nvidia GPUs remain hard to access.
• A key software barrier to using Trainium has been removed in recent months, per developers.
• One client cut inference costs by up to 35% by switching from Nvidia’s H100 to Trainium 2.
• Amazon CEO Andy Jassy says the chip business could be worth $50B annually as a standalone company.

Amazon stock dropped roughly 2% on Tuesday, part of a wider pullback in tech, even as fresh reports pointed to growing demand for its Trainium AI chips.

Amazon.com, Inc., AMZN

The Information reported that developers who have long relied on Nvidia GPUs are beginning to take Trainium more seriously. The main driver isn’t a sudden leap in Trainium’s performance — it’s the ongoing shortage of Nvidia chips.

Demand for Nvidia hardware from cloud providers, AI labs, and large enterprises remains sky-high. That supply squeeze is pushing some buyers to explore alternatives, including chips from AMD, Google, and Amazon.

Until recently, the software ecosystem around Trainium was a sticking point. Developers found it harder to work with compared to Nvidia’s well-established CUDA platform.

“Our response has always been the lack of software support being a barrier,” said Daniel Svonava, CEO of Superlinked. “That’s the thing that changed in the last couple months. That barrier has been removed.”

Cost Savings Driving Real Adoption

Bojan Jakimovski, machine-learning lead at Loka, said interest in Trainium picked up recently as Nvidia GPU access tightened. He noted that one of his clients moved its inference workloads to Amazon’s second-generation Trainium chips.

The results were notable. Testing showed costs could drop by up to 35% compared to Nvidia’s H100 chips — a number that’s hard to ignore for teams running large-scale inference jobs.

---

---

That said, Jakimovski was clear that Trainium isn’t the right fit for every use case. He would still recommend Nvidia hardware for large language model training, which remains one of the most resource-intensive parts of AI development.

So the picture is nuanced: Trainium is becoming a credible option for inference workloads but isn’t positioned to replace Nvidia across the board.

Jassy’s Big Number

Amazon CEO Andy Jassy has been vocal about the company’s chip ambitions. In a recent letter to shareholders, he said the custom silicon business is now “one of the top 3 data center chip businesses in the world.”

He went further, claiming that if the chip division were a standalone company, it could generate $50 billion in annual revenue.

Wall Street remains broadly bullish on Amazon. Analysts currently hold a Strong Buy consensus on AMZN based on 45 Buy ratings and one Hold over the past three months. The average price target sits at $318.23, implying roughly 24% upside from current levels.

AMZN stock was down around 2.08% in Tuesday trading, in line with broader weakness across the tech sector.