By Adam Clark
Artificial intelligence is getting more expensive, and that's good news for Google parent Alphabet and Amazon.com.
The latest AI model release from Anthropic, named Fable 5, has drawn attention for its costs as much as its capabilities. When priced per-token -- the basic unit of AI consumption -- Fable 5 will cost twice as much as its predecessor Opus 4.8. Those high costs are creating an opportunity for Google and Amazon to cement their cloud-computing dominance.
According to Anthropic, Fable 5 costs $10 per one million input tokens and $50 per one million output tokens. Input tokens are the text and other context a user puts into a model, while output tokens are what the model generates in response. GPT-5.5, OpenAI's most advanced model, costs $5 per one million input tokens and $30 per one million output tokens
"The cost per token on Fable 5... can very quickly consume relatively large developer budgets," wrote New Street Research analyst Dan Salmon in a research note.
While that probably will drive some developers to use cheaper AI, there will likely still be demand for the most complex and expensive models -- but they will have to run on infrastructure that limits the costs and can reliably handle the demands. That's where Google and Amazon come in with their AI platforms, known as Vertex and Bedrock, respectively.
"We see a large and increasing tailwind for Vertex and Bedrock as huge models with massive context require large infrastructure capabilities that are impractical for single developers to acquire," wrote Salmon.
The key advantage for Google and Amazon is that they have their own custom chips. Google's Tensor Processing Units and Amazon's Trainium processors can run the latest models, while also potentially delivering more cost-efficient performance. Google might have the upper hand, Salmon notes.
"Whether we see margin-taking or share-battling, custom silicon capable of handling bleeding-edge models will have a clear market advantage," Salmon wrote. "Today we see the greatest advantage for Google due to their TPUs, but AWS is also likely to see benefit from this dynamic as Trainium chips rapidly improve through iteration and co-development with Anthropic and OpenAI."
Write to Adam Clark at adam.clark@barrons.com
This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
June 12, 2026 10:28 ET (14:28 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments