AI startups are being courted with a new kind of incentive: free computing power. OpenAI, Anthropic, and cloud providers are using compute credits and token credits to bring young companies into their ecosystems before those companies settle on long-term infrastructure and model providers.
The sums are large enough to matter at the earliest stage of a company. Some founders are seeing offers that rival the size of a seed financing round, turning infrastructure discounts into a strategic weapon in the race for business customers.
Startups Are Being Pulled In Early
According to a Wall Street Journal report cited by the source article, top AI companies are giving young startups large packages of cloud computing and token credits. Hans Ibarra, founder of AI voice startup Dialogus, says competing offers he has received add up in some cases to more than $3 million in cloud computing and token credits.
The source notes that this is roughly the size of an average US seed round, according to PitchBook. That comparison explains why these offers can shape a startup’s early decisions. For a young company, credits can reduce immediate spending, make experimentation cheaper, and influence which technical stack feels easiest to keep using.
The contest is not limited to OpenAI and Anthropic. Cursor, the AI coding firm acquired by SpaceX, offered a 75 percent discount through July 5. That detail shows how the competition extends beyond foundation model providers into AI coding tools and related software used by startups.
Cloud Providers Want Sticky Customers Too
Cloud companies are also competing for the same startup relationships. A Google Cloud spokesperson said the company provides up to $500,000 in credits, early access to Gemini models, and occasional access to DeepMind engineers. Microsoft and Amazon Web Services offer similar perks, according to the source article.
The strategy is straightforward: get startups using a platform early, then make the tools and workflows difficult to abandon. Once a team builds around a provider’s models, credits, cloud services, support channels, and developer tools, switching may become less attractive even after the free or discounted period ends.
For startups, that creates a useful but complicated choice. Free AI compute credits can help a company move faster and spend less cash in the short term. But the same credits may also pull technical decisions toward a provider before the company has fully tested alternatives.
- OpenAI is using token credits and is actively selling Codex to startups.
- Anthropic is competing with credits tied to its own tools and models.
- Google Cloud is offering credits, early Gemini access, and occasional access to DeepMind engineers.
- Microsoft and Amazon Web Services offer similar perks, according to the source article.
The Discount War Meets Margin Pressure
The timing is notable because OpenAI and Anthropic also need to improve their margins ahead of expected IPOs. Giving away millions in compute or token credits can win customers, but it also comes while these companies are under pressure to show stronger economics.
The competitive backdrop is getting harder. The source article says open and cheaper models are adding pressure, including many out of China. That matters because lower-cost alternatives can make generous credits more necessary for companies trying to defend or expand their customer base.
Anthropic’s revenue surged late last year thanks to Claude Code and Cowork, according to the source. OpenAI reportedly did not catch up until it released GPT-5.4 in March and is now actively selling its Codex tool to startups. In that environment, free credits are not just discounts; they are part of a broader fight to become the default provider for AI development.
Y Combinator Startups Are A Major Prize
Y Combinator companies are especially attractive targets. In May, Sam Altman announced $2 million in token credits in exchange for equity stakes. Anthropic responded with $500,000 and no equity requirement, which was a major increase from its earlier $30,000 offer.
OpenAI then matched Anthropic’s $500,000 with no equity requirement. It also added an optional $1.5 million in exchange for shares. These moves show how quickly the offers escalated once the competition focused on a concentrated pipeline of startup customers.
The source article says Y Combinator has four cohorts per year of roughly 200 companies each. On that basis, OpenAI and Anthropic could together hand out up to $800 million in credits. That figure underlines how large the customer acquisition battle could become if these offers are applied broadly across each cohort.
What The Credit Race Signals
The immediate benefit for startups is clear: AI tools and cloud infrastructure can become cheaper to test and deploy. A startup with limited capital may be able to try more models, build faster prototypes, or delay some direct infrastructure spending.
But the deeper signal is about control of the AI startup stack. OpenAI, Anthropic, Google Cloud, Microsoft, Amazon Web Services, and other providers all want young companies to build on their platforms while habits, codebases, and workflows are still forming.
That is why the offers are so large. The providers are not only selling compute. They are trying to win the point at which a startup chooses its default model, cloud, development tool, and partner ecosystem. Once that choice is embedded, the relationship can become much harder to displace.