Corporate America Shifts to Open-Weight AI Models as Earnings Call Mentions Surge 6x Year-Over-Year
Summary
Updated Oct 6 — Open-weight AI adoption now intersects three converging policy debates: enterprise strategy, national security, and AI safety.
- • Open model mentions in US earnings calls surged 6x year-over-year, AlphaSense data shows
- • Vercel open model token share jumped from 7% to 56% in just eight months
- • AT&T runs 40% of AI workloads on open models, targeting 70% within 12 months
- • US firms adopting Chinese open models like DeepSeek as cheaper frontier alternatives
Updates
Open-weight AI: three converging debates now colliding
Analysis frames the open-weight AI shift as three distinct but converging conversations: enterprise cost and data sovereignty strategy, national security policy, and AI safety — all headed toward a collision as US open models proliferate and government actors engage more directly.
National-security-first US AI czar now in office
A US AI czar with a national security mandate is now in office, adding formal government oversight to the open-weight AI debate at a time when Chinese models like DeepSeek are being widely adopted by major US enterprises, raising geopolitical tension.
Wave of US open-weight models reportedly imminent
Reports indicate a significant cohort of US-developed open-weight models is forthcoming, which could shift enterprise preference away from Chinese alternatives like DeepSeek and materially reshape the competitive open-model landscape.
Details
6x YoY surge in open model mentions across US earnings calls
Mentions of open-weight or open-source AI models in US corporate earnings calls and investor conferences jumped sixfold in August–September 2026 vs. the same period in 2025, according to AlphaSense data cited by the Financial Times — the clearest quantitative signal yet of a structural shift in enterprise AI preferences.
Vercel: open model tokens grew from 7% to 56% of total in 8 months
Vercel's share of tokens processed by open-weight models climbed from just 7% in December 2025 to 56% by August 2026 — an eightfold increase in under a year, driven primarily by cost savings as open models reached performance parity with frontier APIs for common developer workloads.
AT&T: 40% of AI workloads on open models today, targeting 70% within 12 months
AT&T, which processes 45 billion tokens daily, currently runs 40% of its AI workloads on open-weight models and has set a goal to reach 70% within 12 months. The scale of AT&T's deployment makes it one of the largest enterprise open-model adopters disclosed publicly.
Tinder's 10x AI spend surge drove shift of non-technical queries to open models
Tinder's annualized AI spend surged from $1 million in January 2026 to $10 million by July 2026. The tenfold cost increase prompted the company to shift non-technical queries to cheaper open-weight models while retaining frontier models for higher-complexity tasks.
Chinese models including DeepSeek adopted as cost-saving frontier AI alternatives
US companies beyond Silicon Valley are adopting Chinese open-weight models, including DeepSeek, as cost-effective alternatives to proprietary frontier models from OpenAI and Anthropic. The trend introduces a geopolitical dimension given US export-control discussions around advanced AI.
PNC Financial, CH Robinson, Siemens, Digital Realty among mainstream enterprises shifting to open models
Financial services firm PNC Financial Services, logistics company CH Robinson, industrial conglomerate Siemens, and real estate data firm Digital Realty are among the traditional-industry enterprises cited as shifting workloads to open-weight models, confirming the trend is no longer confined to tech firms.
Additional details sourced via Grok live web research from secondary outlets
Details compiled via Grok live web research from PYMNTS (September 28, 2026), Financial News Korea (September 28, 2026), and Techmeme (September 27, 2026) — all secondary reporting on the original Financial Times story. Core figures are consistent across all sources.
Sources: Financial Times (September 27, 2026); corroborated by PYMNTS and Financial News Korea (September 28, 2026) via Grok Research.
What This Means
The sixfold surge in earnings-call mentions of open-weight models signals a structural shift in enterprise AI spending, with cost pressure driving mainstream corporations away from premium frontier APIs toward cheaper open-weight alternatives. The scale of adoption — AT&T running 45 billion tokens daily with plans to reach 70% open-model coverage — shows this has moved from experimentation to core operations across traditional industries. Chinese open-weight models like DeepSeek being preferred by major US companies adds a geopolitical dimension that US export-control policy has not fully resolved. For OpenAI and Anthropic, the trend puts sustained pressure on API pricing and challenges the narrative that proprietary closed models are uniquely capable for enterprise use cases.
Sentiment
Broadly acknowledging the shift to cheaper open models with cost pressure on closed APIs, some skepticism on data sources
“AT&T is doing exactly what OpenAI and Anthropic fear about enterprise. they’re routing 40% of employee AI usage to open models, and intend to boost it to 60-70%. > coding costs down 56% > quality dropped just 2% > 45B tokens/day btw.”
“I keep reading that local models are largely irrelevant in business. That is completely wrong. Rising AI bills are pushing US companies toward cheaper open-weight models, including Chinese alternatives to OpenAI and Anthropic, the FT reports. For example AT&T wants to increase open models’ share of its AI workloads from about 40% to 70% within a year.”
““.. By May, open models accounted for 20 percent of AT&T’s A.I. use. That has since risen to 40 percent and may jump to 60 percent in the coming months ..” @nytimes”
“FT's entire story on American companies doing more work with open weight Chinese models is based on......wait for it......earnings call mentions. That is not data on usage presented through these calls but just mentions. There is a quote later on in the copy from AT&T but strangely enough, the FT reporters don't seem to have asked AT&T the identity of the Chinese models.”
Questioning the strength of evidence in the FT coverage
Split
~70/30 acceptance of the trend vs questioning the underlying data quality or geopolitical angle
Sources
- FT: Corporate America rejects overpriced frontier, embraces open modelsArchive
- Mentions of open models in latest US earnings calls surged 6x YoY, with open models hitting 56% of Vercel tokens in August and 40% of AT&T's AI workloadsFt
- Why Companies Want AI They Can OwnAI Daily Brief