Hark Partners with NVIDIA for Gigawatt-Scale AI Compute

AI startup Hark has forged a multi-year strategic partnership with NVIDIA, securing access to gigawatt-scale computing infrastructure. This collaboration, built on NVIDIA’s next-generation Vera Rubin platform, is set to accelerate Hark’s development of personalized agentic AI.

This significant compute capacity is crucial for both AI training and inference, as Hark prepares for the public launch of its platform. The company aims to bring its initial AI platform to market before the end of summer.

Key Partnership Details

• Partner: NVIDIA
• Compute Platform: Next-generation Vera Rubin
• Infrastructure Scale: Gigawatt-scale computing
• Launch Target: Public platform by the end of summer

Hark focuses on creating multimodal AI systems capable of processing speech, vision, and persistent memory. These systems are designed to deliver deeply personalized user experiences, operating across existing devices and Hark’s custom-designed hardware.

NVIDIA’s AI infrastructure has already supported the creation of Hark Handoff, a computer-use model developed by the startup. The partnership will continue to be vital for distributed inference once Hark’s platform goes live, and for future training of its proprietary foundation models.

Integrated NVIDIA Technologies

• Model Training: Megatron stack
• Efficient Inference: Dynamo
• Pretraining Data: Nemotron
• GPU Cluster Resiliency: NVSentinel

Nico Caprez, NVIDIA’s Vice President of Global AI Infrastructure Growth, stated that personalized AI, especially when combining speech, vision, and and persistent memory, demands substantial computing for both training and inference. Hark founder and CEO Brett Adcock highlighted that the NVIDIA partnership provides the necessary infrastructure to develop highly personalized AI and deploy it rapidly and at scale.

Hark is concurrently advancing its core AI models, native hardware, and new interfaces. Its agentic systems are engineered for natural interaction with humans and the real world.

NVIDIA’s computing power will underpin Hark’s training, inference, and long-term development strategies as it prepares its initial AI platform for market. This alliance positions Hark to scale its ambitious vision for highly personalized AI.