Torus Unveils Foundry: Modular AI Infrastructure Solution Secures $500 Million to Address Inference Power Demands
Torus, a US-based manufacturer, has introduced "Foundry," a modular AI infrastructure solution designed to rapidly deploy AI inference compute capacity. The company simultaneously announced securing $500 million in growth capital, led by Magnetar, to fuel this initiative. Foundry packages essential components like power, storage, cooling, networking, and compute into prefabricated systems capable of delivering 1 to 20 megawatts of new capacity to existing sites. This innovative approach promises compute availability in a mere 12 to 16 weeks from contract signing, a significant acceleration compared to the years typically required for traditional hyperscale data center construction.This development addresses a critical bottleneck in the accelerating AI landscape: the power and infrastructure required for AI inference. As the industry shifts from primarily training large AI models to widely deploying them for real-world applications, the demand for inference compute is projected to grow by 35% annually through 2030, outpacing training demand. Torus's solution aims to unlock unused grid capacity by leveraging existing interconnections between substations and commercial buildings, effectively bringing chips to available power rather than waiting for new power infrastructure.
The introduction of Foundry marks a pivotal step towards democratizing access to high-performance AI compute. By drastically reducing deployment times and optimizing power utilization, Torus could enable faster adoption of AI across various industries, particularly for enterprise applications reliant on inference. This modular approach suggests a future where AI infrastructure is more agile, scalable, and responsive to immediate demand, potentially reshaping how companies acquire and utilize AI capabilities in an increasingly AI-driven world.
Torus Unveils Foundry: Modular AI Infrastructure Solution Secures $500 Million to Address Inference Power Demands
Torus, a US-based manufacturer, has introduced "Foundry," a modular AI infrastructure solution designed to rapidly deploy AI inference compute capacity. The company simultaneously announced securing $500 million in growth capital, led by Magnetar, to fuel this initiative. Foundry packages essential components like power, storage, cooling, networking, and compute into prefabricated systems capable of delivering 1 to 20 megawatts of new capacity to existing sites. This innovative approach promises compute availability in a mere 12 to 16 weeks from contract signing, a significant acceleration compared to the years typically required for traditional hyperscale data center construction.
This development addresses a critical bottleneck in the accelerating AI landscape: the power and infrastructure required for AI inference. As the industry shifts from primarily training large AI models to widely deploying them for real-world applications, the demand for inference compute is projected to grow by 35% annually through 2030, outpacing training demand. Torus's solution aims to unlock unused grid capacity by leveraging existing interconnections between substations and commercial buildings, effectively bringing chips to available power rather than waiting for new power infrastructure.
The introduction of Foundry marks a pivotal step towards democratizing access to high-performance AI compute. By drastically reducing deployment times and optimizing power utilization, Torus could enable faster adoption of AI across various industries, particularly for enterprise applications reliant on inference. This modular approach suggests a future where AI infrastructure is more agile, scalable, and responsive to immediate demand, potentially reshaping how companies acquire and utilize AI capabilities in an increasingly AI-driven world.
Sources
datacentremagazine.com https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFVRmXfENSOIDolUqa2Tk5UNvDhX68LWNoHxnrxgfynmGkS-6-t_fqx0XSJi3vyoHKlJvZP94YEpK-pSYCijBs2qJrA_Zfc5DrMmcMqeiuezM6kJP8thXORpb4cjgWzYeam7S962xbyAxx6ygeBbSiWutRfTt0IGcA5au49d0m3c3nxbgsCHEdc9LFzsYX9xeVwh0xE6fI-
#AIInfrastructure #ModularComputing #AIChips #EnergyEfficiency #DigitalTransformation
