Zuckerberg, After Major AI Talent Acquisitions, Says "Multiple Gigawatt-Scale Data Centers Under Construction"

Zuckerberg's AI War
Number and Locations of Data Centers Remain Undisclosed
Recruitment of Key AI Talent

Mark Zuckerberg, CEO of Meta Platforms, is accelerating efforts to strengthen AI infrastructure by launching the construction of large-scale data centers following a series of high-profile AI talent acquisitions. This move is interpreted as a strategic step not merely to expand AI infrastructure, but to establish an independent and self-sustaining "AI ecosystem" that will enable Meta to compete in the race for artificial superintelligence.

On July 14 (local time), Zuckerberg announced on the social networking service platform Threads that Meta is building several large-scale data centers, with the first scheduled to begin operations next year.

He added, "The first data center, called 'Prometheus,' is expected to go online in 2026," describing the data centers currently under construction as "clusters" on a gigawatt (GW) scale.

He further emphasized, "We are also building several even larger 'Titan Clusters,' one of which alone would occupy a significant portion of Manhattan."

He also noted that the second cluster, 'Hyperion,' could be expanded to a maximum of 5GW within the next few years.

A GW-scale facility refers to a unit of power equivalent to the output of a single nuclear power plant, highlighting the massive scale of the data centers under construction. Operating cutting-edge AI requires vast resources of graphics processing units (GPUs) and advanced parallel computing capabilities, both of which necessitate gigawatt-level power infrastructure.

Zuckerberg also shared a post from SemiAnalysis, a research and analysis firm in the semiconductor and AI sector, which noted that Meta will soon have its first "supercluster" with a capacity exceeding 1GW.

Currently, most data centers operate at the scale of several hundred megawatts (MW), with only a few technology companies such as OpenAI and Oracle developing GW-scale data centers.

Seemingly aware of concerns regarding the scale of investment, Zuckerberg emphasized that Meta's core advertising business remains robust, stating, "We plan to invest hundreds of billions of dollars to build artificial superintelligence, and we have the capital to execute this plan."

However, he did not disclose the exact number or locations of the data centers.

Reuters analyzed this announcement as "an effort to secure leadership in AI technology, in addition to acquiring top-tier AI engineers."

This declaration of large-scale data center construction comes as Meta has been aggressively recruiting AI talent in recent months.

Aiming to develop AI that surpasses human capabilities, Meta established the "Meta Superintelligence Research Lab" and recruited ScaleAI co-founder Alexander Wang, acquiring a 49% stake for $14.3 billion as part of its massive investment in AI talent. Additionally, Meta has recruited researchers from OpenAI, Apple, DeepMind, and Anthropic with compensation packages worth hundreds of millions of dollars, and has fully acquired the voice AI startup PlayAI.