Nvidia's dedicated inference accelerator Groq 3 LPX enters full production to supercharge AI agents
Nvidia Corp. announced full production of its AI inference accelerator, Groq 3 LPX, designed to enhance AI agent performance. The chip, part of the Vera Rubin platform, aims to reduce latency in AI tasks. Nebius Group N.V. is the first customer. Nvidia claims the chip delivers 3,400 tokens per second, four times faster than rivals. Nvidia acquired Groq Inc. for $20 billion in December 2023. SpaceX is also a new customer, using Vera Rubin for its AI architecture.
How this was made

The 30-second read
Why it matters
The launch may expand Nvidia's data‑center TAM and attract enterprise AI workloads seeking lower latency.
Market read
A significant product rollout for a market‑dominant AI chipmaker, likely to influence AI hardware demand.
What to watch
Supply chain constraints and the $20 B acquisition cost could pressure margins in the short term.
Background
Nvidia unveiled the Groq 3 LPX at Hot Chips 2026, positioning it as a dedicated inference accelerator for agentic AI.
Ticker impact
Nvidia announced its Groq 3 LPX inference accelerator entered full production and secured Nebius as the first customer.
Potential upside for NVDA as customers adopt the accelerator; short-term price may rise on the news.
The product targets high‑growth agentic AI workloads and includes a record 3,400 tokens/sec benchmark, indicating a competitive edge.
Market effects
Strengthens the AI hardware segment and may pressure rivals like AMD and Intel.
Positive for US tech equities; limited immediate effect on other regions.
Reinforces Nvidia's role in the global AI compute race.
Counterpoint
Adoption risk if competing architectures prove more cost‑effective; customers may delay deployment.
Key entities
- CompanyNvidia Corp.
US‑listed AI chipmaker.
- CompanyNebius Group N.V.
First customer for the Groq 3 LPX accelerator.


