Meta intends to invest hundreds of billions of dollars in artificial intelligence infrastructure.

 Meta CEO Mark Zuckerberg announced Monday that he plans to invest "hundreds of billions of dollars" in AI infrastructure, the latest major investment announcement to achieve his stated goal of building "superintelligence."

"We're building multi-gigawatt networks," Zuckerberg said in a post on his Facebook and Threads pages.

The new data centers, completely different from existing ones, will be specifically designed to train and host generative AI models.

AI models require massive computing power, and thus advanced computer chips and a significant amount of energy.

Meta has lagged behind in the field of generative AI, which has been dominated by OpenAI since the latter launched its ChatGPT model, along with Google.

All industry leaders are focusing their attention on so-called "general" or "superintelligence" AI, which possesses cognitive abilities beyond those of humans and is capable of making scientific discoveries and inventing technologies on its own.

Mark Zuckerberg explained that "our first 'computing network,' which we've named Prometheus, will go live in 2026. We're also building Hyperion, which will be capable of reaching 5 gigawatts of capacity within a few years."

5 gigawatts is equivalent to the annual electricity consumption of between one and four million American homes.

According to OpenAI CEO Sam Altman, Meta offered "many" OpenAI employees individual signing bonuses of more than $100 million and roughly the same amount in annual salary.

Zuckerberg later confirmed in an internal memo that at least seven employees from the rival company had joined his company, following the example of Scale AI CEO Alexander Wang and several members of rival companies Anthropic and Google.

The billionaire explained on Monday that he intends to build "the most unique and talent-dense team in the entire industry" with "a level of computing power unmatched in the industry."

"We have capital generated from our operations to do that," he continued.

Meta launched one of its AI models, called "Lamma 4," in early April, but it was disappointing.