Groq — research brief

# Groq — research brief

**Slug:** groq  
**Date:** 2026-09-21  
**Industry:** AI Inference Hardware / Neocloud  
**Accent:** `#F55036`

## Teaching arc (verified)

**Free GroqCloud Console ($0)** → **Developer pay-per-token API** (higher limits, Flex, Batch, spend limits) → **Enterprise** (custom models, regional endpoints, LoRA, SSO/SCIM, dedicated support) → **GroqMetal / GroqRack / dedicated capacity** + infra partners (e.g. Bell Canada sovereign network; NVIDIA LPX licensing path).

## FACT sources (primary)

| Claim | Source |
|---|---|
| Plans: Free $0 · Developer pay-per-token · Enterprise contact | [groq.com/groqcloud](https://groq.com/groqcloud), [console billing plans](https://console.groq.com/settings/billing/plans) |
| Free: build/test, community support; Developer: higher limits, chat support, Flex, Batch, spend limits, prompt caching / audit logs | Same + [Billing FAQs](https://console.groq.com/docs/billing-faqs) |
| Enterprise: custom models, regional endpoints, performance tier, scalable capacity, dedicated support, LoRA (enterprise-only) | groqcloud + [LoRA docs](https://console.groq.com/docs/lora) |
| GPT OSS 20B: $0.075 in / $0.30 out per 1M · ~1000 t/s | [Models docs](https://console.groq.com/docs/models) |
| GPT OSS 120B: $0.15 in / $0.60 out per 1M · ~500 t/s | Models docs |
| Whisper Large V3 Turbo $0.04/hr · Whisper Large V3 $0.111/hr | Models docs |
| OpenAI-compatible API `https://api.groq.com/openai/v1` | GroqDocs |
| GroqRack on-prem / air-gapped by request; public, private, or co-cloud | groq.com/groqcloud |
| Platform stack: GroqMetal (bare metal) · GroqCore (inference) · GroqAssured (governance) | [groq.com/platform](https://groq.com/platform) / groqcloud |
| 13 data centers; >5M developers; trillions of tokens/week; $650M growth capital Jun 2026 | [Company newsroom Jun 22, 2026](https://groq.com/newsroom/groq-raises-usd650m-to-scale-its-ai-inference-cloud-business) |
| Scale toward 200 MW by end of 2027 (company claim) | Same newsroom |
| Bell Canada exclusive inference for Bell AI Fabric (May 2025 press) | [Newsroom](https://groq.com/newsroom/groq-becomes-exclusive-inference-provider-for-bell-canadas-sovereign-ai-network) |
| SOC 2, GDPR, HIPAA; optional private tenancy | groq.com/groqcloud |

## SECONDARY / press

| Claim | Source |
|---|---|
| Aug 2024: $640M round at $2.8B valuation (BlackRock-led); >360k developers then | TechCrunch / PR Newswire Aug 2024 |
| Does not train foundation models — hosts open-weight models (Llama, Qwen, GPT-OSS, Whisper, etc.) | Product positioning + third-party explainers; align to models page |

## INFERENCE (labeled on canvas)

- Free tier is demand-gen COGS that converts into Developer token spend when rate limits bind.
- Enterprise + GroqRack/GroqMetal attach when latency SLAs, data residency, or LoRA adapters matter.
- Partner/sovereign deals (Bell) and marketplace-style distribution sit above self-serve API ARPU.

## Do not invent

No unaudited ARR, headcount, or token-volume $ figures beyond company claims cited above.