Meta renting out computing power is like using the old to nurture the new, rather than stopping the pursuit of high-end computing power.

robot
Abstract generation in progress
Mars Finance News: Meta is planning to launch a cloud computing business that rents out computing power. One type is to open up the model capabilities deployed on its own AI infrastructure to external customers; the other is to rent out more underlying “bare computing power.” But based on what seems to be the case so far, the real situation is more likely to be this: Meta is looking to recover cash flow from its existing stock of older, outdated computing power, rather than stopping its pursuit of high-end computing power. In mid to late June, it was reported that Meta has signed an agreement with Crusoe to obtain a total of about 1.6GW of AI computing capacity from two data centers in Texas and Missouri. Meanwhile, Meta 26Q1 has also raised its full-year capital expenditure guidance to $1250–1450 billion. Taken together, the two developments look more like a reallocation of resources across different generations and for different purposes: Meta will continue buying new cards to train frontier models, while older cards (such as the H-series) used for inference for high-traffic products, hosting external models, and other application scenarios can accept renting out part of their capacity—but this does not mean that Meta will slow down purchases of the most scarce high-end cards. (Cailianshe)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned