It’s been a turbulent few weeks, with tariff uncertainty fueling broader concerns about the macroeconomic outlook and, specifically, about the AI spending environment. Even if semiconductors were initially expected to be spared from tariffs, the prospect of an economic downturn affecting the core businesses of major hyperscalers—such as declining advertising revenues at Google and Meta, or slowing software demand at Microsoft—had raised questions about AI capex intentions and investors in AI infrastructure names had naturally started to price in this risk accordingly.
With the Trump administration pausing its tariff initiative last night, two major companies in the AI space are out today with reassuring comments and figures showing that AI infrastructure remains well on track.
First, Google CEO Sundar Pichai confirmed during the company’s Cloud Next event its guidance of $75 billion capex for 2025, primarily focusing on expanding its cloud computing capabilities. The company justified this level of investment by the strong demand for its cloud services despite potential challenges from rising tariffs.
Google also took the opportunity to introduce its seventh-generation custom AI chip, Ironwood, which delivers up to 4,614 TFLOPS and offers 2× the efficiency of last year’s TPUs. Ironwood is engineered for large-scale inference, particularly for large language models (LLMs), with the memory and bandwidth needed to meet the rising demands of Generative AI.
Second, Taiwan Semi, which manufactures most of the AI chips, reported impressive revenue growth of 46% in March, leading to a 42% growth in its first quarter. While some of the March strength could be due to some pull-forward orders before the implementation of US tariffs, the overall numbers are reassuring and show that spending on AI chips kept growing at a fast pace in recent weeks.
With leading AI names such as Taiwan Semi and Nvidia now trading at very reasonable multiples (14x and 22x for 2025, respectively, 12x and 18x for 2026), it’s pretty obvious that the market has priced in significant AI capex cuts. Should the tariff chaos definitely ease and AI spending remain on track, a rerating would be well deserved in our view.
