- OpenAI and Anthropic have dramatically escalated their Y Combinator startup credit programs: Anthropic raised its YC offer from $30,000 to $500,000 in free credits (no equity required); OpenAI initially offered $2 million in token credits for equity, then adjusted to $500,000 free with an optional additional $1.5 million for equity — a back-and-forth reflecting a real-time bidding war for YC’s ~200-company cohorts (four per year) that could become major enterprise customers within 2-3 years; combined, the two companies could commit up to $800 million annually in credits to YC startups alone.
- The subsidy economics are extraordinary: SemiAnalysis research shows Anthropic’s Claude Max plan ($200/month) enables API token consumption worth $8,000 at usage-based pricing, and OpenAI’s ChatGPT Pro 20x plan ($200/month) enables up to $14,000 in API value — meaning both companies are subsidizing power users by 40-70x the subscription price; Google Cloud offers startups up to $500,000 in cloud credits plus early Gemini access and occasional DeepMind engineer time; some founders report receiving over $3 million in combined credits from multiple companies — equivalent to the median US seed round, and sufficient to delay fundraising for months.
- The strategic logic is platform lock-in: startups that build on a specific AI provider’s APIs during the free-credit period tend to stay as they scale, because migration is technically costly and workflows become model-dependent; one founder stated it plainly — “If I’m choosing between a really cheap Chinese model I actually have to pay for and a very expensive Anthropic model I don’t have to pay for, I’m going to pick the Anthropic model”; the credits are also delaying some seed rounds entirely, deepening platform dependency before commercial pressure sets in.
- Anthropic’s enterprise market position accelerated sharply when Claude Code and Cowork went viral late last year, launching the agentic AI era and giving the company a dominant position in the coding-and-research enterprise segment; OpenAI struggled to match Anthropic’s coding capabilities before releasing GPT-5.4 in March and has since deployed large sales teams to push its Codex tool to startups with volume discounts — meaning competition has now shifted from model quality to subsidy depth, a race that favors whoever can burn the most cash the longest.
What Happened?
OpenAI and Anthropic have turned Y Combinator’s startup cohorts into a proxy battleground for enterprise AI market share. In recent weeks both companies dramatically escalated their credit offers: Anthropic went from $30,000 to $500,000 per YC startup with no equity required; OpenAI initially offered $2 million plus equity, then restructured to $500,000 free with an optional $1.5 million for equity. The back-and-forth reflects an intense, real-time competition for startup cohorts of roughly 200 companies per batch, four times a year. SemiAnalysis has documented the underlying subsidy math: at plan-cap usage, both companies deliver 40-70x the subscription price in actual API compute value. Some founders report receiving over $3 million in combined credits from multiple AI companies — a figure that exceeds the median US seed round and is reducing fundraising urgency at the earliest startup stages.
Why It Matters?
The credit war is an admission that the enterprise AI market is not yet won and that customer acquisition costs are extraordinarily high. Both OpenAI and Anthropic face enormous pressure to improve margins ahead of expected IPOs while simultaneously burning cash on subsidies designed to lock in the customers who will eventually generate those margins. The strategic bet is that startups built on a specific AI platform’s APIs are unlikely to switch as they scale, because migration is technically costly and enterprise workflows become deeply dependent on model-specific behaviors. If the bet is right, today’s $500,000 credit to a YC startup could generate millions in annual recurring revenue within 3-5 years. The risk is that the subsidy arms race simply defers the revenue reckoning — and that Chinese open-weight models, rapidly closing the capability gap, could erode the switching-cost moat that makes the lock-in strategy defensible.
What’s Next?
The credit war will intensify as both companies approach their IPO windows, because enterprise customer counts and ARR trajectories are what public market investors will value most. Watch for Google Cloud’s next move: its current $500,000 YC offer is competitive but less aggressive than the OpenAI/Anthropic escalation, and Google has both the balance sheet and strategic incentive to match or exceed to protect its cloud business. The more consequential long-term question is whether token-price collapse (down 90%+ since 2023) combined with subsidy-driven usage creates a structural customer base before the economics of inference become self-sustaining — or whether the industry is building dependency on pricing that is inherently unsustainable, setting up a reckoning when credits expire and real costs appear.
Source: The Wall Street Journal
