Amazon’s Trainium AI Chips Could Challenge Nvidia as AWS Explores Direct Sales to Data Centres

Amazon Web Services (AWS) is evaluating a significant strategic move that could reshape the rapidly expanding artificial intelligence hardware industry.

The cloud computing giant is reportedly considering selling its in-house artificial intelligence processors, known as Trainium chips, directly to external customers.

If the plan moves forward, it would position Amazon as a far more direct competitor to Nvidia, whose graphics processing units (GPUs) currently dominate the global AI infrastructure market.

The possibility of direct chip sales was discussed by AWS Senior Vice President Peter DeSantis during an interview with Bloomberg.

He revealed that the company is actively exploring the idea of making its Trainium AI processors available to outside organisations, enabling them to deploy the chips in their own data centres.

Although discussions are still at a preliminary stage, the development signals Amazon’s growing confidence in the capabilities of its custom-designed silicon technology.

AWS spokesperson Doron Aronson also confirmed to TechCrunch that the company remains open to the prospect of selling its chips in the future.

Historically, AWS has resisted requests to sell its processors directly to third parties, preferring instead to use the technology exclusively within its cloud ecosystem.

However, Aronson indicated that the company is now seriously considering the possibility, suggesting that a strategic shift may be underway.

A direct entry into the chip market would place Amazon in much more direct competition with Nvidia, whose processors have become the backbone of the generative artificial intelligence revolution.

Nvidia currently operates on a vastly larger scale, with an annual revenue run rate exceeding $300 billion.

Nevertheless, industry observers believe that even a $50-billion chip business would establish Amazon as one of the world’s most influential players in the semiconductor sector.

Although the discussions remain in their early phases and Amazon has not identified any potential customers or outlined specific commercial plans, the move reflects increasing confidence in its long-term custom chip strategy.

The company appears to believe that its proprietary processors have matured sufficiently to compete beyond its own cloud infrastructure and potentially attract demand from external enterprises.

A Realistic Commercial Opportunity

The idea of selling Trainium processors to third parties is consistent with comments previously made by Amazon Chief Executive Officer Andy Jassy.

In his annual shareholder letter published in April, Jassy highlighted the remarkable demand for Amazon’s custom artificial intelligence chips and suggested that direct sales to external customers could eventually become a viable business opportunity.

Jassy observed that if Amazon’s chip operations were structured as an independent enterprise that supplied processors both to AWS and to outside customers, the business could potentially generate an annual revenue run rate of around $50 billion.

He emphasised that demand for the company’s AI hardware has grown so rapidly that selling complete racks of AI chips to other organisations is no longer a hypothetical idea but a realistic commercial possibility.

Such a development would represent a substantial transformation in AWS’s business model. Until now, Trainium processors have primarily been used internally to power artificial intelligence workloads on Amazon’s cloud platform.

The company benefits not only from charging customers for computing resources but also from selling a broad range of complementary cloud services, including data storage, networking infrastructure, cybersecurity solutions, and monitoring tools.

Selling chips directly would expand Amazon’s role from cloud service provider to hardware supplier.

The Supply Challenge Facing Amazon

Despite the potential opportunity, Amazon faces a significant challenge: production capacity. Demand for Trainium processors has reportedly exceeded the company’s ability to manufacture sufficient quantities.

In the same shareholder letter released in April, Jassy revealed that the available supply of current-generation Trainium processors had sold out rapidly.

He also disclosed that demand for the upcoming Trainium4 chips has already surpassed anticipated production levels, even before the processors have officially entered the market.

This situation presents AWS with a strategic dilemma. Directly selling chips to external customers could potentially reduce availability for existing cloud clients unless Amazon substantially increases manufacturing output.

Meeting demand from both internal and external customers would require significant expansion of production capabilities and supply chain capacity.

To manufacture its custom processors, Amazon depends on semiconductor partners such as Taiwan Semiconductor Manufacturing Company (TSMC), one of the world’s leading chip foundries.

Notably, TSMC is also a crucial manufacturing partner for Nvidia and several other major technology companies.

As competition for advanced semiconductor manufacturing capacity intensifies, securing sufficient production resources could become increasingly challenging.

Intensifying Competition in AI Hardware

Amazon’s exploration of external chip sales comes at a time when competition in the artificial intelligence hardware sector is becoming increasingly fierce.

Nvidia, long regarded as the undisputed leader in AI computing infrastructure, has recently expanded its ambitions beyond graphics processing units and entered the market for AI-focused central processing units (CPUs).

This move has intensified competitive pressure across the semiconductor industry and heightened rivalry with established chipmakers such as Intel and AMD.

If Amazon ultimately proceeds with its plans to commercialise Trainium processors, it could create an entirely new battleground in the global race for artificial intelligence infrastructure.

AWS would evolve from being one of Nvidia’s largest cloud customers into one of its most formidable hardware competitors.

For now, Amazon has not committed to a specific timeline, product roadmap or commercial strategy. Nevertheless, its willingness to explore direct chip sales indicates that Trainium may soon become much more than an AWS-exclusive technology.

If the initiative materialises, it could significantly alter the competitive landscape of artificial intelligence computing and establish Amazon as a major force in the next phase of the AI revolution.