DeepSeek Made Its 75% AI Price Cut Permanent, Escalating the AI Pricing War

My take: Heads up, because this is one of those that actually saves you money. What started as a promotion through May 2026, DeepSeek turned into its standing price: V4-Pro at $0.44 per million input tokens and $0.87 output, versus the $2.50 and $10 GPT-5.5 charges. More than five times cheaper, and with V4-Flash even more of a steal ($0.14 and $0.28). The company calls it an efficiency gain passed on to the customer, not a discount, and that little phrase says it all 🙃. It runs on Huawei Ascend 950 chips, zero Nvidia, which has a second reading: China is building its own AI hardware ecosystem and it is working. It reminds me of when low-cost airlines forced the big carriers to drop their fares: they did not disappear, but they could never again charge whatever they pleased. And here is my builder advice: if you use AI in your business and you are still paying two-year-old prices without shopping around, you are leaving money on the table. This price war benefits all of us. 💸