DeepSeek’s Permanent 75% Price Cut Changes the AI Model Cost Picture
Chinese AI startup DeepSeek has made one of the most aggressive pricing moves yet in the artificial intelligence race. The company announced that it is permanently cutting the cost of its flagship V4-Pro AI model by 75%, pushing usage prices down to a fraction of what developers were paying only weeks earlier.
That matters because AI companies around the world are dealing with two stubborn problems at once: expensive infrastructure and limited access to high-end AI chips. When a company cuts prices this sharply — and makes the change permanent — it usually points to something important happening behind the scenes.
For developers building AI apps, agents, and services, the new V4-Pro pricing could meaningfully reduce operating costs. Lower model costs can make it easier to run products, test features, and support heavier usage without costs climbing as quickly.
New DeepSeek V4-Pro Pricing Drops to 0.025 to 6 Yuan per Million Tokens
DeepSeek says V4-Pro usage now costs between 0.025 and 6 yuan per million tokens, depending on the type of workload. That is a steep drop from the previous range of 0.1 to 24 yuan per million tokens.
The size of the reduction is the real story here. A 75% cut does not just make the model cheaper. It changes how developers may think about building on top of V4-Pro, especially for products where token usage can scale quickly.
What the Lower DeepSeek V4-Pro Cost Means for Developers
Developers working on AI apps, AI agents, and AI-powered services are often sensitive to usage costs. If a model becomes significantly cheaper to run, it can lower the barrier for teams that want to experiment, expand product features, or support more users.
The pricing drop could be especially meaningful because V4-Pro had previously been much more expensive than DeepSeek’s cheaper Flash model. At launch, Pro access reportedly cost up to 12 times more, largely because advanced AI hardware remained constrained.
Huawei Ascend AI Chips May Be Playing a Bigger Role
DeepSeek did not directly explain what made the dramatic price cut possible. But industry attention is already turning toward Huawei and its Ascend AI chips.
DeepSeek previously acknowledged that limited access to high-end compute capacity forced V4-Pro pricing much higher than its cheaper Flash model. That made Pro access expensive at launch, particularly because advanced AI hardware was still difficult to obtain.
Now, those constraints may be easing.
Huawei’s Ascend 950 chips have become increasingly important for Chinese AI companies after U.S. export restrictions blocked companies like NVIDIA from selling their most advanced AI hardware inside China. If more usable AI compute is becoming available to Chinese firms, it could help explain why DeepSeek is now able to permanently lower V4-Pro pricing.
Why AI Hardware Access Matters for Model Pricing
AI model pricing is tied closely to infrastructure. When high-end compute is limited, running advanced models becomes more expensive. That cost can show up in the price developers pay to use a model.
DeepSeek’s earlier pricing reflected that pressure. The V4-Pro model was priced far above Flash because advanced compute capacity was constrained. If those limits are starting to loosen, even slightly, cheaper inference costs could follow.
The DeepSeek Price Cut Could Intensify the AI Price War
The bigger implication is simple: AI models are getting cheaper fast.
If Chinese AI companies can keep scaling performance while cutting inference costs, the global AI pricing battle could become much more aggressive. That pressure would not only affect rival Chinese startups. It could also challenge larger Western AI providers that charge significantly more for premium models.
DeepSeek’s move puts a spotlight on how quickly pricing can shift when infrastructure conditions change. A permanent 75% cut sends a clear signal that the company sees room to compete more aggressively on cost.
Pressure on Premium AI Model Providers
Premium AI models often come with premium pricing. But if a flagship model like V4-Pro becomes substantially cheaper to use, it raises expectations across the market.
Developers may begin comparing model costs more closely, especially when building apps and services that depend on large volumes of inference. In that environment, higher-priced providers could face more pressure to justify their costs.
Hardware Supply Is Still the Big Question
Even with DeepSeek’s major price reduction, the supply of hardware remains a major uncertainty.
Huawei still faces manufacturing bottlenecks because of restrictions on advanced chipmaking equipment. That means the broader AI infrastructure picture inside China is not fully settled. More affordable model access may be emerging, but hardware availability could still limit how far and how fast companies can scale.
Still, if DeepSeek’s price cuts are an early sign of improving AI infrastructure inside China, the move could mark the beginning of a much larger shift in the global AI market.