Twelve Benefits of CyrusOne’s Intelliscale™ Design for Artificial Intelligence (AI)

AI’s large volume of data and heavy compute requirements, coupled with unprecedented energy demand and need for cooling, is a departure from traditional data center design. CyrusOne’s Intelliscale™ Design was engineered specifically to deliver the computational power, scalability, energy efficiency, and specialized infrastructure needed to support the demanding requirements of AI workloads.

Intelliscale Design Features

Ultra-High Density

Intelliscale Design delivers more than 2,000 watts per square foot and an astounding 600-to-1,200 percent more watts per square foot than typical data center power density.  The Intelliscale data center is built upon an ultra-high-density foundation, optimizing space utilization, and presenting the ideal solution for companies aiming to effectively scale their AI infrastructure. Intelliscale data centers can occupy just 25% of the space of typical data centers, depending on the application and needs and can be single or multi-story.

Environmentally Conscious and Efficient

Intelliscale is designed to be LEED-certified supporting the environmental, social and governance (ESG) desires and requirements of both CyrusOne and our customers. The Intelliscale Design helps reduce the carbon footprint, achieve zero water consumption, and because AI data centers do not need to be within the traditional availability zone (AZ) reduces the regional, cultural and infrastructure concerns association with some data center build outs.

CyrusOne developed a unique cooling technique which uses closed loop chillers to reduce heat as well as conserve water.  This is in addition to liquid, air, and hybrid cooling offerings, and immersion to sustain an ultra-cool operating environment more efficiently while prioritizing protection and conservation of natural resources.

Another feature of Intelliscale is its low power usage effectiveness (PUE). Intelliscale data centers are equipped with specialized, high-performance hardware, including GPUs and TPUs optimized for parallel processing and matrix operations needed in AI applications. When combined with ultramodern resource management techniques Intelliscale ensures the most efficient and effective utilization of resources. Combine this with build efficiency when it comes to size, materials, and labor and Intelliscale provides a more environmentally friendly and efficient facility.

CyrusOne’s traditional data centers can also be retrofit given the Intelliscale design incorporates CyrusOne’s standard kit.

The Standard for Flexibility, Redundancy and Security

CyrusOne has decades of experience, capital and partnerships that have been put to work to innovate and deploy state-of-the-art data centers and technologies that are flexible, scale with the needs of customers, deliver the highest levels of reliability and both physical and cyber security.  As AI applications change Intelliscale can scale up and down with the needs of applications delivering a more flexible and cost-effective solution.  Similarly, customers have options for the level of redundancy needed to support their applications including N, N+1, 2N, and N-2c.  Finally, given the nature of the data used and collected for use in AI and machine learning, CyrusOne continues to provide the highest levels of both physical security and cyber security possible.

Intelliscale Design Benefits

Intelliscale Design delivers several benefits that cater to the unique demands of AI workloads. These benefits include:

1. High Performance Computing (HPC) Capabilities: AI computations, especially deep learning, require massive computational power. Intelliscale Design data centers are equipped with high-performance hardware, including GPUs (Graphics Processing Units) and TPUs (Tensor Processing Units), which are optimized for parallel processing and matrix operations essential for AI model training and inference.
2. Accelerated Training: AI models often involve training on large datasets for extended periods. Our AI data centers can harness the power of multiple GPUs or TPUs to significantly speed up the training process, reducing the time required to train complex models.
3. Scalability: AI workloads can vary in intensity, with demands ranging from lightweight inference tasks to heavy-duty model training. Intelliscale is built to be scalable, allowing organizations to easily scale up or down their computational resources based on the current workload demands.
4. Environmental Cooling: With the increased computing power and density of AI data centers, the cooling requirements grow exponentially. Intelliscale leverages CyrusOne’s unique closed loop chillers, as well as air, liquid chip, hybrid cooling, and immersion to cost effectively cool equipment and the physical environment while supporting ESG initiatives with zero water consumption.
5. Energy Efficiency: AI-specific data centers are designed to optimize energy efficiency, which is crucial given the power-hungry nature of AI computations. Intelliscale uses specialized cooling techniques, power-efficient hardware, and advanced thermal management to minimize energy consumption and operational costs.
6. Specialized Hardware: Intelliscale Design Centers are equipped with specialized hardware, such as GPUs, TPUs, and FPGAs (Field-Programmable Gate Arrays), that are tailored for AI tasks. These hardware components are optimized to perform the complex matrix operations and deep learning algorithms that AI workloads require.
7. Optimized Network Infrastructure: AI workloads often involve transferring and processing large amounts of data. Our solutions offers high-speed interconnects and optimized networking infrastructure to ensure fast data transfer between servers, which is crucial for distributed training and real-time inference.
8. Dedicated Storage Solutions: AI models can generate and process massive amounts of data. CyrusOne’s data centers are equipped with high-capacity and high-speed storage solutions to store training data, model checkpoints, and other related assets efficiently.
9. Reduced Latency: AI applications like real-time image recognition, natural language processing, and autonomous vehicles require low-latency processing. Our data centers are designed to minimize network latency and provide rapid responses for time-sensitive tasks.
10. Security and Compliance: AI models often deal with sensitive data, making security a top priority. CyrusOne’s data centers incorporate advanced security measures to protect both the data and the computational infrastructure. They also help organizations follow relevant data protection regulations.
11. Customization and Optimization: AI workloads can be highly specialized. Intelliscale data centers offer customization options to optimize the hardware and software stack for specific AI applications. This ensures that the infrastructure is tailored to the unique requirements of the AI tasks at hand.
12. Resource Isolation: CyrusOne provides resource isolation mechanisms to ensure that different AI workloads running concurrently do not interfere with each other's performance, allowing for consistent and predictable execution.

Intelliscale Design is built for AI - providing the computational power, scalability, energy efficiency, and specialized infrastructure needed to support the demanding requirements of AI workloads, enabling organizations to develop and deploy advanced AI applications that benefit from CyrusOne’s innovative, sustainable AI practices, processes and build-out.