FirstMoveWED 07 OCT 2026
TODAYWORLDMARKETSAIAI SKILLSSPORT
First Move  ·  AI Hardware & Infra  · 
The build-out — Wednesday morning, 07 October

NVIDIA Unveils LPX Rack with Rubin GPUs and Groq 3 Chips at GTC 2026 for Faster AI Data Centers

The AI compute build-out accelerates with new datacenter hardware integrations and international power capacity expansions, signaling a focus on efficiency and scale.

EDITION 2026-10-07 · EVERY CLAIM SOURCED · GROUNDED VIA SEARCH AT PUBLISH
ShareWhatsAppXLinkedIn

The story

NVIDIA unveiled its latest AI data center hardware at GTC 2026, aiming for increased speed and efficiency. Central to the announcement is the LPX rack, which integrates Rubin GPUs and Groq 3 LPU chips.

This all-in-one solution is designed to reduce costs and enhance performance for AI workloads. The LPX rack incorporates technology licensed from Groq in a $20 billion deal, featuring the Groq 3 LPU chip with substantial on-chip memory to minimize delays.

Alongside the LPX, NVIDIA also introduced the Vera-Rubin NVL72 rack and the Spectrum-6 SPX rack, completing a suite of offerings for AI providers. These developments underscore a push towards faster processing and improved energy utilization, enabling companies to extract more value from their AI infrastructure.

Silicon

LPX rack (system)

Maker: NVIDIA

What: Integrates Rubin GPUs and Groq 3 LPU chips for enhanced AI data center performance and efficiency.

For Whom: AI providers and data centers.

AI TOP ATOM

Maker: GIGABYTE

What: Compact personal AI supercomputer delivering 1 PFLOPS of FP4 performance for local AI development.

For Whom: Developers and small to medium businesses (SMBs) for local AI model building and fine-tuning.

W775-V10

Maker: GIGABYTE

What: High-performance desk-side AI workstation powered by NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, offering 20 PFLOPS of FP4 performance.

For Whom: Secure local AI development for regulated data and proprietary models.

Photonic Fabric platform

Maker: Marvell

What: Replaces electrical interconnects with light-based transport, aiming for up to three times performance for the same energy.

For Whom: Future and existing AI systems to improve data movement between GPUs.

The build-out

ProjectWhoScaleWhere
AI Factory campus developmentApplied DigitalUp to 1 GW of potential power capacityFinland, marking the company's first international expansion.

Supply & policy signals

ON.energy's AI UPS (Uninterruptible Power Supply)

Implication: Addresses grid instability by preventing load swings from AI data centers, a critical bottleneck for AI expansion.

Reporting + analyst voices: grounded via Google Search at publish time.