News summary produced by Claude AI
Artificial intelligence has emerged as a significant and rapidly growing operational expense for U.S. companies, prompting many to explore cost-reduction strategies. Lindy.ai, a San Francisco-based startup developing AI-powered email and calendar management tools, recently announced it migrated entirely to the Chinese model DeepSeek-V4 after determining that Anthropic’s services represented its largest single expense—exceeding payroll costs for its workforce. The company reported achieving a tenfold reduction in expenses through this transition.
Chinese AI models, while estimated to lag American counterparts by six to 12 months in capabilities, have established a competitive advantage in the open-source sector. These models are available for free download and modification, creating what industry observers describe as a dominant position in open-source development. Adoption is widespread across the sector, with major companies including Airbnb and Perplexity utilizing Chinese models such as Alibaba’s Qwen. Data from OpenRouter indicates DeepSeek usage increased from approximately 9 percent to nearly 20 percent since January, while uptake of models from MiniMax, Xiaomi, and Tencent has also risen.
Companies accessing these models do so primarily through paid hosting platforms based in the United States, such as Featherless and OpenRouter, which maintain user data domestically despite utilizing foreign AI systems. The economic calculation centers on cost per token—the unit of AI computational work. Industry representatives note that many routine, high-volume tasks perform adequately with lower-cost alternatives, allowing companies to reserve premium American models for specialized applications requiring advanced reasoning capabilities.
Not all companies have made the transition. Some startups prioritize model quality over cost savings, particularly early-stage ventures requiring rapid development. Additionally, current subsidized pricing from American firms like Anthropic and OpenAI may not persist indefinitely. Economists tracking corporate spending patterns suggest American AI companies may respond competitively by introducing higher-quality open-source offerings or adjusting pricing structures, though uncertainty remains regarding whether such measures will prevent further migration toward Chinese alternatives as profitability pressures mount on U.S. firms approaching public offerings.