Gravity
no clean $ figure — heavy open-weight adoption + ultra-cheap API (impact proxy: #1 US iOS app, ~$600B Nvidia 1-day loss Jan-2025). New 2026-07-28 proxy data: Chinese open-weight models (DeepSeek the most-cited example) reportedly command ~30-46% of enterprise LLM token usage and captured share on OpenRouter that shifted from ~30% to ~70% over the past year, at a reported ~1/10th (one source: 57x) the price of US frontier APIs — directionally large third-party economic footprint, but still no total-market-$ baseline sourced well enough to convert this into a defensible $/yr figure without guessing the denominator
Sourced figures 5
- board.yaml deepseek gravity open-weight adoption + ultra-cheap API; #1 US iOS, ~$600B Nvidia 1-day loss Jan-2025
- Google News RSS — DeepSeek enterprise adoption/market-share sweep (attrib. MarketScale/finance.biggo.com) Chinese AI models handle 'nearly a third of enterprise tokens at one-tenth the cost of US rivals'; separate report cites 'token usage hits 46%'
- Google News RSS — DeepSeek enterprise adoption/market-share sweep (attrib. OfficeChai) US models' share of OpenRouter volume 'collapsed from 70% to 30% over the past year' as Chinese models (DeepSeek prominent) gained share
- Google News RSS — DeepSeek enterprise adoption/market-share sweep (attrib. KuCoin/finance.biggo.com) 'DeepSeek Surpasses OpenAI and Anthropic in U.S. Enterprise Adoption Amid AI Cost Crisis'; Uber's LLM budget reportedly 'exhausted in four months' pre-switch to cheaper Chinese alternatives
- Google News RSS — DeepSeek enterprise adoption/market-share sweep (attrib. VentureBeat/finance.biggo.com) DeepSeek V4 cut prices further ('permanent price cut', output pricing cited under $1/million tokens, described as a 75% cut vs prior pricing)
the working
Threads 1
A claim is one metric we assert about a node — the value is the summary, this page is the receipt. Sourced from kestrel; supersedes-history not yet shown. Back to DeepSeek.