AI’s “IPO Battle of the Century”: OpenAI Strives to Catch Up as Its Gap with Anthropic Narrows

This year, the competitive landscape in the large AI model sector is undergoing a significant reversal. Anthropic, which once dominated the enterprise market, is now facing a strong challenge from OpenAI — and the ultimate stage for this contest is the approaching IPOs for both companies.

According to the latest data from OpenRouter, among the approximately 120,000 enterprises using both Anthropic and OpenAI products, spending between the two companies was roughly equal as of September. At the beginning of this year, Anthropic alone accounted for three-quarters of the total expenditure from this group.

The core driving force behind this reversal is the release of OpenAI’s GPT-5.6 series this summer — a model lineup designed around cost efficiency and followed by significant price cuts immediately upon launch. Simultaneously, Anthropic’s data retention policies for its flagship model Fable 5 have deterred some enterprise customers.

Both companies are also facing surging capital expenditure pressure and are preparing IPOs: Anthropic may go public as early as this November, with OpenAI expected to follow next year. Wall Street is closely watching to see if both firms can support their respective valuations, each exceeding a trillion USD, with sustainable business models and revenue growth.

GPT-5.6: OpenAI’s Price War Counteroffensive

OpenAI’s counteroffensive was marked by the launch of the GPT-5.6 series in June. The series includes the Sol, Terra, and Luna models, covering different capability levels and offering enterprise clients more flexible cost options. Shortly after release, OpenAI slashed the price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%.

According to The Wall Street Journal, several founders and industry analysts have regarded the launch of GPT-5.6 as an inflection point in the market landscape. Peter Walker, Head of Insights at OpenRouter, mentioned: "GPT-5.6 has opened a new track for OpenAI, and Anthropic’s current model lineup has no corresponding equivalents."

David Hsu, founder and CEO of software development platform Retool, said that earlier this year his company was switching between OpenAI and Anthropic, but the launch of GPT-5.6 was the “main catalyst" for the shift to primarily using OpenAI models. “We want to find the cheapest models because the cheaper the model, the higher our revenue,” Hsu said. He estimates that costs for using OpenAI models are currently about 20% lower than Anthropic’s.

Fable 5 Data Policy: Anthropic’s Pitfall

Beyond OpenAI’s aggressive pricing, Anthropic’s own policy missteps have also accelerated the loss of some customers. When releasing Fable 5 in early June, Anthropic announced it would, by default, retain user data for 30 days for trust and safety purposes. This policy presents compliance challenges, especially for industries dealing with sensitive data.

David Hsu noted that nearly all of Retool’s default contract terms stipulate immediate data deletion, making Fable 5’s data retention policy unfit for their use, so the company has never used the model. While Anthropic later attempted to address this by offering select clients data control options, the damage was already done.

Enterprise Clients: Cost Pressure Reshapes Decision Logic

On a broader level, enterprises’ decision logic is undergoing structural change. As more employees experiment with AI tools, enterprise cost pressure continues to mount, and companies are increasingly realizing that not every task requires the most advanced model. Some enterprises have begun incorporating low-cost Chinese open-source models as part of their multi-model strategy.

David Zhu, co-founder and CEO of AI sales platform startup Reevo, said his company is shifting its AI usage from Anthropic to OpenAI for two reasons: cost considerations and reducing dependency on a single model supplier. “The honeymoon period with being deeply tied to a single model supplier is over,” Zhu said, though he also admitted preferences could shift again at any time: “Things change too fast.”

Ara Kharazian, an economist at fintech company Ramp, observed the data of 70,000 clients under his employer in mid-September and found that corporate spending on OpenAI models surpassed Anthropic for the first time since December of last year. However, this lead was short-lived — by the end of the same week, Anthropic had regained the top spot by a narrow margin.

Anthropic: Market Enthusiasm Remains, New Model Counterattack

Despite pressures on market share, Anthropic’s appeal among the developer community remains considerable. On Wednesday morning this week, a large crowd lined up outside a warehouse in downtown San Francisco, waiting to enter Anthropic’s “Claude Founder House” event, with some arriving an hour early. The previous day, some waited up to three hours only to be turned away due to a full house; the scene was compared to Coachella music festival queues on social media.

Michael Szklarski, co-founder of video game startup ReadyM, said his company typically requires the most cutting-edge models and prefers using Anthropic’s Fable 5.1 but wishes to discuss with Anthropic engineers how to further reduce usage costs.

Under competitive pressure, Anthropic began releasing its next-generation Claude 5.5 series in late September, emphasizing lower costs and higher efficiency. The leading edge in AI is always fleeting, and the price war for enterprise clients is far from settled.