DeepSeek Tops US Business Spending Tracker

Chinese artificial intelligence startup DeepSeek led a closely watched US corporate spending index in June. According to the "trending software vendors" list compiled by corporate expense management platform Ramp, DeepSeek ranked ahead of event management firm PheedLoop and open-source model serving company Fireworks AI, as reported by the South China Morning Post on June 4.

The Ramp index tracks the month a US company first makes a payment to a software vendor, measuring new customer acquisition rather than overall usage volume. That a Chinese AI model topped the list—outpacing Western leaders OpenAI and Anthropic—has emerged as an unexpected market signal.

First-Purchase Growth vs. Market Share

This surprise is tempered by broader market data. Leading the trending vendor list differs significantly from dominant market share. According to the Ramp AI Index, DeepSeek’s actual adoption rate among US businesses stood at just 0.1% as of April, whereas Anthropic and OpenAI held 34.4% and 32.3% respectively, commanding a combined two-thirds of the market.

Consequently, DeepSeek is not the most widely used model, but rather the one experiencing the fastest growth in new corporate buyers. While the absolute volume remains small, the velocity of these first-time purchases is notable. Because Ramp monitors actual transactional records rather than survey responses, the index reflects realized corporate spending rather than purchase intent.

Data Residency Is the Real Signal

The fact that these transactions occurred directly introduces a critical detail: the destination of the payments. Ara Kharazian, lead economist at Ramp Economics Lab, noted that US companies are paying DeepSeek directly, bypassing the option to self-host the startup's open-source models on their own secure infrastructure.

These direct payments route corporate data to and from China-hosted servers. "In what is likely the clearest sign that companies are seeking cheaper alternatives to OpenAI and Anthropic," Kharazian observed, "some are willing to use lower-cost Chinese models, permitting US data to be transmitted to and from China-hosted servers." By choosing direct API access over secure self-hosting, these firms have prioritized lower pricing over data residency.

This choice elevates the development beyond a simple pricing war. For corporate workloads governed by HIPAA, FedRAMP, or financial regulatory frameworks, routing data to China-hosted servers via a direct API can trigger immediate compliance violations. While current transaction volumes are minor, legal and regulatory liabilities expand as direct payments grow. Trading data residency for short-term cost savings may ultimately incur significant long-term costs.

What 0.1% Is Really Shaking Is Price

The catalyst driving companies to cross this regulatory boundary is cost. This is not DeepSeek's first surge of interest in the US market. According to Ramp's analysis, the developer experienced a brief adoption cycle in January 2025, when its share of US business spending reached 0.3% before reverting to 0.1%.

At that time, the barriers to adoption were substantial. DeepSeek's API suffered frequent outages immediately after launch, and Bloomberg reported that hundreds of corporations blocked access to the service due to data security concerns. Furthermore, only 11 days after DeepSeek's debut, OpenAI released o3-mini, its lowest-cost reasoning model, effectively neutralizing the price differential and eliminating the incentive to accept security risks.

However, that economic equation is shifting once again. As the ongoing China AI price war drives inference costs down to nearly a quarter of Western alternatives, the financial incentive to accept these risks has returned for some enterprises. Although Ramp did not disclose precise market share figures for June, and the 0.1% adoption rate remains marginal, the shifting direction of first-time corporate purchases and the resulting flow of data present a significant question for the industry.