Amazon wants to sell its custom AI processors to customers outside its own cloud. The move could reshape competition in one of technology’s most important industries.
What happened?
For years, Amazon has been quietly developing its own artificial intelligence chips. Now it wants to sell them to the outside world.
On its October 29 earnings call, CEO Andy Jassy said Amazon plans to start selling its Trainium AI chips to customers outside Amazon Web Services in late 2027. BigGo Finance (earnings call) · Tech Times
Until now, customers have mainly accessed Trainium by renting computing power through AWS. The new plan would let outside companies buy Amazon’s hardware for their own data centers.
That puts Amazon on a potential collision course with Nvidia, the dominant supplier of the processors powering advanced AI systems.
And Amazon isn’t alone. Two days earlier, on October 27, Oracle announced it will offer Google’s Tensor Processing Units (TPUs) in Oracle Cloud starting in 2027, the first major cloud outside Google to do so. That’s different from selling chips directly, but it’s another sign of the same shift. Oracle
The AI chip market is becoming more competitive.
Why it matters
Artificial intelligence requires enormous computing power. Training sophisticated models and running AI applications depend on specialized processors capable of performing billions of calculations.
Nvidia has benefited tremendously from this demand. But companies like Amazon and Google have been developing alternatives to reduce costs and gain more control over their infrastructure. Now those alternatives are beginning to move beyond their original cloud ecosystems.
Imagine an airline that builds its own engines to reduce costs. Now imagine that airline starts selling those engines to other airlines. That’s essentially what Amazon is exploring.
Instead of simply using its chips internally, Amazon could become a supplier to other businesses building AI infrastructure.
The bigger picture: the next AI competition
Here’s where things become particularly interesting.
The next AI infrastructure battle may be about more than producing the fastest chip.
It could be about creating the most affordable and accessible computing ecosystem. Nvidia’s advantage extends beyond hardware: its CUDA software platform lets developers build and optimize applications for Nvidia processors. For Amazon to compete, developers must find its chips relatively easy to use.
That creates an important second-order effect. As alternative processors become more accessible, demand could grow for software that lets AI applications run across different chip architectures. Companies might eventually choose processors based on cost, performance and availability rather than staying dependent on one supplier.
However, switching hardware is complicated. Software compatibility, networking, memory and reliability all affect real-world performance. Cheaper chips alone won’t guarantee success.
The investment connection
Amazon’s plan could influence several industries:
- Nvidia may face more competition as alternative AI processors become more widely available.
- Semiconductor designers and manufacturers could benefit from growing demand for specialized chips.
- Data center operators may gain more choices when building computing infrastructure.
But there’s another side to the story. Amazon expects about $220 billion in capital expenditures in 2026, much of it for AI and cloud infrastructure. And even with AWS revenue growing 39.3% in the third quarter, Amazon reported negative free cash flow of $42.4 billion over the previous twelve months. CNBC · BigGo Finance (earnings call)
Jassy also pointed to the trade-off inside the chip plan itself: selling chips brings in revenue sooner, but using them inside AWS can generate more total revenue and cash over five to six years. BigGo Finance (earnings call)
Growing demand doesn’t automatically translate into stronger cash flow when infrastructure costs are enormous.
What we’re watching next
Three questions stand out:
- Will Amazon actually begin selling Trainium chips outside AWS in late 2027?
- Can alternative processors compete with Nvidia’s software ecosystem and overall performance?
- Will growing competition eventually make AI computing more affordable?
The bottom line
For years, the AI boom has been closely tied to Nvidia’s processors. That dominance isn’t disappearing overnight. But Amazon and Google are showing that the future of AI infrastructure may involve more suppliers, more computing choices and greater competition.
The next chapter of artificial intelligence may not be determined solely by who builds the smartest models, but by who makes the computing power behind them more accessible and affordable.
And that competition is only beginning.
Stay sharp, stay organized.
Educational commentary, not personalized investment advice. Companies named are examples, not recommendations. Amazon’s outside chip sales are a future plan, not a current product. Sources: Amazon Q3 2026 earnings call · Oracle, October 27 announcement.