US Corporates Pivot to Open-Weight AI to Curb Costs

Key points

• Executive mentions of open-weight models in earnings calls surged sixfold in August and September 2026 compared to the same period in 2025, according to AlphaSense.
• AT&T now runs 40% of its AI workloads on open models, aiming for 70% within a year, citing significant cost savings at scale.
• Tinder’s CTO noted AI spending rose from $1 million to $10 million per year between January and July 2026, prompting a shift to open models for non-technical queries.
• Open-weight models accounted for 56% of tokens processed through Vercel’s AI Gateway in August, up from 7% in December 2025.
• Chinese companies dominate the open-weight market, though US and European firms like Nvidia, Mistral, and Reflection AI are also developing competitive open models.

Background

This shift follows a narrowing performance gap between open and closed models, with earlier reports indicating the gap has shrunk to about four months. Chinese open-source models like Qwen and DeepSeek have seen massive global adoption, surpassing US rivals in downloads. Geopolitical tensions and US export controls have accelerated the development of self-sustaining open AI ecosystems, particularly in China, while BRICS nations have proposed collaborative open-source AI zones.

How outlets are covering it

The Financial Times emphasizes the broad corporate adoption of open-weight models across industries, highlighting cost savings and performance parity for routine tasks. PYMNTS frames the shift as a financial decision by CFOs, noting that the trade-off between cost savings and infrastructure responsibility is becoming critical as AI integrates into enterprise operations. Business Insider presents a more polarized view, with Washington University’s CIO Scott Wilson arguing that frontier labs like OpenAI and Anthropic are overvalued due to rising competition from cheaper Chinese models, while venture capitalist Vinod Khosla counters that closed models offer lower long-term costs through integrated infrastructure. 36Kr highlights a broader industry trend of price cuts and efficiency improvements, noting that OpenAI and Anthropic are slowing frontier model development for safety reasons while optimizing existing models for cost and performance.

Why it matters

The adoption of open-weight AI models threatens the revenue growth and valuations of major AI labs like OpenAI and Anthropic, which are preparing for significant funding rounds and IPOs. This shift could reshape the AI supply chain, reduce dependence on proprietary systems, and accelerate the integration of AI into enterprise workflows, potentially altering the competitive landscape and global AI governance.

What to watch

Expect continued pressure on frontier AI labs to reduce prices or improve efficiency as open-weight models gain market share. Companies will likely continue to optimize their AI infrastructure for cost and security, while the debate over the long-term viability of closed versus open models will intensify among investors and industry leaders.