AI Industry Faces Price War as Firms Shift to Lower-Cost Models

A new price war is emerging in the artificial intelligence sector as an increasing number of major and startup companies turn to lower-cost AI models, including open-source variants and those developed by Chinese firms. This trend is exerting pressure on leading industry players such as OpenAI and Anthropic to reduce the prices of their services.

According to the Wall Street Journal, companies are seeking to cut their inflated AI expenses by using new tools that can automatically switch between different AI models based on the task at hand. Cheaper models are employed for routine tasks, while more advanced models are reserved for complex analytical needs.

Executives utilizing these tools report that this approach can reduce the costs of certain AI-dependent operations by up to 95%.

Chinese Models Gain Momentum

The Wall Street Journal highlighted that this new system benefits from low-cost models developed by Chinese companies such as Alibaba and DeepSeek, alongside open-source models that companies can modify and train internally.

Dan Robinson, founder of Detail, a company specializing in bug detection, stated that his firm has shifted about 90% of its workloads from Claude and Gemini models to customized models and Chinese GLM models, which have proven effective at significantly lower costs.

Analysts note that this shift toward cheaper alternatives is already affecting spending patterns in the sector, indicating that competition now extends beyond technical performance to operational costs.

Pressure on OpenAI and Anthropic

Reports indicate that OpenAI is considering substantial price reductions for its services in anticipation of similar moves by its competitor Anthropic, amid growing customer demands for more affordable AI solutions.

These pressures come at a time when both companies continue to invest billions annually in the computing infrastructure required to develop and operate advanced AI models, potentially complicating future profitability.

This development revives the debate over whether low-cost models will commoditize AI access or if leading companies will maintain their edge through continuous model improvements.

Vishal Misra, Associate Dean for Computing and AI at Columbia University, remarked that many everyday applications do not require the most powerful models available, suggesting that companies’ ability to charge high premiums for AI services may gradually decline.

US Firms Enter the Competition

In response to this trend, Microsoft has introduced a range of smaller, more efficient models, while Nvidia launched the low-cost Nemotron model family and supported startups developing open-source models.

Meanwhile, Chinese models continue to expand their presence within US companies.

Data from the Vercel platform shows that DeepSeek’s share of AI usage rose from 1% in April to 17% in May, making it the most used company on the OpenRouter platform since mid-May.

Efficiency Race Intensifies

Many companies have started developing their own models based on open-source software, training them on internal data, which has helped reduce costs and, in some cases, improve performance.

Specialized platforms have also emerged that distribute tasks among multiple AI models based on cost and complexity, maximizing the use of cheaper models and switching to more advanced ones only when necessary.

Observers believe that the upcoming competition in the AI sector will be decided not only by model capabilities but also by companies’ ability to deliver strong performance at the lowest possible cost, setting the stage for a new phase of price wars in one of the world’s fastest-growing industries.