A Chinese AI model has done something few industry observers predicted just eighteen months ago: it has topped the list of trending software vendors tracked by Ramp, the corporate spend management platform that monitors purchasing behavior across more than 50,000 US businesses. DeepSeek’s appearance at the top of Ramp’s June 2026 trending vendors index is not merely a data point about software preferences, it is a signal of how dramatically cost pressure and capability convergence are reshaping the American enterprise AI market, and how the US-China AI rivalry is playing out in procurement decisions far below the level of government policy.
A Surprise Even for the Analyst Tracking the Data
Ramp economist Ara Kharazian, who compiles the monthly index, did not hold back in expressing his surprise. “I didn’t expect American firms to use DeepSeek,” Kharazian said in comments accompanying the report. That candor matters. Ramp is not a technology advocacy organization interested in talking up Chinese AI. It is a financial infrastructure company whose value proposition is to give businesses accurate visibility into their spending. When its own economist is caught off guard by the data his platform is generating, the signal is worth taking seriously.
What makes the data particularly meaningful is that Ramp is tracking direct payments to DeepSeek, not businesses self-hosting the company’s open-weight models on their own servers, which would leave no trace in payment records. These are companies writing checks, or authorizing corporate card transactions, to DeepSeek as a vendor. That distinction separates this adoption wave from the broader open-source usage story that has defined much of DeepSeek’s international footprint. Enterprises are not just downloading DeepSeek’s models to run in-house. They are subscribing to the service as a commercial vendor relationship.
The Adoption Curve: Collapse, Then Resurgence
The trajectory of DeepSeek’s penetration among Ramp businesses tells a story of volatile but ultimately ascending adoption. The first wave came in January 2025, when DeepSeek’s R1 model generated enormous global attention for matching or exceeding frontier Western models at a fraction of the reported training cost. At that peak, roughly 0.3% of Ramp businesses were using DeepSeek. The number then fell sharply to around 0.1%, as initial curiosity faded, concerns about data security circulated in corporate legal and compliance departments, and several US legislative efforts to restrict or ban the service generated headlines.
That pullback reflected real institutional hesitancy. The national security debate around DeepSeek has been substantive and ongoing, with critics pointing to the Chinese company’s data handling policies and the legal obligations Chinese firms face under national security legislation. Those concerns have not disappeared. But the renewed rise in adoption, culminating in DeepSeek topping the June 2026 trending index, suggests that for a meaningful segment of American businesses, cost and performance considerations are now outweighing those reservations, or that enterprises have found compliance frameworks they are comfortable operating within.
According to reporting from the South China Morning Post, the businesses driving this latest adoption wave are concentrated in professional services sectors: US law firms, management consultancies, and financial services companies. These are not startups optimizing for cost at the expense of everything else. They are regulated, risk-conscious institutions with legal and reputational skin in the game. Their adoption of DeepSeek as a paid vendor relationship represents a considered commercial decision, not an experiment by a developer team working around procurement policy.
Price Is the Argument That Keeps Winning
To understand why DeepSeek is gaining ground in enterprise procurement, it is worth looking at what it is competing against. OpenAI currently reaches 32.3% of Ramp businesses, while Anthropic, which has invested heavily in enterprise positioning and its Constitutional AI safety narrative, sits at 34.4%. Both figures represent extraordinary commercial penetration for relatively young AI companies. But both OpenAI and Anthropic operate at price points that create meaningful friction for cost-sensitive enterprise buyers, particularly in sectors like legal services where AI usage can scale rapidly across large document volumes.
DeepSeek’s pricing structure, designed for a Chinese domestic market where model costs have been in freefall, creates a stark contrast. Chinese AI providers have been locked in an aggressive price war since mid-2024, with Alibaba Cloud’s AI products and competitors like Baidu repeatedly cutting API prices to near-zero or free tiers for high-volume tokens. That competitive dynamic has produced models and services that are structurally cheaper to operate than their American counterparts, and American enterprise buyers are beginning to notice.
The broader competitive context matters here. ByteDance has raised its 2026 AI capital expenditure to $30 billion, while Alibaba has been constructing 10,000-card AI computing clusters powered by domestically developed silicon. The scale and pace of Chinese AI infrastructure investment is producing frontier-competitive models at costs that American providers, operating on more expensive GPU infrastructure under export control constraints, are struggling to match on price.
What This Means for the AI Geopolitical Debate
DeepSeek’s rise in American enterprise spending arrives at a complicated diplomatic moment. The Trump-Xi summit earlier this year produced limited but notable progress on AI safety dialogue, even as the underlying technology competition intensified. US export controls continue to restrict Nvidia’s most advanced chips from reaching Chinese buyers, a policy architecture that has simultaneously constrained Chinese AI development and accelerated domestic chip investment by companies like Huawei and Alibaba.
Yet the Ramp data suggests that hardware decoupling at the supply chain level is not translating into software decoupling at the enterprise adoption level. American businesses are separating the question of where AI chips come from, a matter for trade policy and national security officials, from the question of which AI services deliver the best value for their workflows. That separation may not survive increased regulatory scrutiny. Several US states and federal agencies have already moved to restrict government use of DeepSeek, and it is plausible that legislation extending those restrictions to regulated industries like financial services and legal practice could follow.
For now, however, the market is speaking in the language that corporate spend data makes unavoidable. DeepSeek, whose trajectory from open-source darling to commercial vendor has unfolded faster than most analysts anticipated, has reached a position that would have seemed implausible at the start of 2025: it is the trending software product across a platform tracking more than 50,000 American businesses, outpacing even the most established names in Silicon Valley AI for growth velocity. Whether policymakers choose to intervene in that market dynamic is the next chapter of a story that enterprise procurement data is now forcing into the open.