29 3月 2024
Microsoft and OpenAI are planning an unprecedented supercomputer that uses millions of specialized server chips and could cost up to $100 billion. This will be the most expensive data center endeavor in history, collaborating with OpenAI on a supercomputer installation that could exceed $100 billion in costs.
The ambitious project, set to span six years, aims to culminate in the launch of an advanced AI supercomputer dubbed "Stargate," with a projected operational date set for 2028.
Yet, the supercomputer is just one facet of this extensive five-phase initiative.
As part of their foray into AI-driven robotics, Microsoft and OpenAI are positioning themselves at the forefront of the industry — a move underpinned by Microsoft's significant financial infusion into OpenAI and consistent support for their AI developments since 2019. Microsoft's substantial cloud infrastructure currently powers the computational might behind ChatGPT.
With the conclusion of this massive project, Microsoft and OpenAI are looking to curate a suite of tools capable of training significantly advanced AI models, even beyond today's iterations of ChatGPT.
"In our constant quest to advance AI capabilities, we're future-proofing our infrastructure to push the boundaries of AI applications," said a Microsoft spokesperson.
An Overview of the Data Center Endeavor's Five Stages
The upcoming data center represents a continuation of Microsoft's strategic partnership with OpenAI. The tech behemoth is scouting US locations for this new facility.
Currently positioned within the third phase of the initiative, Microsoft and OpenAI are focused on constructing numerous supercomputers to boost ChatGPT's already impressive capabilities.
The fourth phase has Microsoft developing a supercomputer that, while smaller than Stargate, is anticipated to surpass most current models in power. The company aims to harness this supercomputer by 2026, possibly to enhance OpenAI’s voice recognition or facilitate text-to-video AI functionalities.
Industry analysts project the cumulative cost of these endeavors at roughly 115 billion allocated to the final two phases alone.
The Astronomical Cost of AI Data Centers Explained
Typical data centers come with a far less exorbitant price tag, generally falling between $250 million to a few billion dollars. The unprecedented estimated cost of Microsoft's proposed facility is indicative of the unique demands of artificial intelligence technology.
AI data centers necessitate considerable power, state-of-the-art hardware cooling mechanisms, an extensive inventory of chips (likely in the millions), and amplified cloud storage and data processing capacities.
With generative AI workloads demanding significantly more computational resources, these factors inevitably drive up the cost of cloud services, explained Tracy Woo, an analyst in information technology, to the Wall Street Journal.
With these industry dynamics at play, it wouldn't be surprising if future data center projects command price tags soaring above $100 billion.