Open-Weight AI Models Gain Significant Market Share and Spending
Open-weight models are rapidly capturing token volume and enterprise spending through Vercel's AI Gateway.
1Rapid Adoption of Open-Weight Models
Open-weight models have seen a massive surge in usage, processing over seven in ten tokens via Vercel's AI Gateway by September. Their share of total spending quadrupled in just three months, rising from one in twenty dollars in June to nearly one in four dollars by the end of September.
2Gateway Economics and Price Deflation
While total gateway spending in September was 13 times higher than in December 2025, the average price per token dropped by 47 percent. This significant cost reduction highlights how competitive pressures are reshaping the AI landscape, forcing vendors to lower prices as users increasingly switch between different primary AI providers.
3Market Dynamics and Vendor Switching
Nearly one in three teams changed their primary AI provider during September alone. This high churn rate suggests that market competition is effectively limiting vendor lock-in. While TypeSafe AI’s Jev model gained rapid popularity among 35 percent of paid teams, its low cost resulted in minimal overall token and spending impact.
4Dominance of US-Based AI Models
Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6 Astra remain the primary drivers of AI spending, capturing 11.03 percent and 4.86 percent respectively. Despite the rise of alternative providers, these established American models continue to command the largest portion of enterprise budgets, maintaining their strong position within the current AI ecosystem.
5The Rise of Chinese AI Providers
Chinese AI models are gaining traction, with DeepSeek V4.1 Flash ranking third in total spending at 4.19 percent. Moonshot AI’s Kimi model captured 13.2 percent of spending, while Z.ai’s GLM-5.3 accounted for 4.5 percent. Notably, two-thirds of spending on Chinese models now flows through US providers serving those specific AI technologies.
6Regulatory Risks and Export Controls
China recently proposed adding 40 technology items to its export control list, including 23 related to AI. Models like Moonshot’s Kimi K3 and Alibaba’s Qwen 3.8-Max may fall under these new parameter thresholds. This creates significant uncertainty for US businesses that rely on these models for their core AI infrastructure and operations.
7Strategic Vulnerabilities for US Businesses
Chris McGuire from the Council on Foreign Relations warns that Chinese government control over licensing agreements poses a major risk. Since Chinese labs are increasingly licensing open-source models for revenue, the ability of Beijing to curtail these agreements at any time makes relying on these specific AI models a precarious business strategy.
8Future Regulatory Framework Requirements
Experts suggest the US must adopt a comprehensive regulatory framework to govern AI technology usage overseas. This includes integrating Chinese AI models hosted or used within the United States into domestic regulations. Without such policies, American businesses remain exposed to shifting trade barriers and the unpredictable nature of international AI model licensing.