Groq
LPU inference-only silicon optimized for deterministic low-latency token generation
Groq designs the LPU (Language Processing Unit), a deterministic, SRAM-based inference architecture that delivers very high tokens/sec and low latency, deployed via its GroqCloud service. In 2025 it raised large rounds at a multibillion-dollar valuation and signed major deals including Saudi Arabia's HUMAIN, positioning as an inference-first alternative to GPUs. Its current chips are fabbed on a mature GlobalFoundries node (not leading-edge), sidestepping some HBM/CoWoS constraints.
Product
LPU inference accelerator, GroqCloud
Edge
Deterministic, SRAM-based, no HBM
2025 deals
HUMAIN (Saudi Arabia)
Fab
GlobalFoundries (mature node)
How it fits the stack
Groq with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
Groq in the AI stack. Groq with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 2 entities, 1 relationships
- Groq —competes with→ Nvidia
Context — capital, rivals, policy · · 1