Niv-AI Unveils Tech to Tame AI's Power Hungry GPUs, Secures $12M Seed Funding

Key Takeaways

• Niv-AI exits stealth mode with $12 million in seed funding to address GPU power inefficiencies in AI data centers.
• The company's technology focuses on millisecond-level GPU power monitoring and management to optimize energy usage.
• Niv-AI aims to create an 'intelligence layer' between data centers and the power grid, enabling better resource allocation and preventing over-consumption.
• The startup is deploying rack-level sensors in partnership with design partners to gather detailed power usage data.
• Niv-AI plans to develop an AI model to predict and synchronize power loads across data centers.

The explosive growth of artificial intelligence is creating unprecedented demands on data center power infrastructure. The sheer volume of GPUs required to train and run advanced AI models results in significant energy consumption, often leading to inefficiencies and limitations on processing capabilities. Data centers are struggling to keep up with these surges, leading to throttling of GPU performance by as much as 30% to avoid overloading the electrical grid.

Niv-AI, a Tel Aviv-based startup founded by Tomer Timor and Edward Kizis, is tackling this challenge head-on. The company's core innovation lies in its ability to precisely measure GPU power consumption at the millisecond level. This granular level of detail allows for a much deeper understanding of the power profiles associated with different deep learning tasks.

By deploying rack-level sensors in its own facilities and in collaboration with design partners, Niv-AI is gathering the data needed to build a comprehensive model of GPU power usage. This data will then be used to develop mitigation techniques that enable data centers to unlock more of their existing capacity without exceeding power limits.

The ultimate goal is to create an AI-powered "copilot" for data center engineers. This system will be able to predict and synchronize power loads across the data center, optimizing energy consumption and preventing sudden surges that can strain the grid. The company intends to have the system running in select US data centers within the next eight months.

Niv-AI's investors include Glilot Capital, Grove Ventures, Arc VC, Encoded VC, Leap Forward, and Aurora Capital Partners. Their solution not only addresses the immediate problem of power inefficiencies in AI data centers but also lays the groundwork for a more sustainable and scalable future for AI development.

The company's vision extends beyond individual data centers. Niv-AI aims to create an "intelligence layer" between data centers and the electrical grid, facilitating better communication and coordination between the two. This will help to ensure that data centers can access the power they need while also minimizing their impact on the grid's stability.

Why it matters

Niv-AI's technology has the potential to significantly reduce the energy footprint of AI, making it more sustainable and accessible. By optimizing GPU power usage, the company can help data centers to unlock more of their existing capacity, reducing the need for costly infrastructure upgrades and promoting more efficient resource allocation. This is crucial as AI continues to permeate various aspects of modern life, making efficient power management a critical component of its continued growth.