Amazon Eyes Alt AI Models After Anthropic Price

What changed in Amazon's Anthropic billing?

Amazon switched its Anthropic Claude integration on AWS to token-based billing in June 2026. Under the new structure, Amazon pays for Claude based on the number of tokens consumed rather than computing hours used. The Information first reported the change, citing two people with knowledge of the discussions between the two firms.

Claude models on AWS now bill through per-million-token rates or through Claude Consumption Units, priced at $0.01 each. Reports indicate some enterprise costs across the industry have climbed past $500 million monthly as token-based billing becomes standard.

An Amazon spokesperson pushed back on the framing, saying it is incorrect that changes from the expanded collaboration will increase costs.

How big is the Amazon–Anthropic deal?

The April 2026 agreement between Amazon and Anthropic is substantial. Here's what the deal contains, according to AI Weekly's reporting:

• $5 billion invested immediately by Amazon
• Up to $20 billion more tied to milestones
• More than $100 billion committed by Anthropic to spend on AWS technologies over 10 years
• Up to 5 gigawatts of Trainium compute capacity pledged for training and serving Claude

Amazon has invested more than $13 billion in Anthropic in total, building on $8 billion in prior commitments before the April round.

You might also like

Jeremy Strong Plays Zuckerberg in Social

JetBlue Drone Strike at JFK: What Happened

Sergey Brin Lost Big on NYC Rent-Stabilized Flats

SpaceX Deploys Dozens of Engineers to Grok AI

Why does the billing model matter to Amazon's bottom line?

The billing structure determines how revenue flows inside AWS. When an AWS customer runs Claude on Bedrock, Amazon splits that revenue with Anthropic. When a customer runs Amazon's own Nova model family, Amazon keeps 100% of the revenue.

That gap is the direct reason Amazon SVP Peter DeSantis has been public about closing the frontier-model gap. He told CNBC in June that Amazon's own models haven't been at the very frontier, but that the company intends to compete directly with Anthropic and OpenAI's frontier models within the next year.

Here's what we know so far: the token-pricing shift has made that internal model push more financially urgent, not less.

What alternative AI models is Amazon considering?

Amazon has been developing its own foundational AI models within AWS, branded as the Nova family. The company has pledged up to 5 gigawatts of compute power through AWS. DeSantis's public comments about competing at the frontier suggest Amazon's internal AI development has advanced enough that leadership feels comfortable making those promises publicly.

The token-based pricing shift is not unique to Amazon and Anthropic. OpenAI and other competitors have also moved to token-based pricing structures, making cost comparisons across providers more straightforward for enterprise buyers.

What other pressures is Anthropic facing?

In June 2026, the US Commerce Department imposed export controls on Anthropic's most advanced models. The restrictions reportedly stemmed from security concerns that Amazon executives themselves raised.

Separately, the White House later lifted some of those export controls, allowing Anthropic to release its model to more than 100 US institutions, according to The Information.

Anthropic has also drawn attention inside other major tech companies. Internal documents show Meta putting limits on Claude and Codex use, citing distillation concerns. And Salesforce employees have raised concerns about Anthropic's expanding presence inside Slack — a topic covered in our reporting on Anthropic Claude for Slack.

How does this compare to broader AI pricing trends?

The shift toward token-based billing is part of a wider industry move. The table below summarizes the key financial terms in play:

This pricing pressure mirrors what's happening across the enterprise software stack. Apple's recent price hikes on Mac and iPad hardware show that hardware and infrastructure costs are being passed downstream to enterprise buyers broadly, not just in AI.

The cost dynamics also affect how companies think about building versus buying AI capability — a tension visible in Amazon's internal Nova push and in how humanoid AI builders are weighing cloud AI spend against proprietary model development.

What happens next?

Amazon SVP Peter DeSantis stated in mid-June 2026 that Amazon intends to compete directly with Anthropic and OpenAI's frontier models within the next year. That is the most concrete public milestone the sources confirm.