Meta, parent company of Facebook, is planning another round of layoffs in March, according to the Financial Times. This comes just one week after Meta revealed its earnings results for 2022 and its economic outlook for 2023. In that reporting, Meta announced a $4 billion cut to its CapEx, which includes data center assets.
The data center budget cuts and predicted layoffs in March are part of Meta CEO Mark Zuckerberg’s ‘Year of Efficiency’ management theme for 2023. This belt-tightening has spread to other hyperscalers, including cloud service providers (CSPs) AWS, Azure, and Google Cloud Platform. All three companies have slashed their workforces during the last several months, leading tech and financial analysts to predict that the cloud compute sector will slow its growth. With this slowdown comes a greater emphasis on AI technology and facilitating its use. For instance, Meta CFO Susan Li has spoken about the firm’s AI-focused data center architecture efforts.
“This is a new architecture that we've been developing, and it’s going to replace the prior data center architectures, and it’s much more flexible in terms of accommodating both A.I. and non-A.I. workloads,” said Li during Meta’s Q4 2022 earnings follow-up call.
Li went on to highlight features of Meta’s phased data center buildouts:
- Low-commitment base plans for initial capacity
- Fast subsequent buildouts as demand increases
- Designs supporting high-density racks (more computing power for less space)
For those in the data center game, these aspirations aren’t new but are rather the plan for all data center builds. Yet building around a specific technology, AI, poses an interesting approach that may lead Meta to create a template for data centers to follow for incremental buildouts.
It’s important to note this talk of high-density racks wouldn’t be credible without contributions from chipmakers such as AMD, Intel, Arm, Samsung, and Micron. AMD’s innovations with Ryzen 7000 and MI300 make higher rack density in a smaller footprint more than just aspirations.
Chipmakers have seen data center semiconductor chip sales save their profit structures as PC sales slow. Nvidia’s Q3 2022 beat estimates due to data center business growth. And AMD’s Q4 2022 received a notable boost from sales on the data center side of its business.
Meta’s high-density, low footprint approach could depend on development of chips that facilitate AI workload processing. AMD’s Ryzen is the first x86 chip to have an AI processor directly on the chip. With the firm’s surging market leadership, other chipmakers may follow this business model.