Amazon opens the doors to its Trainium lab, home to the chip Anthropic, OpenAI and Apple have chosen

Amazon gave TechCrunch journalists an exclusive tour of its Trainium lab right after announcing its largest investment yet, $50 billion in OpenAI. That is no coincidence: Trainium is the chip underpinning AWS's new strategy as an AI infrastructure player.

Trainium is a line of custom AI accelerators that Amazon has been developing under the AWS brand since 2020. Trainium 2, announced in 2023, is positioned as a direct competitor to the NVIDIA H100. It is optimized for training large language models and promises comparable performance at a significantly lower cost per operation. Amazon is pouring billions into its own silicon precisely because NVIDIA dominates the AI hardware market, which makes its chips scarce and gives it pricing power.

Notably, Trainium's customers include companies that are both competitors and partners of Amazon in AI. Anthropic, a startup in which Amazon has invested up to $4 billion, actively uses Trainium to train its Claude family of models. OpenAI, which has just received $50 billion from Amazon, is also among the platform's users. The most unexpected name on the list is Apple: a company that traditionally builds vertically integrated products on its own hardware has turned to AWS Trainium for some of its AI workloads.

It is this combination of names, Anthropic, OpenAI and Apple, that turned the lab tour into an event. AWS is effectively telling the market: if the three AI organizations most demanding about hardware quality have chosen Trainium, it is neither a coincidence nor marketing.

The chip is developed by Annapurna Labs, an Israeli startup that Amazon acquired in 2015. Annapurna created the Graviton chips and the Nitro hypervisor. Trainium is the same team's next bet, this time in AI accelerators.

The fundamental difference between Trainium and NVIDIA is the business model. Trainium is available exclusively through AWS: you cannot buy it and install it in your own data center. Amazon monetizes the chip through cloud compute hours, not direct hardware sales. For customers, this is attractive: access to powerful accelerators without capital expenditure. For Amazon, it ties the world's largest AI labs to its own infrastructure.

The $50 billion investment in OpenAI is not just a financial bet on the market leader. It is a commitment by OpenAI to use AWS as its key cloud partner for training and inference. Trainium is the physical embodiment of that agreement.

Amazon is systematically building a position as the default AI infrastructure: through investments in leading labs, custom silicon and the scale of AWS. If Trainium proves performance comparable to NVIDIA's at a lower cost, the pressure on NVIDIA's dominance in AI will become significantly more real. That even Apple, the company with the most closed hardware stack in the industry, uses this chip suggests the strategy is already working.