Amazon rarely builds the AI model that makes the headlines. It builds the ground the whole industry stands on. In the AI era, that may be the most powerful position of all.
Amazon, founded in 1994 by Jeff Bezos and led today by chief executive Andy Jassy, approaches artificial intelligence the way it approached cloud computing two decades ago. Rather than betting everything on being the smartest model, it is building the infrastructure and the platform that everyone else, including its rivals, relies on. Through Amazon Web Services, it offers custom chips, a marketplace of models, its own foundation models, and the tools to build with all of them. The strategy is to be indispensable, and it is working.
This is the same playbook that made AWS the backbone of the internet. Provide the essential plumbing at scale, make it easy to use, and let a million businesses build on top. Applied to AI, it positions Amazon to benefit no matter which models ultimately win.
What distinguishes Amazon is how much of the AI stack it owns. At the bottom sits custom silicon. Amazon designs its own training and inference chips, Trainium and Inferentia, which give it an alternative to relying solely on third-party processors and let it drive down the cost of AI for its customers. Trainium has moved into volume, with newer generations ramping, and that silicon independence is a strategic asset few competitors can match.
Above the chips sits Bedrock, Amazon's managed platform for building with AI, alongside SageMaker for machine learning and Amazon's own Nova family of foundation models, which now includes versions with very large context windows. And at the consumer edge, Amazon rebuilt its voice assistant as Alexa Plus, a generative-AI assistant now reaching tens of millions of households. From the chip to the chatbot, Amazon has a presence at every layer, which is a rare and durable kind of strength.
One of Amazon's smartest decisions was to make Bedrock a place where customers can choose almost any model rather than being pushed toward one. Through Bedrock, businesses can run Anthropic's Claude, Amazon's own Nova models, and a wide range of others, picking the best engine for each job. A very large number of customers, over a hundred thousand, now run Claude on Bedrock alone.
This neutrality is a feature, not a hedge. In a market where no one knows which model will lead next year, the company that lets customers switch freely captures the value regardless. Amazon does not need to win the model race. It needs to be the place where the race is run, and it has built exactly that. For enterprises wary of locking themselves to a single provider, that openness is a powerful reason to build on AWS.
The clearest sign of Amazon's ambition is its deep partnership with Anthropic. The two companies have committed to each other on a historic scale, with Anthropic pledging to run enormous workloads on AWS and Amazon investing billions and building dedicated infrastructure to support it. The centerpiece is an effort, often called Project Rainier, to construct one of the largest computing clusters in the world, built on more than a million of Amazon's own Trainium chips and scaling toward power capacity measured in gigawatts, to train and serve Anthropic's models.
It is a brilliant arrangement for both sides. Anthropic gets vast, cost-effective compute tuned to its needs. Amazon gets a flagship customer that proves its silicon at the highest level, deepens a relationship with one of the best model makers, and showcases what AWS can do. It binds a leading model lab and a leading infrastructure provider together in a way that strengthens both.
Amazon is a reminder that in any gold rush, the companies selling the tools and the land often do best of all. By owning the chips, the platform, and a marketplace open to every model, Amazon has made itself essential to the AI economy without having to win any single model contest. Its scale, its customer reach, and its silicon give it staying power that is hard to overstate.
For business leaders, the takeaway is that infrastructure is strategy. Amazon turned that insight into the dominant position in cloud computing, and it is running the same play in AI with discipline and enormous resources. Whoever builds the best models, a great deal of the world's AI will run on Amazon's machines, and that is a quietly commanding place to stand.
One of Amazon's most strategic moves in AI is the least visible: designing its own chips. Through its Trainium processors for training and Inferentia for inference, Amazon has built a credible alternative to relying entirely on third-party silicon. This matters for two reasons. It lets Amazon drive down the cost of AI for its customers, and it reduces its dependence on any single supplier in a market where the best chips are scarce and expensive.
The newest generations of Trainium have moved into serious volume, and Amazon has paired them with the networking and systems needed to link huge numbers of chips into coordinated clusters. Owning the silicon end of the stack gives Amazon a lever few competitors have. When it can offer comparable capability at lower cost, it strengthens the case for building AI on its cloud rather than anyone else's, and it captures more of the economics of every workload that runs on its hardware.
This is classic Amazon. The company has long preferred to build the unglamorous foundations that others depend on, from warehouses to cloud servers, and then let a thousand businesses build on top. Custom AI chips are the same instinct applied to the defining technology of the moment, and they position Amazon to benefit from the AI boom regardless of which models or applications ultimately win.
Amazon's platform strategy centers on Bedrock, its managed service for building with AI, and its defining feature is choice. Through Bedrock, a company can run Anthropic's Claude, Amazon's own Nova models, and a wide range of others, picking whichever best fits each task. A very large number of customers, well over a hundred thousand, run Claude on Bedrock alone, a sign of how readily businesses have embraced the approach.
This neutrality is a deliberate and shrewd strategy. In a market where no one can be sure which model will lead next year, the company that lets customers switch freely captures the value no matter how the race unfolds. Amazon does not need to win the contest to build the single best model. It needs to be the place where businesses come to use whichever model is best, and that is exactly what Bedrock has become.
Amazon has rounded out the platform with its own capable Nova models, including versions able to handle very large amounts of context at once, and with tools like SageMaker that help companies build and manage their own machine learning. The result is a full menu, from raw infrastructure to ready-made models to development tools, that meets businesses wherever they are in their AI journey.
Amazon also took on one of the harder challenges in consumer AI: rebuilding a product that hundreds of millions of people already owned. Its reimagined voice assistant, powered by generative AI, brings far more natural and capable conversation to the devices in people's kitchens and living rooms, and it has rolled out broadly to a large audience. Upgrading an established product that people use every day is in some ways harder than launching a new one, and doing it at this scale is a notable achievement.
The effort matters strategically because it gives Amazon a direct, everyday relationship with consumers through AI, complementing its dominant position in cloud infrastructure. Few companies operate across both the foundational layer that powers the industry and the consumer layer that reaches people at home, and that breadth gives Amazon an unusual vantage point on how the technology is actually used.
Above all, Amazon's strength rests on AWS, the cloud platform that already runs an enormous share of the world's computing. For a great many businesses, the natural place to build AI is the cloud they already trust with everything else, and that gives Amazon a built-in advantage that is hard to overstate. The relationships, the security, the familiarity, and the global infrastructure are already in place. Adding AI to that foundation is far easier than building the foundation from scratch.
The deep partnership with Anthropic crowns the strategy. With Anthropic committing to run vast workloads on AWS and Amazon building dedicated, enormous infrastructure to support it, the arrangement binds a leading model maker to a leading infrastructure provider in a way that strengthens both. It proves Amazon's silicon at the highest level, deepens a crucial relationship, and showcases what AWS can do.
For business leaders, Amazon is a reminder that in any technology gold rush, the companies supplying the tools and the land often prosper most of all. By owning the chips, the platform, and a marketplace open to every model, and by anchoring it all in the world's leading cloud, Amazon has made itself indispensable to the AI economy without having to win any single contest. That is a quietly commanding place to stand, and the company built it on purpose.
Part of what makes Amazon's position so strong is the sheer scale and reach of its infrastructure. AWS operates data centers around the world, which lets it serve customers wherever they are and keep their data close to home when regulations or preferences demand it. That global footprint is expensive and slow to build, which is precisely why it is such a durable advantage. A company starting today could not replicate it quickly at any price.
Amazon has signaled that it intends to keep investing on a massive scale, including large commitments to build AI capacity for government and public-sector use. These are long-horizon bets, the kind a company makes only when it believes demand will keep growing for years. They reflect confidence that AI is not a passing wave but a lasting shift in how computing works, and that the infrastructure to support it will be needed for a long time.
The patient, platform-first approach is quintessential Amazon. The company rarely chases the most glamorous headline. It builds the foundations, lowers the costs, widens the choices, and lets the value accumulate over years. Applied to AI, that discipline has made it one of the indispensable companies of the era, the quiet backbone on which a great deal of the industry runs.
For business leaders, the lesson is that infrastructure is strategy. Amazon turned that insight into the dominant position in cloud computing, and it is running the same play in AI with patience and enormous resources. Whoever builds the best models, much of the world's AI will run on Amazon's chips, in Amazon's data centers, through Amazon's platform, and that is a remarkably powerful place to stand.
Jason Kumpf follows the AI industry for what it means to business. He is Head of US Revenue at Razorpay, a board advisor, angel investor, and speaker. More about Jason.