The Uptime Institute’s 2024 Global Data Center Survey made it clear that AI is the megatrend impacting all data centers. But the fulfillment of AI demand will require data centers to operate hybrid architectures while finding a way to achieve sustainability.

Data Centers Understand the Potential Benefits of AI

Most operators recognize the benefits of AI but foresee immediate challenges in adding it to their data center repertoire. Some lack the space or plumbing to implement liquid cooling while others can’t get enough power to run demanding AI applications. Even for those data center operators with the budget in place to buy GPUs and other AI hardware, they could face long delays due to supply chain bottlenecks.

Despite these barriers, many are speculating that AI workloads are the future. They are investing heavily in what they perceive to be the data center of the future. Many are experimenting and developing their own AI tools and large language models (LLMs) internally. According to the Uptime Institute survey, the number of operators that have deployed AI in production is growing rapidly. This includes massive colocation companies, hyperscalers, and many of the Fortune 500. The big drivers of AI deployment are facility efficiency, elimination of human error, and higher staff productivity.

New services for data centers are also evolving. AI cooling optimization is one example. It offers a way to increase the effectiveness of cooling while lowering costs by targeting air where it needs to be and avoiding endlessly running cooling units when the cool air isn’t needed.

All of this would lead one to believe that people are gaining more confidence in AI. However, the survey found trust in AI to be declining for the third year in a row. That said, the majority said they would still trust an adequately trained AI model to make operational decisions in their data centers. Some trepidation remains in areas such as decision-making transparency, accountability, cybersecurity risk, and how AI could add another point of failure.

Data Centers Need Hybrid Architectures

The cloud forever changed the data center model. Instead of everything residing in-house, operators began to make decisions based on economics or convenience that placed many workloads in the cloud. AI will continue this trend. Many AI workloads will be hosted in the cloud while many others will reside within the data center. Thus, hybrid architectures will play an even bigger role in our collective AI future.

According to Uptime Institute, more than half of workloads (55%) are now outside of the data center. That’s up from 42% in 2020 and predicted to rise to 58% by 2026. Meanwhile, capacity expansion in the cloud and on-prem continues. So, although the slice of the overall pie for data centers gets slowly smaller, the amount of capacity inside data centers is increasing. Expect colocation workloads, too, to rise steadily. Large providers offer services and pricing that is hard for some in-house operators to beat.

“The size and capacity of these hyperscale developments is unprecedented, as they take their place on the global data center map alongside established and fast-growing data center hotspots,” said Andy Lawrence of Uptime Institute.” 

The Achievement of Sustainability Means Better Tracking of Metrics

Yes, data centers want AI. They want more power for their operations. They are demanding the latest GPUs and AI-related hardware. All of this tends to negate the net-zero carbon goals that many enterprises and data centers have set in recent years. So, while there is a headlong rush toward AI implementation, there is a growing realization that a sustainable middle ground has to be achieved – how to balance AI growth with environmental responsibility.

The first point to tackle, according to Uptime Institute, is measurement. Its survey revealed that fewer than half of data center owners and operators actively track sustainability metrics and meet regulatory requirements. The only metrics widely gathered are power consumption and PUE. Those are inadequate as they don’t give an encompassing view of sustainability or progress being made toward it.

What we have, then, is a collision course. Business leaders boldly laid out targets to achieve net-zero by 2030. They issued statements about sustainability being their top priority. Now they are issuing seemly contradictory statements about the need to implement AI fast, add more power, increase server density, and deploy liquid cooling. At some point in the near future, these opposing vectors will have to be reconciled. It remains to be seen which driver will prevail. Hopefully, innovation will play a role in relieving any friction.

Staffing Challenges Remain

Uptime Institute noted that staffing challenges remain a challenge in the data center. The shortage of skilled labor could be the hidden factor that slows down AI momentum and curtails significant progress in the attainment of sustainability goals. It is one thing for management to lay out visions for a sustainable AI tomorrow and quite another to have the manpower and resources at hand to implement it.

“More effort is needed to expand labor pools and skillsets to match the pace of capacity growth,” said Lawrence.