Meta weighs AWS veteran as AI compute race intensifies

Meta Platforms is in talks to hire Dave Brown, a 19-year Amazon Web Services veteran, according to multiple reports cited by the CNBC Investing Club, as large technology groups compete for engineering talent and infrastructure capacity in artificial intelligence. The potential move comes as CNBC reported that Meta is also in preliminary discussions with Anthropic over a computing-power lease that could be worth up to $10 billion over two years.

Brown led AWS’s compute and machine learning services division, a role that included oversight of Amazon’s custom Trainium AI chips, according to CNBC. His departure would place another senior infrastructure executive at the center of the AI talent market, where cloud providers, chip designers and model developers are competing for staff who can reduce the cost and power intensity of large-scale AI systems.

In an interview with the CNBC Investing Club earlier this year, Brown said AI infrastructure would be shaped by improvements in “price performance,” meaning more computing output for each dollar spent. He said a chip or processor that delivers more performance for less money can give customers a strategic advantage.

That approach is central to the economics of generative AI. Training and running advanced models require large clusters of processors, high-speed networking and substantial electricity. Custom silicon, such as Amazon’s Trainium chips, is designed to improve efficiency for specific AI workloads rather than relying only on general-purpose processors.

Meta’s cloud ambitions

If Brown joins Meta, CNBC said his background could support the company’s interest in a cloud-computing business. Meta Chief Executive Mark Zuckerberg said at the company’s May shareholder meeting that such a strategy was “definitely on the table,” according to CNBC. Jim Cramer said earlier this month that Meta plans to sell computing power to outside customers, creating an additional revenue line.

The New York Times reported Friday that Anthropic was working on a deal to lease computing power from Meta, with a potential value of as much as $10 billion over two years. CNBC’s Kate Rooney confirmed the report and said the companies were in very preliminary talks, citing a person familiar with the matter.

A computing lease would allow an AI developer to use another company’s data-center capacity rather than building or buying all of the infrastructure itself. For the provider, the arrangement can turn existing or planned capacity into contracted revenue, though the economics depend on utilization, power costs, chip performance and the length of the customer commitment.

Markets pull back as AI names weaken

The reports landed during a weaker week for US equities. The CNBC Investing Club said stocks fell Friday, with pressure in Amazon, Alphabet, Meta and Microsoft weighing on major indexes. The S&P 500 was on track to lose more than 1% for the week, the Nasdaq was headed for a decline of more than 2%, and the Dow was roughly unchanged, according to the club.

Semiconductor and AI infrastructure shares faced early selling pressure, CNBC said, although Micron, Sandisk, GE Vernova, Eaton and Qnity Electronics rebounded as buyers came in at lower prices. Cramer said investors were broadening technology exposure into aerospace, where GE Aerospace rose nearly 3% and Honeywell Aerospace gained more than 2%.

Cramer also said healthcare could remain constructive, pointing to Abbott Laboratories for review and citing Johnson & Johnson after what he called a strong quarter. He said industrial stocks may be harder because of data-center exposure and pointed instead to transport names including FedEx and FedEx Freight.

Earnings season accelerates next week, with about 80 S&P 500 companies scheduled to report, according to CNBC. The calendar includes industrial, automotive and defense companies, as well as Alphabet and Intel. CNBC said Alphabet will be the first hyperscaler to report and that investors will be listening for commentary on capital spending and its June equity offering of $84.75 billion.

This story draws on original reporting from CNBC.