Microsoft and Amazon Stocks Drop. This Analyst Downgrades Them ‘With a Heavy Heart.’

Rothschild & Co. Redburn analyst Alexander Haissl downgraded Amazon.com and Microsoft to Neutral from Buy and lowered his price target on Microsoft shares. (Getty Images for Amazon Web Services)

Key Points

• Rothschild & Co. Redburn downgrades Microsoft and Amazon.com to Neutral from Buy due to concerns about generative AI economics.
• Analyst Alexander Haissl states that generative AI infrastructure requires six times more capital for the same cloud 1.0 value.
• Each dollar invested in generative AI infrastructure yields about 20 cents of net present value, compared to $1.40 for cloud 1.0.

After months of unflagging enthusiasm for artificial-intelligence stocks, it’s time to take “a more cautious stance” on hyperscalers like Microsoft and Amazon.com , according to Rothschild & Co. Redburn analyst Alexander Haissl, who downgraded stocks to Neutral from Buy in a note Monday.

Haissl cut his price target on Microsoft shares to $500 from $560 and maintained a $250 price target on Amazon.

Amazon stock declined 2.8% to $226.43 while Microsoft dropped 2.5% to $494.89.

The downgrade comes “with a heavy heart,” Haissl wrote, noting that the firm had maintained an above-consensus bullish view on both stocks since initiating coverage in 2022.

“Until now, we held the view that economics would eventually revert to cloud 1.0 levels (the pre-Gen-AI period up to early 2024), when hyperscalers captured most of the value,” the analyst explained.

That doesn’t appear to have panned out. The argument that “GenAI is just like early cloud 1.0” seems “increasingly misplaced,” Haissl continued. He stressed that the economics underpinning the bullish argument are “far weaker than assumed.”

The data center buildout is far more capital-intensive than building out cloud infrastructure, he added, noting that GPU deployments “require roughly six times more capital to generate the same cloud 1.0 value, with risks skewed to the downside.”

Essentially, the enormous amount of capex required skews the risk/reward balance more toward the negative. Haissl asserted that each dollar invested in generative AI infrastructure produces only about 20 cents of net present value, a measure that compares the value of cash inflows and outflows to gauge profitability. This compares to roughly $1.40 for mature cloud 1.0 projects, Haissl said.

We’ve seen this argument before, that AI doesn’t generate sufficient returns to justify massive capital spending. A MIT report and commentary from figureheads like Sam Altman over the summer sparked the market paranoia that has caused investors to increasingly sour on AI’s prospects.

Following an AI-fueled peak in late October, the Nasdaq has fallen more than 5% as investors back away from AI stocks, citing lofty valuations.

In Microsoft’s case, Haissl argued that growth in Azure, the company’s flagship AI platform, is increasingly tied to lower-value workloads, while “value increasingly flows to model providers like Anthropic or OpenAI.”

And while you may assume Amazon is in a better place, generative AI as a whole has become the industry’s “low value growth engine”—contributing more than half of AWS’s and Azure’s revenue growth, but only through heavy capex, Haissl wrote.

The real constraint to consider is “the capex required relative to the growth delivered,” the analyst continued. He pointed to the higher risk of overbuild, a fear that crept into the market earlier this year particularly in relation to Microsoft, when TD Cowen indicated the company was canceling data-center leases after building out too aggressively.

Crucially, “Gen-AI scales on a bloated, inefficient stack, while cloud 1.0 scaled only after achieving efficiency,” Haissl wrote.

Heavy is the heart—and the stocks.

Write to Mackenzie Tatananni at mackenzie.tatananni@barrons.com