OpenAI and Anthropic are flooding startups with free computing credits, and the numbers are getting serious. One founder at an AI voice company said he received competing offers totaling more than $3 million. For context, that is roughly what a typical US startup raises in its entire first funding round, according to data from PitchBook.
The bidding war sharpened around Y Combinator, the startup accelerator that has produced companies like Airbnb, Stripe, and Dropbox. In May, OpenAI's Sam Altman offered every startup in the current batch $2 million in credits in exchange for a small equity stake. Anthropic countered with $500,000 and zero equity required, a dramatic jump from its previous $30,000 offer to the same group. OpenAI then matched at $500,000 without equity, and added an optional $1.5 million for those willing to give up some ownership.
The cloud providers are running the same play in parallel. Google Cloud offers up to $500,000 in credits along with early access to its models and occasional time with its DeepMind researchers. Microsoft and Amazon offer similar packages.
None of this is charity. This is the same strategy Amazon used to become the dominant cloud provider: get startups hooked early, and they rarely leave. Once a young company builds its product on a specific AI system, every prompt, every workflow, and every piece of software gets tuned to that provider's way of working. Switching later means rebuilding from scratch, which costs engineering time and money few growing companies want to spend.
There is a subtler element in the OpenAI deal structure worth noting. When credits are exchanged for equity, OpenAI is not just acquiring customers. It is acquiring ownership in the companies it supplies, meaning it profits twice: once from the subscription fees paid later, and again from any success the startup achieves. As AI inference costs continue to fall, the credits OpenAI hands out today will cost it progressively less to deliver tomorrow, while the ownership stakes it receives in return retain or grow their value.
For established businesses watching this, the lesson is not that credits are available to claim. Most of these programs target early-stage technology startups, not traditional operators. The relevant lesson is structural: the software tools your company builds on today are accumulating switching costs. The enterprise equivalents of these credit programs already exist through Microsoft, Google, and Salesforce, bundled into existing contracts in ways that are easier to accept than to examine.
Vendors are running switching-cost accumulation campaigns while buyers are running capability evaluations. Those are two different games, and the buyers often do not realize which one they are playing until renewal time arrives and the cost of leaving suddenly becomes visible.
Both OpenAI and Anthropic are preparing for public stock listings and need improving margins to make that story credible. Giving away credits now is expensive in the short term, but a customer who has built an entire product on your platform pays full price for years. The free credits are the acquisition cost. The lock-in is the business model.