- Vendor Ranking: DeepSeek leads Ramp’s June trending software vendors list as US firms test cheaper AI access.
- Spending Signal: Ramp tracks first-time vendor purchases, so the list indicates momentum rather than total market share.
- Data Risk: Ramp-linked evidence suggests some firms are paying DeepSeek directly and sending data through its hosted service.
- Market Context: April adoption data kept Anthropic and OpenAI far ahead, while DeepSeek’s lower prices sharpen procurement tradeoffs.
Expense-management company Ramp has released its June 2026 trending software vendors list with Chinese AI startup DeepSeek ranked first, putting a cheaper AI provider at the top of a new-spending signal from US businesses. The list does not show overall enterprise AI market share, but it does show fresh buyer interest as companies weigh model costs against data-control concerns.
Ramp tracks vendors that customers buy from for the first time each month, so the list measures breakout spending rather than total software use. Ara Kharazian, lead economist at Ramp Economics Lab, wrote that he “I didn’t expect American firms to use DeepSeek” before pointing to direct hosted use that raises data-residency and security questions.
Some US firms are paying DeepSeek directly, not only running open-source model code on their own infrastructure. Direct payments put prompts and outputs in DeepSeek’s hosted service, so security teams have to judge both cost savings and data-control exposure before treating the ranking as a simple adoption win.
Ramp did not provide market-share percentages for DeepSeek, OpenAI, or Anthropic. Without those figures, the June result points to new vendor trials rather than proof that DeepSeek has overtaken larger US AI providers in business adoption.
What Ramp’s June Data Shows
Companies buying DeepSeek access directly create a different risk profile from teams that self-host models. Models run on a company’s own infrastructure can keep prompts and outputs inside that environment; hosted use sends and receives data through DeepSeek’s service.
Kharazian made that distinction explicit in Ramp’s June analysis.
“To be clear: this is not just self-hosted open source usage. Firms are sending and receiving data through DeepSeek directly.”
Ara Kharazian, lead economist at Ramp Economics Lab (via Ramp Economics Lab)
Security teams now have two linked questions to answer: whether lower model costs justify a new vendor, and whether hosted use moves sensitive corporate data outside internal controls. Kharazian’s comment narrows the issue because the spending signal points to service access, not only experimentation with public model weights.
Ramp’s evidence base is large enough to make the June result worth tracking, even though it is not a full map of the AI market. Its spending platform represents roughly 70,000 US businesses, up from about 50,000 at the start of 2026, and processes business expenses through corporate card and bill-pay products.
First-time vendor spending can act as an early procurement signal. It still leaves out pilots, bundled contracts, direct cloud credits, and AI use that does not appear as a Ramp-paid line item.
Other lower-cost infrastructure names also surfaced in the same June list. Model-serving provider Fireworks AI, inference platform fal AI, model infrastructure provider DeepInfra, and GPU cloud provider Vast.ai give the ranking a broader alternative-infrastructure flavor without proving that any of them has displaced larger AI providers.
Momentum Is Not Market Share
Ramp’s own history keeps the June result in perspective. In January 2025, DeepSeek briefly reached 0.3 percent adoption among US companies before falling back to 0.1 percent, far from the leading US providers in Ramp’s adoption data.
April data kept that gap alive. DeepSeek’s business adoption remained tiny while Anthropic, the Claude developer, and OpenAI, the leading US AI provider, held 34.4 percent and 32.3 percent of Ramp’s spending-based AI adoption index.
Earlier Ramp stats from April makes the June ranking look like momentum around new vendor trials, not enterprise AI dominance. Paid AI use can be a budget signal without capturing the entire global market.
DeepSeek can be gaining fresh buyer attention while still operating from a small base. June’s vendor list identifies companies breaking into new buyer flows; it does not show whether DeepSeek’s overall business footprint has moved beyond that earlier low adoption level.
Cheaper Models Raise a Data-Risk Tradeoff
DeepSeek made a 75 percent V4-Pro discount permanent in May 2026, and the recent V4-Pro price cut gives procurement teams a concrete comparison against premium US models. DeepSeek V4 arrived at the end of April, which helps explain renewed interest in the model line.
DeepSeek V4 pairs Pro and Flash variants under MIT-licensed open weights, a one-million-token context window, and a lower-cost API pitch. Usage fees charged as AI systems process text can multiply quickly in agent-style workflows, which makes pricing a practical procurement decision.
Kharazian’s March 2025 analysis identified uptime, privacy, local-deployment friction, and falling US provider prices as previous barriers to DeepSeek adoption. June spending behavior suggests some buyers are testing the hosted service anyway as usage-based AI fees become easier to spot in operating budgets.
Lower token prices can push experimentation into normal procurement workflows, while hosted access keeps the data-control question attached to every cost-saving argument. July’s next Ramp update will show whether the first-time buyer flow persists or stays separate from broader business adoption.
