Rising AI costs push Coinbase and others to build in-house coding agents

Korean tech outlet Digital Today reports that rising AI costs are pushing Coinbase and other companies to turn to in-house coding AI agents, reducing their dependence on external large model services.

The report notes that the cost of using large model APIs is climbing quickly as enterprise adoption scales, becoming a major source of AI spending pressure and prompting technically capable companies to re-evaluate the trade-off between buying and building.

The appeal of in-house coding agents is control: companies can tailor agent behavior to their own codebases and engineering standards while keeping data and reasoning inside the organization, avoiding continuous per-token charges.

The report names Coinbase as a representative company in this shift, showing that even a leading crypto exchange sees coding agents as a direction worth internal R&D investment.

The trend matters for the AI tooling market: if more companies follow, external model API providers will face greater pricing and product pressure, and the enterprise AI development tool landscape could be reshaped.

What to watch next is how the cost-performance balance between in-house coding agents and commercial models evolves, and whether more companies disclose the investment and returns of such internal projects.