Recent reports suggest that data centers are under pressure from a surge in demand for generative AI, imposing some costs on customers.
Although ChatGPT is gaining momentum after its launch and immediate success, generative AI is a capital-intensive operation that requires higher levels of computing power.
According to real estate company JLL, this increase in demand has also led to rising infrastructure costs, which have soared this year.
hard to find space
In North America, for example, demand for colocation space is outstripping supply, leading to rising costs, according to JLL, which analyzed data center market trends.of report notes As a result, customer costs increased by 30%.
“Major cloud service providers are growing rapidly to support new AI requirements and the need for more computing power, and many markets are finding it difficult to find the space and power to accommodate smaller requirements. “This is becoming increasingly difficult,” the report says.
It further highlights that this led to “a significant surge in leasing in Q2 2023 due to increased capacity demand to meet high-density data center requirements for AI development.”
Also read: ChatGPT gets audio and image capabilities, making it more like Apple’s Siri
Further increases in costs
JLL’s report aligns closely with other studies conducted on the impact of generative AI on data centers globally. Research firm Tirias Research predicts that demand for data center infrastructure and the cost of generating AI could exceed $76 billion by 2028. This is more than double the current estimated cost for Amazon Web Services (AWS), which accounts for a third of all cloud usage. infrastructure services market.
A separate study from TD Cowen earlier this year showed data center lease costs reaching “record levels” as hyperscalers continue to expand their AI development capabilities.
The study concluded that the industry is experiencing a “tsunami of AI demand” and that some service providers are unable to keep up with demand from hyperscalers.
However, JLL points out the need for: Improving data center efficiency and adapting the infrastructure to accommodate “high power density clusters.”
According to the ITPro report, infrastructure overhauls may result in changes to cooling capabilities and energy consumption.Recent Silicon ANGLE The report also found that data centers are consuming large amounts of water to cool production AI servers as technology usage increases.
Microsoft reported a 35% jump in data center water consumption in 2022 compared to 2021. According to the report, the company used more than 1.7 billion gallons (6.4 billion liters) of water in 2022 alone. There are over 2,500 Olympic-sized swimming pools. ”
Data centers also consume a lot of energy, so it’s not all about high water consumption.
According to a JLL report, some larger requirements can increase cluster density to 50-100 KW per rack. This represents a sharp increase compared to current hyperscaler demand, the report adds.
“Many colocation providers are increasing the voltage delivered to the floor to 415 volts, which can reduce the initial cost of powering these high-density clusters,” JLL said in the report. There is.
“Given the sustainability goals of hyperscalers and colocation providers, further innovation will be required to improve cooling and energy efficiency for AI applications,” he added.
What should I do?
Adam Nethersall, vice president of Kao Data, said in an interview with ITPro that generative AI-related demand is straining data centers, creating a space shortage in the U.S., but what can be done to accommodate the increase in activity. He said it would have a “positive impact” on Europe. UK data center provider.
“Capacity shortages in the US data center market are likely to be exacerbated by both the introduction of generative AI and the associated growth of hyperscale platforms, while in more localized regions of the UK and Europe. We are seeing a positive impact on the sector,” he said. .
He added that demand continues to surge. HPCAI, and GPU-powered technologies were accelerating the development of new Tier 2 data hubs.
According to Nethersole, this is causing a “complete redesign or re-engineering of data center capacity to accommodate high-density computing.”
“This includes the introduction of new high-density, pre-configured systems such as Nvidia’s SuperPOD, increased adoption of liquid cooling in some cases, and continued commitment to ultra-energy-efficient colocation capacity powered by 100% renewable energy. This includes specific requirements.”
JLL’s report highlighted the impact of generative AI development on customers who require “small scale” colocation services, i.e. companies that require data center infrastructure for operational purposes other than generative AI development.
In the long run, Nethersall said, Generation AI During development, some customers may be prohibited from accessing the colocation service.