Meta's Mark Zuckerberg says energy constraints are holding back AI data center buildout
"We would probably build out bigger clusters than we currently can if we could get the energy to do it"
News - The Compute, Storage & Networking Channel, Facebook
The latest news from the AI, data center, telco, chip, and cloud sector
-
-
Meta plans $37bn digital infrastructure investment in 2024
Facebook parent company anticipates high levels of spending to cater for AI demand
-
Meta to operate "600,000 H100 GPU equivalents of compute" by year-end
Including 340,000 H100 Nvidia GPUs at its data centers
-
Meta spends $7.1bn on data centers, servers, and office capex in Q1, still assessing "data center restructuring initiative"
Expects to spend big on generative AI
-
Meta's data center redesign due to it scrapping AI chip rollout, turning to Nvidia GPUs
Company hires Microsoft exec to lead new AI chip efforts
-
Google increases server life to six years, will save billions of dollars
While Meta ups to five years
-
Meta opens logistics hub in Las Vegas warehouse district
Facility will handle logistics for Meta’s data centers
-
As share price crashes, Meta says it plans more data center buildout, but will extend server life
Will build new data centers designed for "next-generation AI hardware"
-
Meta/Facebook to use dedicated Microsoft Azure cluster for AI supercomputing
Opting to use the cloud, rather than internally building it
-
Meta to use time cards from Orolia to synchronize data centers
Using an alternative version of a Facebook design
-
Facebook/Meta builds huge Nvidia DGX A100 system, will become "the fastest AI supercomputer in the world"
AI Research SuperCluster will grow to include 16,000 GPUs
-
OCP: Meta updates Facebook's data center switches, moves to SAI open source API
And shows new hardware built by its OEM partners
-
Facebook/Meta to deploy AMD Epyc server chips
A major blow to Intel
-
Facebook is developing in-house AI chips for video transcoding, recommendations
Just like Google, Facebook expands its silicon ambitions
-
Facebook starts work on final phase of $1.5bn Fort Worth data center campus
Started in 2015, the data center campus will finally be completed in 2022 and employ 200 staff
-
Marvell to buy Inphi in $10bn semiconductor deal
Yet another billion-dollar chip deal
-
Facebook and Carnegie Mellon team up for AI-led energy storage research
Releasing a huge data set to kickstart catalyst research
-
Facebook details new OCP server 'Sonora Pass'
An improvement in modularity and security
-
Cumulus' Linux OS first to support Facebook's modular Minipack
The Linux based OS will be available pre-installed on Facebook's network hardware
-
Intel, Google and others join forces for CXL interconnect
CXL is designed to create a high-speed, low latency interconnect between the CPU and workload accelerators
-
Facebook nabs Google chip exec to lead silicon initiative
Could be developing FPGAs and ASICs for the server market
-
Facebook explores building its own ASICs and FPGAs
Its Own Compute Project
-
OCP Summit: Open Network Linux is ready for mass adoption
Open source project brings together Facebook, Google and Big Switch Networks
-
Intel announces new AI chip, with input from Facebook
Nervana Neural Network Processor aims to handle deep learning workloads in the data center
-
White Space 70: Reports from the field
The editorial team talks about the places we’ve been and the things we’ve seen
-
White Space 69: New and improved
Now with 33 percent female presence
-
Telecom Infra Project recruits Microsoft, BT, Airtel and more
While Voyager switch gets some new users
-
White Space 65: So long, 2016
What an eventful year we’ve had