NVIDIA Announces Groq 3 LPX Production Milestone for Agentic AI Inference

Event Summary

NVIDIA described Groq 3 LPX as in full production at Hot Chips on August 24, 2026. Nebius announced that it would be the first AI cloud to adopt the platform, initially for a subset of models on Token Factory. The announcements establish a production and integration milestone, not proof of broad cloud availability or uniform throughput in customer workloads.

Context & Narrative

NVIDIA and Nebius frame the LPX design around low-latency token generation for multi-step agents. NVIDIA cited a 3,400-output-token-per-second result on a specific Gemma 4 31B setup, while Nebius cited the same benchmark and described a planned integration path. NVIDIA's release also states that many products and features remain subject to availability, so EpochArc treats performance and rollout effects as vendor-reported and prospective.

Key Findings

• Fact Grade B

  NVIDIA announced Groq 3 LPX as in full production on August 24, 2026, and Nebius announced it as the first adopting AI cloud.

  Sources [1][2][3]
• Impact Grade B

  NVIDIA reported up to 3,400 output tokens per second on a specified Gemma 4 31B benchmark configuration, aimed at latency-sensitive agent workflows.

  Sources [1][2]
• Limitation Grade B

  Throughput depends on the tested model and rack configuration, while cloud rollout and features remain subject to availability and partner delivery.

  Sources [1][3]

Impact Assessment

• Capability Leap +1 · Medium-term

  Creates a production milestone for specialized low-latency inference hardware intended for multi-step agent workloads.

  Affected Groups: AI cloud operators, agentic system developers, enterprise users

• Economic Disruption +1 · Medium-term

  Nebius's announced integration indicates early commercial interest in specialized rack-scale inference, subject to delivery and adoption.

  Affected Groups: semiconductor industry, cloud providers

Consensus & Sources

• 1
  NVIDIA Groq 3 LPX Now in Full Production With World-Class Speed for Agentic AI

  URL: https://nvidianews.nvidia.com/news/nvidia-groq-3-lpx-now-in-full-production-with-world-class-speed-for-agentic-ai

  Reference Evidence Citation logged Live source
• 2
  NVIDIA Groq 3 LPX coming to Nebius Token Factory

  URL: https://nebius.com/blog/posts/nvidia-groq-3-lpx-nebius-token-factory

  Reference Evidence Citation logged Live source
• 3
  Nvidia says Groq racks will be online this year following $20 billion purchase

  URL: https://www.cnbc.com/2026/08/24/nvidia-says-groq-racks-will-be-online-this-year-after-20-billion-deal.html

  News Report Citation logged Live source