News / US / cnbc
Nvidia Accelerates AI Infrastructure With Commercial Launch of Groq 3 LPX Racks
Electronic Technology · Semiconductors · cnbc · 2026-08-24
NVDA, NBIS, NYYY, AMD, SOXX, CBRS, SOXS, TSM
Nvidia is moving to production with its Groq 3 LPX racks, a key component of its $20 billion acquisition aimed at boosting low-latency AI inference capabilities.
What Happened
Commercialization of Groq Technology: Nvidia has officially moved its Groq 3 LPX rack into full production, marking a major milestone following its $20 billion acquisition of Groq assets. These systems are slated for deployment at Nebius later this year, signaling a rapid integration of the startup's specialized architecture into Nvidia's broader data center ecosystem.
Focus on Low-Latency Inference: The new racks are designed to address the growing demand for low-latency inference, which is critical for responsive AI agents and coding applications. By utilizing 500 megabytes of on-die SRAM, these chips effectively bypass traditional memory bottlenecks, allowing cloud providers to offer premium, high-speed service tiers.
Competitive Market Positioning: Nvidia faces stiff competition in the inference space from players like Advanced Micro Devices and Cerebras, which are also targeting low-latency workloads. Despite this, Nvidia maintains that Groq technology is intended to complement, not replace, its core GPU offerings, ensuring the right processor is used for specific parts of the AI workload.
Future Revenue and Scaling: CEO Jensen Huang continues to push an aggressive strategy, projecting $1 trillion in cumulative sales for the Blackwell and Vera Rubin systems through 2027. The company is currently allocating significant data center capacity to these new architectures as it prepares to report its latest financial results.