That’s only partly true. Amazon appears to be abandoning the most expensive and least defensible layer of the AI stack in order to strengthen the businesses where it already enjoys structural advantages: cloud computing, retail operations, advertising and logistics.
In other words, Amazon may be giving up on winning the AI popularity contest in favor of winning the AI economy. The tech and retail giant is scaling back development of several products within its Nova family, including its Premier and Omni models as well as Reel and Canvas. Some have reportedly shifted into maintenance mode, while engineering resources are being redirected toward a smaller number of advanced initiatives.
On the surface, the moves resemble a retreat. Viewed through a business lens, however, they look more like capital discipline. It also raises the stakes in Amazon’s competition with Walmart, which is applying AI to shopping, supply chains and employee productivity without attempting to compete directly in the frontier-model race.
See more: Amazon and Walmart Face a New Gatekeeper for Loyalty
Retailers Get Specific About the AI They Need for Their Operations
Amazon does not need to build the world’s most celebrated AI model to become one of the biggest economic winners from AI. It needs companies to rent its computing infrastructure, deploy models through AWS and use artificial intelligence to buy more ads, move more inventory and lower the cost of fulfilling orders.
Amazon is restructuring its artificial intelligence operation, scaling back several homegrown models and concentrating resources on a smaller number of advanced projects. The shift comes as Amazon prepares to spend heavily on AI infrastructure while trying to turn the technology into practical advantages across its cloud, advertising, logistics and retail businesses.
Whether an enterprise chooses Anthropic, Meta, Amazon or another provider, AWS benefits when customers consume more compute. Amazon Bedrock becomes more valuable as organizations seek a single environment to deploy, govern and switch among multiple models.
Amazon’s strategic rival is not only OpenAI or Google. It is Walmart.
See also: Amazon and Walmart’s Summer Sale Wars Deliver a Win (With An Asterisk)
Walmart is applying AI to product search, supply chains, employee productivity and store operations without making a major bid to become a frontier-model leader. That contrast matters because it points toward the next phase of retail AI competition.
Amazon and Walmart are unlikely to win based on which company trains the most impressive general-purpose model. They will win based on which company can connect increasingly available intelligence to proprietary commercial systems.
For Amazon, those systems include fulfillment centers, marketplace sellers, advertising inventory, Prime relationships and AWS. For Walmart, they include stores, local inventory, supplier relationships, employee workflows and purchase data.
The competitive advantage is shifting from model ownership to operational context.
Consumers now use mobile devices for 53% of purchases, but according to new data from PYMNTS Intelligence and Visa Acceptance Solutions, the bigger shift is happening before checkout. The report, “Global Digital Shopping Index: The AI-Powered Shopper Has Arrived,” revealed the ways in which the smartphone has become an in-store force. Shoppers now use their phones to discover products, compare prices, read reviews, check inventory, access loyalty offers and pay in stores.
