DeepSeek says it has 20,000 H-equivalent units and wants to spend into GPUs fast
teortaxesTex · x · 2026-07-23
DeepSeek said it had about 20,000 H-equivalent compute units by early June, with many machines only arriving recently. The company plans to keep expanding aggressively and buy large batches of hardware over the coming months—“basically all NVIDIA.”
The quote also says:
- the main gap with the U.S. is resources, not personnel stability;
- the team wants to convert as much capital as possible into GPUs;
- if the money can be spent within six months, that would be ideal.
In short, the post highlights how fast DeepSeek is scaling compute and how directly it is channeling funding into GPU purchases.
More from Companies & People
- Neill Blomkamp’s 13-minute AI film NIGHTBORNE was made in under two weeks — minchoi · 2026-07-23
- Liang Wenfeng says China has no cost advantage in data annotation — teortaxesTex · 2026-07-23
- V4 GA is delayed, and 150B-active training is now penciled in for late 2026 to April 2027 — teortaxesTex · 2026-07-23
- A surreal AI startup drama thread about public beef and falling apart — DionysianAgent · 2026-07-23
- OpenAI seeks AI safety researcher for real-world misalignment investigations — tomekkorbak · 2026-07-23
- Crawlravenseo pitches a way to turn old search data into AI-era opportunities — ayushtweetshere · 2026-07-23