As we dive into the buzz and brilliance of generative AI in 2024, it’s impossible not to get swept up in the excitement. Icons like ChatGPT are leading the charge, making it clear why virtually every online business is eager to harness AI.

But let’s take a moment to spotlight the unsung heroes behind this digital transformation: the mighty data centers. These are the powerhouses making AI magic possible, equipped to handle hefty power and cooling demands. As we explore the transformative impact of AI on business operations, let’s tip our hats to the facilities that keep our AI dreams alive and kicking.

Handling massive power and cooling requirements

AI workloads run on GPU servers, which require considerable power and generate considerable heat. Deft has been supporting high-capacity, power-dense workloads for decades, long before the AI hype.

AI-enabled data center setup in Deft's data center
This is an example of a customer that has 30kW cabinets, all air-cooled, and within NVIDIA specifications for airflow and system temperatures. We installed special vent tiles for these GPU-heavy workloads. The tiles allow for extra cooled air to circulate the cabinets. We also didn’t pack the racks with servers to maintain proper standards of airflow and power usage.

Many data centers need to undergo painstaking upgrades before they can safely handle the power and cooling requirements of AI workloads. Fortunately, Deft already has the right power and cooling systems to support AI today. Looking ahead, we’re also testing newer technologies like in-rack liquid cooling.

How Deft’s data centers support AI today

  • High-density GPUs: AI brings denser racks into multi-tenant data centers. This poses challenges for data center design and management. Deft has power-dense rows of GPUs with proper cooling and low latency networking which meet the stringent certification requirements from NVIDIA.
  • Contiguous deployments: Deft can deploy GPU clusters in contiguous rows instead of playing Tetris with your GPU server deployments. For tasks that require multiple GPUs to work together (like distributed training or machine learning), having contiguous rows reduces the distance your data needs to travel. The result = low latency and better performance.
  • Cost-effective cooling solutions: A key architectural challenge is positioning data-crunching AI racks in specific areas of the data center that can be properly cooled, but more importantly, have proper heat dissipation to minimize any impact on other customers’ cooling requirements.
  • Robust data storage and management solutions: AI workloads need access to large amounts of data. This requires robust data storage and management solutions that can handle enormous surges. Deft’s 100/400-Gbps network is prepared to handle high volumes of data with 100% uptime SLAs.
  • Security and compliance: Your AI setup must comply with various laws and regulations. We pass strict security audits every year as part of our SOC 2 compliance, in addition to supporting data encryption at rest and in transit.
  • Maintenance and monitoring: Once your AI infrastructure is in place, Deft monitors performance on your behalf. We have 24/7 support to help you customize your AI infrastructure, select and set up hardware, and test everything to make sure it works as expected. Then, we regularly apply patching, check system health, and fine-tune performance. This allows you to focus on what’s important to your business — model training and tuning — and not worry about the day-to-day care and feeding of your hardware, storage, and connectivity.

Unlike other data centers that say they can support AI/ML workloads in their data centers, Deft can show you that we already do today. Book a private data center tour or schedule a time to discuss how we can support your AI/ML workloads.

Deft, a Summit company

Deft, a Summit company
2200 Busse Rd.
Elk Grove Village, IL 60007
+1 (312) 829-1111