Jim Cramer admits he was wrong about Amazon — now calls its AI chips a real threat to Nvidia

As Cramer told CNBC, these Trainium chips “have more long-term value than I thought and are more of a threat to Nvidia than I thought.” He added that the cheaper price tag for Amazon’s offering could help explain why Nvidia’s stock didn’t pop following its positive earnings call on May 20.

Cramer turned from skeptic to supporter of Amazon shortly after sitting down with the company’s CEO, Andy Jassy, for a CNBC interview. When chatting about Amazon’s chips, Jassy said the officially reported “$20 billion annual run rate” likely “underestimate[s] really the size of it.”​

Jassy went on to explain that that figure could easily balloon to a “$50 billion annual run rate” by selling the chips they’re producing this year to AWS and third parties. He also name-dropped the AI labs OpenAI and Anthropic, both of which have “multiyear, multi-gigawatt commitments to Trainium.”

Year-to-date, Amazon’s stock has been riding the broader AI surge, climbing roughly 13% at the time of writing.

Inside Amazon’s AI chip arsenal

​So, what are these Amazon chips that make investors like Cramer so excited?​

Amazon actually has three different chips that it’s marketing for different purposes in the AI infrastructure niche.

As mentioned above, the Trainium line is gaining a lot of attention because it’s a threat to Nvidia’s dominance with GPUs. These AWS-built specialized chips do the heavy lifting for generative AI and LLMs, but are way cheaper than Nvidia’s offerings.

Just how cheap? According to Amazon’s stats, Trainium2 offers “30 to 40% better price performance” than GPUs.​

And that’s not all Amazon is working on in its chip division. It also offers the “Graviton” product as a general-purpose processor.​

Graviton grabbed mainstream attention when Facebook’s parent Meta announced a huge partnership with Amazon. Although we don’t know the specifics of this deal, Morningstar reported that Meta plans to use “tens of millions” of Graviton chips.​

Then there’s the “Nitro” line of chips, which handles the behind the scenes work in AWS data centers. You could think of these chips as giving Amazon an “efficiency boost” by handling tasks like storage and security so the main processors can focus on running customer workloads.

Is Amazon really the next Nvidia?

​Amazon’s chips are already so big that they’re drawing the attention of established players like Nvidia.

CNBC reported Nvidia’s CEO Jensen Huang openly criticizing newer, more affordable chips from the likes of Amazon. As Huang said at the recent Computex 2026 conference, “Choosing the wrong architecture just because the chips are cheaper doesn’t translate.” Instead, Huang emphasized using hardware designed to deliver the most AI work per watt of power.