Reddit may cut Google off from training data in a deal worth about $60 million a year
Rangesh06 · reddit · 2026-07-22
Reddit may be considering cutting off Google’s access to its content for AI training, despite an estimated $60 million a year deal signed in 2024.
The post argues the move would reflect Reddit’s growing awareness that its corpus is highly valuable to LLMs, noting a claim that Reddit is currently one of the most-cited sources in LLM training data, ahead of Wikipedia, YouTube, and Google itself.
Related event: Reddit Considers Blocking Google AI Access Amid Traffic Decline(5 posts)→
More from Infra
- 12 KV Cache Reduction Techniques Every AI Engineer Should Understand, Explained — blaizedsouza · 2026-09-11
- The shadow GPU capacity market is formalizing, with Meta selling excess compute to outside buyers — DavidLinthicum · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11