Adds blog

Welcome to the home for real-time coverage of markets brought to you by Reuters reporters. You can share your thoughts with us at markets.research@thomsonreuters.com.

GENERATIVE AI'S ROI STILL UNCLEAR, WARNS EDMOND DE ROTHSCHILD CIO

Private financing is growing among AI players - as are valuations - but the issue of return on investment for generative AI technology has never been addressed in earnest, says Benjamin Melman, global chief investment officer at Edmond de Rothschild Asset Management.

In a letter to clients he says private financing has grown massively, particularly among AI players competing in what he calls a "frenetic race" to build new infrastructure.

"But if the high valuations within the sector are to be reckoned by, the issue of ‘return on investment’ was never addressed in earnest," he writes.

He refers to a study by Goldman Sachs that showed the Big Five tech firms - namely Microsoft, Amazon, Google, Meta, Oracle - will have poured $380 billion this year into capex, mainly to support generative AI.

"The study also highlights that for most ‘hyperscalers’, the return on investment is higher than the cost of capital. In other words, operating margins are large enough to handle the expansion. But this does raise the question of future earnings forecasts," says Melman.

He also flags Cloud/AI segments which currently display an investment/sales ratio of almost 35%, compared with a historical average around 15%, raising barriers to entry but increasing systemic risk if one supplier suddenly cuts its budget.

"Interestingly, after Oracle issued a substantial bond, Moody's flagged several potential risks: high reliance on large AI contracts, debt that would increase faster than its EBITDA, negative cashflows for an extended period, and high leverage of 4x before Oracle’s EBITDA begins to surpass its debt," writes Melman.

The CIO also flags the fact that physical and power capabilities are proving to be limited.

"....notably owing to saturated power grids around large hubs (North Virginia, Dublin, Frankfurt) and to the extended timeframe needed to develop a new datacenter."

Summing up, he says AI requires massive, energy-hungry, and costly infrastructure that takes a long time to become profitable.

(Lucy Raitano)

*****

EARLIER ON LIVE MARKETS

ARE CHIP SHORTAGES THE NEXT PROBLEM FOR EUROPEAN AUTOS? CLICK HERE

FUND MANAGERS STARTING TO TAKE OFF FX HEDGES CLICK HERE

BANKS, DEFENCE TUMBLE, ESSILORLUXOTTICA SHINES CLICK HERE

EUROPE BEFORE THE BELL BANKS ARE BACK ON THE AGENDA CLICK HERE

MORNING BID: ADD CREDIT RISK TO A BUBBLE, AND STAND WELL BACK CLICK HERE