In cooperation with Meta Platforms’ (NASDAQ:META) fourth-quarter results and guidance that stunned Wall Street this week, the Mark Zuckerberg-led company has increasingly focused on “efficiency,” mentioning it more than 40 times in its earnings call.
With a push toward greater streamlining, Meta ( META ) Chief Financial Officer Susan Li now expects the company’s annual expenses to be between $89 billion and $95 billion in 2023, down from a previous forecast of $94 billion to $100 billion. While some of that is related to lower headcount, a good portion of the drop in spending comes from lower capital spending, particularly in the data center, which could have implications for Arista Networks (NYSE: ANET), Nvidia (NASDAQ: NVDA) and Clean Storage (NYSE: PSTG).
Analysts at Wells Fargo said in a research note that “from a lateral perspective” investors could view Meta’s reduced capital investment, albeit expected, as a “net negative valuation” for Arista Networks ( ANET ), Nvidia ( NVDA ) and Pure Storage ( PSTG).”
Analysts noted that Meta’s ( META ) guidance for capital spending was between $30 billion and $33 billion, down from a previous forecast of $34 billion to $37 billion, a 2% decline from last year. Much of this is believed to be due to a shift to a new data center architecture that is considered more “cost-effective” and can support AI and non-AI workloads.
On Meta’s ( META ) earnings call, Li described the transition to the new architecture as a way to give the company “more options” as it better understands its AI needs over time.
Li also said the company will “optimize,” or be more efficient, with its approach to building data centers. Now Meta (META) will build basic plans with less initial capacity and less money spent up front, but then add capacity if needed.
These comments seem to indicate that some of Meta’s (META) suppliers may be affected.
Led by CEO Jayshree Ullal, Arista ( ANET ) likely got 10% or more of its revenue from Meta ( META ) in 2022 as the company built its AI fabric networks.
For Nvidia ( NVDA ), the company deployed 16,000 GPUs to help build the Meta ( META ) Research SuperCluster supercomputer, one that could be the “world’s fastest.”
RSC is being trained with natural language processing and computer vision for research and could one day train models “with trillions of parameters,” Meta said last year.
In addition, Pure Storage ( PSTG ) also contributed to the RSC, the Wells Fargo team said, providing its all-Flash storage arrays for the supercomputer.
For years, investors have called for Meta (META) to become more efficient with money, with some analysts having referred to as the broader tech sector that spent like “1980s rock stars.”
Finally, the company listened and investors rewarded it. However, some of his clients may not be so lucky.