The Blubber

OpenAI's CEO Sam Altman has said an energy breakthrough is necessary for future artificial intelligence, which will consume vastly more power than people have expected, per Reuters.


The necessity for increased power in the field of artificial intelligence (AI) arises from the substantial energy consumption of semiconductors handling AI applications, surpassing expectations. According to Altman, the semiconductors powering AI software demand significant electrical power, leading to the anticipation of larger data centers hosting these advanced computers. These data centers, known as hyperscale data centers, are integral to supporting robust and scalable applications for major tech companies like Google, Amazon, Facebook, IBM, Apple, and Microsoft.

In addressing the energy requirements for these expansive hyperscale data centers, questions arise about the sources of electricity. Notably, these data centers, such as Apple's $2 billion Mesa facility in Arizona and Google's Council Bluffs center spanning over two million square feet, necessitate substantial electrical power, with Google's demand exceeding 100 MW. Cisco's data highlights that, despite representing less than 10% of the market, over half of all data processed in 2020 passed through hyperscale facilities.

Examining the hyperscale data center market, specific Tier 1 markets are identified, including Northern Virginia, Silicon Valley, Dallas, Chicago, Phoenix, New York, and Atlanta in the United States. In EMEA (Europe, the Middle East, and Africa), the FLAP markets – Frankfurt, London, Amsterdam, and Paris – along with Dublin and Madrid are crucial. APAC (Asia-Pacific) Tier 1 markets encompass Greater Tokyo, Singapore, Sydney, and Hong Kong.

The flexibility of hyperscale companies in choosing their location is contingent on robust connectivity. DCD emphasizes that, with strong connectivity, these companies can opt for rural areas based on the availability of reliable energy and affordable land.