OpenAI and Anthropic Are Turning Credits Into the New AI Customer Acquisition Weapon

OpenAI's GPT-5.6 pricing and Anthropic's Sonnet 5 launch make a blunt point: frontier AI is entering a distribution war. The winner is not necessarily the model with the loudest benchmark headline. It is the vendor that can make a developer or enterprise team build habits around its pricing structure, credit policy, and workflow defaults.

That matters for Microsoft, Alphabet, Amazon, and NVIDIA because the real economic question is no longer whether AI is useful. It is which platform captures the recurring token demand once the first free or discounted experiments turn into production usage.

OpenAI published GPT-5.6 with a three-tier price ladder: Sol at $5 input and $30 output per million tokens, Terra at $2.50 and $15, and Luna at $1 and $6. Anthropic answered with Claude Sonnet 5 at introductory pricing of $2 input and $10 output per million tokens, while also making Sonnet 5 the default model across Free and Pro plans.

Anthropic also kept pushing the access side of the funnel. The company said Claude Fable 5 returned globally on July 1 and that Pro, Max, Team, and select Enterprise customers could use up to half of their weekly limits through July 7 before moving to usage credits. The pattern is familiar: subsidize the first wave, then meter the heavier users once they are hooked.

The first customer that matters is not the biggest enterprise logo. It is the team that starts routing a real workflow through a model and then keeps doing it. If a vendor can win that first workflow with credits, limits, and a low default price, it can often raise effective lifetime value later through higher-volume tiers, usage credits, or platform lock-in.

That creates a near-term margin trade-off, but it also widens the gap between model companies and pure application layers. The application layer can still win if it owns distribution or data, but the model vendors are showing they understand the same playbook.

If token prices keep falling while usage keeps rising, Microsoft, Alphabet, and Amazon can still win through distribution and cloud attachment. If demand does not accelerate enough, the price war just compresses margins and pushes the AI labs toward a harder capital cycle.

The key read is simple: if credits and discounts are only a temporary customer acquisition tool, margins can recover later. If they become the only way to grow, the whole AI stack gets harder to underwrite.