The next battle in AI will have less to do with model quality and more to do with customer lock-in.
For the past few years, OpenAI and Anthropic have competed relentlessly on model performance. But according to Samuel Colvin, CEO of AI startup Pydantic, the economics are changing.
“A year ago, what they cared about was revenue,” Colvin told me recently. “Now, when one assumes they’re both trying to IPO, their profit margin becomes really important.”
The problem is that competing solely on model quality is expensive. Frontier labs must spend billions training ever-better models that are soon emulated. That’s not a great recipe for durable profits.
Instead, Colvin believes OpenAI and Anthropic are increasingly focused on building products that are harder for customers to leave.
“They are doing their very best to find ways of locking people in that are not related to model quality,” he said. “That’s where I think Claude Code and Codex and all that work is coming from.”
AI coding services already look like better businesses than chatbots. Developers can quickly consume huge numbers of tokens while running Claude Code or Codex on complex projects, generating far more usage (and revenue) than a typical chat session.
These tools could also become exceptionally sticky. Claude Code and Codex help companies generate software at unprecedented speed, creating codebases that grow beyond what human developers can realistically manage. Companies could find themselves dependent on the same AI tools to maintain, update, and understand the software those tools helped create.
That may help explain why Anthropic and OpenAI are racing to turn Claude Code and Codex into broader AI-powered work platforms rather than standalone coding products. Anthropic is pushing in this direction with Cowork, while OpenAI is planning to merge Codex into ChatGPT.
The catch is that customers are moving the other way.
Consider Walmart’s home-grown coding assistant, Code Puppy. It’s designed to avoid dependence on any single AI provider, and give the retail giant more control over its codebase. The system can switch between models from OpenAI, Anthropic, Google, and others, helping Walmart control costs and reduce vendor lock-in.
That’s the tension increasingly shaping the AI market. OpenAI and Anthropic want sticky, high-margin products that keep customers inside their ecosystems. Enterprise buyers want flexibility, portability, and lower token bills.
The winners of the next phase of AI may be the companies that best navigate this conflict.
Sign up for BI’s Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.
