Why the price war matters for makers and artists
For creators and developers, the shift in enterprise spending signals a brutal new reality: the era of free or heavily subsidised credits is over. As US firms pivot toward cheaper Chinese models like Deepseek to slash inference costs, the market is moving toward a strict price-to-performance calculus. Artists and makers using these tools must expect model prices to climb across the board, ending the age where open access was guaranteed by flat-rate subsidies.
Deepseek’s surge and the security caveat
In June 2026, Deepseek claimed the top spot on Ramp’s list of fastest-growing software vendors. This metric tracks breakout expansion relative to a company’s existing size. Ara Kharazian, chief economist at Ramp, clarifies that this growth stems from US businesses paying Deepseek directly and routing data through its platform, rather than from the security benefits of self-hosting open-source code.
Kharazian remains wary of this trend. He highlights significant security and competitive risks inherent in using Chinese models directly and questions whether this surge will endure. Deepseek launched its V4 model at the end of April 2025. While it does not match the total performance of top Western counterparts, its cost is a fraction of the price, creating a scenario where the performance gap is far smaller than the price gap.
The trajectory has been volatile. Deepseek enjoyed a brief hype cycle in January 2025, reaching 0.3 per cent adoption among US companies according to the Ramp AI Index, before dropping quickly to 0.1 per cent. These figures are drawn from real transaction data covering more than 50,000 companies.
While Western AI labs still lead by a wide margin, Deepseek’s breakout in May fits a broader pattern of Chinese models gaining ground due to their superior price-performance ratio. A report from December 2025 indicated that Chinese models, including Deepseek and Alibaba’s Qwen, surpassed US rivals in Hugging Face downloads for the first time. They now account for over 44 per cent of all downloads for popular new models.
The emerging token economy
Kharazian identifies cost consciousness as the primary driver behind Deepseek’s strong showing in June. This sentiment is mirrored in the growth of inference platforms such as Fireworks AI, fal AI, and DeepInfra, which allow companies to run open-source models without paying premium rates to providers like OpenAI or Anthropic.
These developments mark early signs of a token economy. Firms are increasingly likely to select models based on cost efficiency rather than brand prestige. Concurrently, the debate over AI’s return on investment has intensified, with many businesses struggling to quantify the value. As model prices have risen across all providers, the era of heavily subsidised flat rates is clearly nearing its end.
Contrary to some fears, Ramp’s data does not support the notion of an “SaaSpocalypse”—the idea that AI will destroy established software products. Design tools like Figma and Paper remain in high demand, despite the success of competitors like Anthropic’s Claude, which recently launched its own design spinoff.
Key takeaways
- Deepseek dominated Ramp’s trending vendor list in June 2026, driven by US companies seeking direct access to cheaper Chinese models rather than self-hosting benefits.
- While Western labs maintain a performance lead, the significant price gap makes Chinese models increasingly attractive, with downloads surpassing US rivals in late 2025.
- The market is shifting toward a token economy where cost-efficiency dictates model selection, and the age of subsidised flat rates is effectively over.
- Despite fears of AI obliterating existing software, core tools in design and productivity remain in strong demand alongside new generative capabilities.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
