The price war OpenAI started is now a race to zero, and a Chinese lab just made sure no one can afford to charge premium rates again.
The Summary
- DeepSeek is making permanent its 75% discount on the V4-Pro model, dropping developer pricing to one-quarter of the original rate after the promotional period ends May 31.
- The company announced the change directly on X, framing it as an "official adjustment" rather than a temporary promotion extension.
- This locks in pricing that makes DeepSeek's flagship model cheaper than most competitors' mid-tier offerings, forcing the entire API market to recalibrate what "affordable AI" means.
The Signal
DeepSeek isn't backing down. The V4-Pro pricing adjustment takes effect after May 31, converting what started as a promotional discount into the new baseline. For developers, this means the temporary deal they've been testing becomes the permanent economic foundation they can build on. No rug pull, no gradual price creep, just a 75% haircut that stays.
The timing matters. DeepSeek ran the discount long enough for developers to migrate workloads, rewrite budgets, and commit to the new price point. Now those customers are locked in, and competitors face an ugly choice: match the price and crater margins, or hold the line and watch developer mindshare evaporate. The Hacker News thread hit 287 points with 167 comments, a clear signal that the developer community sees this for what it is: the new floor.
"The 75% discount isn't a promotion anymore. It's the market price."
What makes this different from typical cloud pricing games:
- DeepSeek announced the permanence publicly, making it hard to walk back
- The discount applies to the flagship model, not a loss-leader tier
- Competitors now have to explain why their APIs cost 4x more for similar capability
This is the AI infrastructure version of AWS circa 2008: relentless margin compression in service of market share. But DeepSeek isn't Amazon. It's a Chinese lab with different incentives, different cost structures, and less pressure to show quarterly profit growth. That asymmetry is the real weapon. OpenAI and Anthropic have to justify their pricing to investors who want to see a path to profitability. DeepSeek just has to keep the API running.
The second-order effect hits hardest: every startup that built a business model around "affordable AI" just saw their moat drain. If your pitch was "we make GPT-4 class models accessible at reasonable prices," you're now competing with a permanent 75% discount on a model that benchmarks well. The arbitrage window is closing.
The Implication
If you're building on AI APIs, lock in your DeepSeek integration now before the next pricing move. The permanent discount is a signal that the company is playing for territory, not margins. Expect other providers to respond within weeks, either with matching cuts or new bundling strategies. The developers who move fast get to rebuild their unit economics before the next wave hits.
For everyone else, watch what happens to the premium AI pricing tier over the next six months. DeepSeek just made "flagship model at commodity prices" the new normal. The companies that can't compete on cost will have to compete on something else: speed, reliability, specialization, or vertical integration. Undifferentiated model access is becoming a race to the bottom, and the bottom just got 75% lower.