# Groq

> Markdown twin of https://roboterminal.app/companies/groq · https://roboterminal.app/companies/groq.md

## On shared sources

- [NVIDIA](https://roboterminal.app/companies/nvidia) (4)
- [OpenAI](https://roboterminal.app/companies/openai) (3)
- [Anthropic](https://roboterminal.app/companies/anthropic) (2)
- [Cerebras](https://roboterminal.app/companies/cerebras) (2)
- [SambaNova](https://roboterminal.app/companies/sambanova) (2)
- [Alibaba](https://roboterminal.app/companies/alibaba) (1)
- [AWS](https://roboterminal.app/companies/aws) (1)
- [Baseten](https://roboterminal.app/companies/baseten) (1)
- [Braintrust](https://roboterminal.app/companies/braintrust) (1)
- [DeepSeek](https://roboterminal.app/companies/deepseek) (1)
- [Etched](https://roboterminal.app/companies/etched) (1)
- [fal](https://roboterminal.app/companies/fal) (1)
- [Google](https://roboterminal.app/companies/google) (1)
- [J.P. Morgan](https://roboterminal.app/companies/j.p.-morgan) (1)
- [Meta](https://roboterminal.app/companies/meta) (1)
- [MiniMax](https://roboterminal.app/companies/minimax) (1)
- [RunAnywhere](https://roboterminal.app/companies/runanywhere) (1)
- [Runway](https://roboterminal.app/companies/runway) (1)
- [Tesla](https://roboterminal.app/companies/tesla) (1)
- [Together AI](https://roboterminal.app/companies/together-ai) (1)
- [TypeSafe](https://roboterminal.app/companies/typesafe) (1)

## Pulse

- [@stretchcloud on X](https://x.com/stretchcloud/status/2112891658748535040)
  Twenty five trillion tokens. That's what Google says it processed in the first 72 hours after Gemini 4 Argon went live, per Logan Kilpatrick, and the number only makes sense next to the baseline under…
- [@pequityresearch on X](https://x.com/pequityresearch/status/2109434157839327738)
  Expert Takeaways: Inference > GPU Rental Rates Up ~20%: Driven by HBM supply constraints. Neo-cloud providers are passing hardware cost increases straight to tenants via shorter 1–3 year contracts rat…
- [@stretchcloud on X](https://x.com/stretchcloud/status/2102200707541540958)
  The pattern I keep seeing this month: once a model layer commoditizes, the fight moves one layer down to whoever serves it fastest. RunAnywhere, a YC W26 company, launched Wally this week, an inferenc…
- [@ricci_nov on X](https://x.com/ricci_nov/status/2100747855128674789)
  Nvidia's Groq 3 LPU carries no DRAM: 512MB of on-die SRAM at 150 TB/s, against 22 TB/s from the 288GB of HBM4 on each Rubin GPU. It is a decode-only co-processor, with Rubin doing prefill. Bandwidth a…
- [@stretchcloud on X](https://x.com/stretchcloud/status/2100443378186166632)
  The routing problem in a multi-harness agentic stack has been an LLM call problem. You route to Claude Code, Codex, Opencode, or another harness by putting logic in a system prompt and burning tokens …
- [@SemiAnalysis_ on X](https://x.com/SemiAnalysis_/status/2098046327154258390)
  BREAKING NEWS🚨: NVIDIA IS BEING INVESTIGATED BY THE DOJ FOR ANTITRUST CONCERNS OVER THE GROQ LPU DEAL Last year, NVIDIA signed a "non-exclusive license" and hired most of Groq's important people to f…
- [@ricci_nov on X](https://x.com/ricci_nov/status/2096408874828050903)
  SambaNova on Llama 3.1 70B: 16 SN40L chips, versus its own estimate of 336 Cerebras dies for 445 TPS and 576 Groq LPUs for 250 TPS. Each socket holds 520 MiB SRAM, 64 GiB HBM and 1.5 TiB DDR. Vendor c…
- [@stretchcloud on X](https://x.com/stretchcloud/status/2095304264994177524)
  The pattern in AI inference is consistent: the fast-and-cheap version of a frontier model ships 2-4 weeks after the frontier version. H3 Max Turbo confirms this for video generation.H3 Max launched on…
- [The Pulse™ | Week 35: The ceiling is not the plan](https://thefogsignal.substack.com/p/the-pulse-week-35-the-ceiling-is-not-the-plan)
  Nevada authorized 7,000 robotaxis, and Tesla expects 2,500. NVIDIA guaranteed $105B behind OpenAI's Ohio lease and invested $1.5B. Groq raised at half.

---

Source: Roboterminal · https://roboterminal.app/companies/groq.md