The Chinese AI upstart is making its flagship model dramatically cheaper, a move that signals rising confidence in local compute supply, intensifies pressure on rivals and sharpens the global battle over the economics of artificial intelligence.

Chinese artificial intelligence startup DeepSeek has announced that a 75 per cent price cut on its flagship V4-Pro AI model will become permanent, keeping the model’s API costs at just one-quarter of their previous level.

The decision marks a significant escalation in China’s fast-moving AI market, where model performance, chip access, pricing power and developer adoption are increasingly locked in a single contest. For enterprise users and developers, the move lowers the cost of building AI-powered applications. For competitors, it raises a harder question: how long can high-end AI remain expensive when aggressive challengers are prepared to turn intelligence into a commodity?

According to the company, V4-Pro API pricing has been reduced to between 0.025 yuan and 6 yuan per million tokens, depending on the usage type. That compares with the previous range of 0.1 yuan to 24 yuan per million tokens. A token is the basic unit of text processed by an AI model — essentially the language “fuel” consumed when users send prompts and receive responses.

The scale of the cut is notable. It suggests DeepSeek is no longer treating low pricing as a temporary adoption strategy, but as a structural weapon in the AI platform economy.

Cheaper Intelligence, Bigger Adoption

In artificial intelligence, price matters almost as much as performance.

For banks, fintechs, manufacturers, media companies, software developers and government agencies, the cost of tokens determines how widely AI can be embedded into products and operations. Expensive models encourage selective use. Cheap models encourage experimentation, automation and scale.

By permanently cutting V4-Pro pricing, DeepSeek is attempting to expand its developer ecosystem and accelerate enterprise usage. If the model is good enough for high-volume tasks — document processing, coding assistance, customer support, search, translation, analytics and content generation — then lower pricing could help it win market share even in segments where global rivals retain stronger brand recognition.

This is the classic platform play: reduce the entry cost, increase usage, capture developers, and build dependency around tools, workflows and integrations.

The Huawei Ascend Factor

DeepSeek did not confirm whether the permanent price reduction was linked to improved access to Huawei’s Ascend 950 AI chips, which the company had previously identified as central to maximising V4’s performance.

That question matters because compute supply is now one of the defining constraints in global AI competition.

U.S. export controls have restricted Nvidia’s ability to sell its most advanced AI semiconductors into China. That has created a strategic opening for Huawei’s Ascend chips, which have become an important domestic alternative for Chinese AI companies. However, separate restrictions on advanced chipmaking equipment have limited Huawei’s ability to scale production as quickly as demand requires.

When DeepSeek introduced V4, it said the Pro version would cost substantially more than its less powerful Flash model because of constraints in high-end compute capacity. The company also indicated that pricing could fall sharply once Huawei Ascend 950 supernodes became available in large quantities in the second half of the year.

The new permanent price cut suggests either that DeepSeek has gained more confidence in its compute pipeline, has improved model efficiency, or is willing to accept thinner margins to accelerate adoption. In reality, it may be a combination of all three.

China’s AI Strategy Moves from Catch-Up to Cost Disruption

DeepSeek’s pricing decision fits a broader pattern in China’s technology economy. Chinese firms have often competed not only by matching global technology, but by compressing costs at scale. In solar panels, batteries, electric vehicles, telecom equipment and e-commerce infrastructure, China has repeatedly shown how rapid industrial scaling can turn once-expensive technologies into mass-market tools.

AI may now be moving into the same phase.

The strategic logic is clear. If China cannot always access the world’s best chips, it must get more performance out of domestic hardware and make AI services cheaper enough to attract mass use. Lower model pricing can offset hardware constraints by creating a larger ecosystem around locally available compute and software optimisation.

This is not simply a commercial move. It has geopolitical consequences. The country that makes AI cheaper and easier to deploy can shape the next wave of digital infrastructure across enterprises, governments and emerging markets.

Pressure on Global AI Rivals

A permanent 75 per cent cut will put pressure on rival Chinese AI companies and may influence global pricing expectations.

The AI industry has so far been built around enormous capital spending, expensive compute clusters and a race to train increasingly capable models. That cost structure has supported premium pricing by leading providers. But if capable models become dramatically cheaper, the market may begin to split.

At the top end, frontier models will continue to command premium prices for highly complex reasoning, coding, multimodal tasks and enterprise-grade reliability. But below that frontier, pricing competition could become brutal. Many customers do not need the most powerful model for every task. They need affordable, reliable performance at scale.

DeepSeek’s move pushes the market toward that reality.

The Developer Economy Is the Real Battleground

For DeepSeek, the prize is not merely revenue from API calls. It is developer loyalty.

Once developers build products around a model’s pricing, latency, interface and performance characteristics, switching costs begin to rise. A cheaper model can quickly become embedded in thousands of applications, especially among startups and small companies that are sensitive to cost.

This is why pricing is strategic. A lower token cost can bring in a larger base of users, generate more feedback data, expose model weaknesses faster and accelerate product improvement.

The AI company that becomes the default infrastructure for builders gains influence far beyond its immediate revenue.

Risks Behind the Price Cut

The move is not without risk.

A dramatic price reduction can raise questions about profitability, sustainability and quality. If demand surges faster than compute capacity, service reliability may suffer. If margins are too thin, the company may need external funding or cross-subsidies to sustain the strategy. If competitors respond with their own cuts, China’s AI market could enter a price war that benefits users but punishes weaker companies.

There is also the issue of trust. Enterprise customers will not choose an AI provider based on price alone. They will also consider data security, uptime, compliance, model governance, integration support and long-term viability.

DeepSeek must therefore prove that its cheaper AI is not merely affordable, but dependable.

BRANDECONOMY Insight

DeepSeek Is Forcing AI to Confront Its Real Business Model

DeepSeek’s permanent 75 per cent price cut is more than a promotional discount. It is a direct challenge to the economics of artificial intelligence.

The global AI boom has been defined by giant capital expenditure, scarce chips, premium APIs and extraordinary investor expectations. DeepSeek is asking whether that model can survive a world in which capable AI becomes dramatically cheaper.

If the DeepSeek V4-Pro model can deliver strong performance at a fraction of earlier pricing, the implications will be far-reaching. Developers will experiment more. Enterprises will deploy AI more widely. Competitors will face pricing pressure. And the market will begin to distinguish between models that are truly premium and models whose value rests mostly on scarcity.

For China, the move also carries industrial significance. U.S. chip restrictions were designed to slow China’s AI ascent. But they may also be accelerating a domestic efficiency race — forcing Chinese firms to optimise models, use local chips more aggressively and compete on cost.

For Africa and Nigeria, the lesson is practical. Cheaper AI lowers the barrier to adoption for banks, media companies, public institutions, fintechs, schools, health providers and SMEs. If model costs continue to fall globally, emerging markets will have a better chance to use AI not as a luxury technology, but as productivity infrastructure.

The next AI race may not be won only by the company with the biggest model. It may be won by the company that makes intelligence affordable enough to become ordinary.