The rush to deploy artificial intelligence, especially generative AI, will drive hyperscale data center providers like Google and Amazon to nearly triple their capacity over the next six years.
That's the conclusion of Synergy Research Group, which tracks the data center market. In a new report, Synergy notes that while there are many exaggerated claims about AI, there is no doubt that generative AI is having a significant impact on the IT market.
Synergy's latest six-year forecast shows that the average capacity of new hyperscale data centers will soon be more than double that of currently operating hyperscale data centers. And the total capacity of all operating hyperscale data centers will nearly triple between 2023 and 2028.
The study is based on an analysis of the data center footprint and operations of 19 of the world's leading cloud and internet services companies. By mid-2023, these companies operated 926 major data centers around the world. Synergy's future data center projections include adding an additional 427 facilities over the next few years.
Synergy says the impact of advances in generative AI has not only driven an increase in the number of data centers, but has also led to a significant increase in the amount of power required to operate those data centers. As the number of GPUs in hyperscale data centers proliferates, primarily due to AI, the power density of associated racks and data center facilities must also increase significantly. This requires hyperscale operators to rethink some of their data center architectures and deployment plans.
So if managing something like this is a headache for AWS, what about the average company running servers that are 5 years old?
Companies are rushing to implement generative AI to improve their business, but the cost of purchasing and operating the hardware is holding many back. Nvidia's DGX servers are custom-built for generative AI, with embedded hardware that can easily run in the six-figure range. With that much money, he could buy about 10 regular servers. Which will companies prioritize?
Additionally, there are costs involved in operating them. Nvidia GPUs aren't known for low power consumption. It's quite the opposite. These are the largest power consumers in data centers. Therefore, deploying generative AI hardware may be too much of a burden for budget-conscious businesses, especially mid-sized businesses.
Additionally, AI operates differently than traditional line-of-business applications. There's the process-intensive task of training, which requires GPUs, and then there's inference, which runs the trained model on GPUs. Once your model is trained, you likely won't need to modify it for several months. When this happens, very expensive hardware is left unused and its value decreases.
Can companies do this on their own without using a hyperscale cloud provider? “Theoretically yes, but the cost would be prohibitive and access to the right expertise would be significantly reduced. may be limited,” said John Dinsdale, chief analyst and research director at Synergy Research Group.
Therefore, the emerging trend for generative AI in enterprise IT is to outsource the training part of the AI and do the much less process-intensive inference in-house. Why invest hundreds of thousands of dollars in hardware you won't use often when you can rent it from Google or his AWS?
This is known as AI as a service and is an emerging offering from hyperscalar companies like AWS and Microsoft. You can expect more from this.
