A fierce customer-acquisition war between OpenAI and Anthropic is driving a shift in how AI startups fund early stages, with both companies giving away millions of dollars in free computing power and token credits. OpenAI’s Sam Altman offered $2 million in API credits to startups in a Y Combinator cohort in exchange for equity via uncapped SAFE agreements. Anthropic countered with $500,000 in Claude credits with no equity requirement, a 16x jump from its previous standard allocation. OpenAI then revised its strategy, offering a baseline of $500,000 with no strings attached, while keeping an optional $1.5 million tranche available for an equity stake. Some early-stage AI voice and coding startups report receiving aggregate offers topping $3 million in total cloud computing and token credits. The strategy aims for ecosystem lock-in, as API token usage and cloud computing represent the single largest operational expense for AI-native startups. Building an AI product involves prompts, fine-tuning, and complex workflows optimized for a specific LLM architecture, creating high switching costs. Frontier labs also seek to secure future enterprise revenue by converting startups into paying clients once free tiers burn down, while facing pressure to clean up margins ahead of anticipated initial public offerings and competition from open-source models.