• Meta Platforms is extending the operational life of certain non-AI data-center servers from six to seven years due to a persistent memory-chip shortage, according to internal memos cited by The Wall Street Journal.
  • The move aims to preserve capacity and avoid service disruptions amid global supply constraints expected to last through 2027.
  • Despite the shortage, Meta continues aggressive AI infrastructure investment, with 2026 capital expenditure projected in the low-to-mid hundreds of billions.

Meta Platforms is stretching the lifespan of some of its non-AI data-center servers to seven years, up from the typical six, as a persistent memory-chip shortage continues to squeeze hardware supply. According to internal company memos reviewed by The Wall Street Journal, the adjustment is a practical response to constrained chip availability, designed to prevent outages or capacity gaps. A Meta spokesperson declined to comment on the specific server extension but reiterated the company's focus on operational efficiency.

The decision underscores a broader industry struggle: global memory-chip supply remains tight, driven by surging demand for AI workloads and limited production capacity. For Meta, the shortage means prioritizing purchases of high-performance hardware for AI systems while running older non-AI servers longer. 'We have to be pragmatic about where we allocate scarce resources,' a person familiar with the matter said.

This hardware stretch comes as Meta doubles down on AI investment. The company has signaled 2026 capital expenditure in the low-to-mid hundreds of billions, with operating costs also climbing as it scales AI-powered features across Facebook, Instagram, and WhatsApp. Meanwhile, ad revenue remains strong, providing the cash flow to fund these ambitious plans.

Market reaction has been muted, with shares trading largely flat following the news. Analysts view the server extension as a short-term fix rather than a strategic shift. 'It's a sign of the times,' said one semiconductor analyst. 'Every hyperscaler is facing similar supply constraints. The question is how long they can kick the can down the road.'

Other tech giants are pursuing similar measures, including longer hardware lifecycles and supplier diversification. But the memory shortage is expected to persist through 2027, according to industry estimates, meaning these workarounds may become more common.

Correction: An earlier version of this article misstated the source of the server extension information. The details were from internal memos, not a company press release.