(Bloomberg)– The case for generative artificial intelligence is no longer clear and hyperscalers should be approached with caution, according to an analyst who downgraded two of Wall Street’s most favored tech giants.
Microsoft Corp. and Amazon.com Inc. saw their ratings dropped from buy to neutral on Tuesday, as Rothschild & Co Redburn’s Alexander Haissl downgraded the stocks for the first time since coverage began in June 2022, according to data compiled by Bloomberg. Shares of both companies fell more than 2% in New York.
Most read on Bloomberg
The industry narrative that “trust us – the AI generation is like the first Cloud 1.0” seems “increasingly misplaced,” Haissl wrote in a note, saying the underlying economic factors are “much weaker than expected.”
Haissl’s downgrades follow a selloff in the tech-heavy Nasdaq 100 index, which has wiped out nearly $1.8 trillion since a peak in late October, sending the index down 5.1% through Monday’s close. Investors have pulled out of AI-related stocks due to concerns over strained valuations.
With this call, Haissl broke with most of his peers. More than 90% of analysts who cover the stocks have equivalent buy recommendations on them — without a single sell rating, according to Bloomberg data.
Former director of automotive research at Berenberg and Credit Suisse Group AG, Haissl began his career as a police officer in Vienna. He joined the company, then called Redburn Ltd., in 2022 and initiated coverage on Amazon and Microsoft with a buy rating citing their untapped cloud potential. While the cloud computing and e-commerce giant more than doubled over the period, Microsoft grew 98%.
Haissl’s coverage covers only three other stocks. He started with a sell rating on Oracle Corp. in September – shares of the software company have fallen about 25% since then. However, his call to downgrade Snowflake Inc. to neutral in March 2024 did not have the same result: the stock rose about 60%. It also has a neutral rating on MongoDB Inc.
Tech companies are spending tens of billions of dollars to build out their AI infrastructure. In recent weeks, investors have become increasingly concerned about how quickly depreciating assets such as graphics processing units and servers are losing value.
“Generation AI margins already assume longer payback schedules, 5 to 6 years, compared to just three years at the start of the cloud era,” Haissl said. “On a comparable basis, this means that the capital intensity for the AI generation is significantly higher, while the pricing power is significantly lower. »
