LinkedIn bucks AI data center spending trend

LinkedIn has decided against aggressive expansion of its AI data centers for its current fiscal year, a move that contrasts with many other major tech companies. Instead, the professional networking platform is prioritizing efficiency, keeping its investment in GPUs steady and its compute and storage footprint flat. This strategy is enabled by finding ways to use existing GPUs twice as efficiently over the past six months, a feat achieved through meticulous optimization of its data center usage.

Executives highlighted that this disciplined approach motivates engineering teams to innovate while developing new generative AI features. By optimizing its infrastructure and employing techniques like model distillation, LinkedIn estimates its efficiency efforts have saved approximately $24 million in the past 12 months, equivalent to powering 1,100 GPUs continuously for a year. This focus on 'tokenomics' signals a broader industry shift from mere spending to achieving more with existing resources.