Cloud roundup: AI increases data center costs; hyperscalers bogart capacity

In this cloud roundup, increased use of artificial intelligence (AI) is driving up data center costs, said The Wall Street Journal. In addition, new data revealed that data center capacity use by hyperscalers and colocation companies is on track to outpace that of enterprise on-premise data centers, according to Synergy Research Group.

Increased demand for AI is driving up data center costs

As the use of AI skyrockets, some data centers are responding by increasing prices for customers to offset the additional costs associated with supporting AI, according to The Wall Street Journal.

"Artificial intelligence applications burn more energy than conventional software because they are designed to read through a far larger amount of data. Over several days, a single AI model can consume tens of thousands of kilowatt-hours," wrote Angus Loten in The Journal. "Generative AI models, such as the technology underlying OpenAI's ChatGPT chatbot, can be up to 100 times bigger than standard AI tools."

Demands on processing power, the need for advanced chips, supply chain challenges and the increased use of AI are among the expenses data centers are trying to offload on customers, said the WSJ. Increasing construction, labor and utility costs are also driving up prices.

"In Northern Virginia, the world's largest data center market with more than 275 facilities, the amount of power available for lease this year shrank to 38.4 megawatts, from 46.6 megawatts a year ago, according to the latest analysis by commercial real estate services firm CBRE Group," wrote Loten.

CBRE reported that data center customers in the Northern Virginia region are paying as much as $140 a month per kilowatt of power, up from $130 in 2022 – an increase of nearly 8%. By comparison, in Silicon Valley, customers pay as much as $250 a month per kilowatt of power, a 43% increase from $175 in 2022.

Capacity demands from hyperscalers and colocation companies outpace that of on-premise data centers

Hyperscalers are using 37% of global capacity of all data centers, according to 2022 data by Synergy Research Group.

About half of the capacity used by hyperscalers stems from their own data centers, and the remaining half is from leased data centers. Non-hyperscale colocation facilities use 23% of capacity. On-premise enterprise data centers used 40% of data center capacity in 2022, down significantly from 60% in 2017.

Synergy predicts hyperscalers will use more than 50% of capacity by 2027 and that on-premise enterprise data center capacity will decrease to less than 30%.

Over the next five years, the firm forecasts that "while on-premise share of the total will drop by over two percentage points per year, the actual capacity of on-premise data centers will decrease only marginally. Colocation share of total capacity will remain relatively constant."

Over the last decade, enterprise spending on cloud services has increased by about 42% per year to $227 billion in 2022, up from $10 billion in 2012. In contrast, enterprise spending on data center hardware and software has increased only about 2% each year, said Synergy.

"As enterprises radically reshaped their IT investments and crimped spending on their own data centers, the leading cloud providers rapidly built out huge global networks of hyperscale data centers," said the firm. Enterprises are also moving more of their data center equipment offsite to colocation facilities.

Social networking, e-commerce and online gaming are among the digital consumer services that have played a large role in hyperscaler growth. There are about 900 data centers operated by hyperscalers, according to Synergy.

Related posts:

— Kelsey Kusterer Ziser, Senior Editor, Light Reading