More companies are growing concerned that AI costs are rising to burdensome levels, regardless of whether they are conglomerates or startups. As a result, price has emerged as a heavyweight variable in the AI industry, alongside performance.

Demand is increasing for relatively cheaper AI models, and leading AI firms such as OpenAI and Anthropic, which are seeking IPOs as early as this year, appear to be under pressure to cut prices.

Citadel Securities said in a recent report that a switch to cheaper models appears to have affected indicators showing AI spending is declining.

The report said that even the most powerful technology cannot avoid practical constraints such as cost curves, capacity limits and diminishing returns. OpenAI CEO Sam Altman also said at a recent company event, "Cost has suddenly become a big issue."

Media reports are also increasing on companies adopting AI with a focus on cutting costs.

AI legal tech company Harvey worked with inference platform Fireworks AI and mixed Anthropic's Claude Opus with GLM 5.1, an open-source model developed by Chinese AI company ZhipuAI. It cut inference costs by a factor of 3 without lowering quality.

TechCrunch reported that Harvey co-founder Gabe Pereira said, "The definition of quality is shifting from using the most powerful model for every task to using the model that produces correct answers most efficiently."

No-code AI agent platform Lindy switched its base model from Anthropic to DeepSeek v4. According to The New Stack, Lindy founder and CEO Flo Crivello said on social media platform X (Twitter), "We moved 100 percent of Lindy traffic to DeepSeek v4. Not only did we save millions of dollars, performance actually improved in core use cases. A transformative change for the business."

Crivello previously said in April that AI inference had become the biggest cost item at Lindy, exceeding labor costs. Lindy chose DeepSeek v4 after evaluating open-source models for 6 to 9 months. The transition work was not easy. It was far more complex than expected. Crivello said, "It required 100 times more work than expected," adding that key challenges were evaluation to verify model performance in real work environments and rewriting prompts. Lindy said the situation could change again. Crivello said, "If Anthropic dramatically lowers prices in its next model, we could go back."

A recent Wall Street Journal report said that as price becomes a sensitive issue, interest has also risen in tools that allow companies to switch as needed among external AI models, in-house AI systems and open-source models.

The analysis said tools that let users apply ChatGPT or Claude to complex tasks and cheaper models to less demanding work deliver tangible cost savings. Some say they can cut the cost of tasks carried out with AI help by up to 95 percent.

Bug-detection startup Detail moved 90 percent of its workload from Anthropic's Claude and Google's Gemini to a custom model and GLM developed by Chinese company ZhipuAI. Founder Dan Robinson told the WSJ, "When we find technology that is proven effective and preferred by engineers, we look for ways to implement it cost efficiently," adding, "Right now there is truly an abundance of great open-source models."

According to OpenRouter, which processes AI queries, Chinese AI company DeepSeek has been the most used AI company since mid-May. The company said open-source token usage among its key clients grew 4 times faster than closed models from fall 2025 to spring 2026.

For OpenAI and Anthropic, which are leading the AI market, the rising number of companies scrutinising prices does not appear to be something they can simply watch.

With Anthropic also expected to cut prices, OpenAI is considering a move to lower AI fees later as a pre-emptive card. The WSJ reported that OpenAI has invested huge sums over the past year to secure computing resources at costs far below current market prices, and believes it is in a favourable position even if price competition intensifies.

For OpenAI and Anthropic, growing corporate sensitivity to AI costs is an awkward development. Both scenarios — more companies using cheaper AI models, or having to lower costs themselves to maintain their customer base — could hurt profitability. Ahead of IPOs, the two companies need to reduce as much as possible the "large-scale deficit structure" that investors point to as a problem.