News
Nvidia VP Ian Buck says LPX release prioritized over CPX due to resource allocation limits
Wednesday, March 18, 2026 at 04:39 PM
Nvidia VP Ian Buck explained that the decision to prioritize the LPX chip release over CPX this year was driven by internal resource allocation and engineering focus constraints rather than technical failure.
Context
At Nvidia’s GTC 2026 event, Vice President Ian Buck addressed the company’s shift in its product roadmap, stating that the prioritized release of the Groq 3 LPX over the previously announced Rubin CPX was driven by resource allocation. Buck explained that "there's a limit to how many chips Nvidia can—where we want to swarm," indicating a strategic decision to focus engineering talent on high-impact inference solutions. While the Rubin CPX was originally unveiled in September 2025 as a dedicated GPU for massive-context applications like generative video, it has now been placed on hold.
Instead, Nvidia is moving forward with the Groq 3 LPX, an accelerator rack featuring 256 interconnected LPUs licensed from startup Groq. These systems are designed to deliver up to 35x higher inference throughput per megawatt for trillion-parameter models compared to prior architectures. By pivoting to the LPX for its Vera Rubin platform, Nvidia aims to capture immediate demand for low-latency agentic AI while managing a record-breaking fiscal year that saw total revenue reach $215.9 billion.
Sources (5)
Inside NVIDIA Groq 3 LPX: The Low-Latency Inference Accelerator for the NVIDIA Vera Rubin Platform | NVIDIA Technical BlogNVIDIA Unveils Rubin CPX: A New Class of GPU Designed for Massive-Context Inference | NVIDIA Newsroom
NVIDIA Corporation - Financial Reports
Nvidia Puts Groq LPU, Vera CPU And Bluefield-4 DPU Into New Data Center RacksNvidia removes Rubin CPX accelerators from its roadmap — Groq 3 LPUs take center stage as CPX is removed | Tom's Hardware
Related Companies
Nvidia
NVDA