Meta's stock roars back to life as it notches its best week in years

By Christine Ji

Investors are bullish on Meta's new low-cost AI pricing and infrastructure plans

Meta is reportedly on track to deploy 7 gigawatts of computing capacity this year, scaling to 14 gigawatts by 2027.

Shares of Meta Platforms are riding a fresh wave of investor enthusiasm following a slew of artificial-intelligence developments this week, including the introduction of the company's new Muse Spark 1.1 model.

Meta's stock (META) gained 6% on Friday, finish the week up 14.8% and seal its best weekly performance since at least February 2024, according to Dow Jones Market Data. The company's shares have now crossed into positive territory on a year-to-date basis.

When Meta's stock went on its strong weekly run back in February 2024, investors were reacting positively to the early results of its "Year of Efficiency," a cost-cutting initiative meant to show discipline after an expensive and ineffective venture into the metaverse.

Shares of Meta are now up 1.4% since the beginning of the year. This week's breakout performance hints that Meta's stock could finally shed its label as an AI loser - opening the door for the company to potentially spend even more on its AI strategy.

Muse Spark 1.1 is Meta's first model to boast near-frontier agentic coding capabilities and a paid application-programming interface. On Thursday, CEO Mark Zuckerberg emphasized in an X post that the model would be available at "a very low price," fueling social-media speculation that the tech giant could kick off an AI-token price war and squeeze competitors. The new Meta Model API offers every account $20 in free credits paired with a pay-as-you-go rate of $1.25 per million input tokens and $4.25 per million output tokens.

Radio Free Mobile founder Richard Windsor believes the Muse Spark 1.1 launch supports recent reports that Meta is planning a new business line selling computing access. "There is growing evidence that it will unveil a new business of selling compute to [third] parties given the very large returns that are currently on offer," Windsor wrote in a Friday note.

Muse Spark is now almost on par with top AI coding models "but at 25% of the price," he added, which would make it an attractive option for the mass market.

Additionally, a Thursday Reuters report said that Meta is on track to begin mass production of its custom chips, designed to reduce the cost of AI workloads. Meta is also reportedly planning to deploy 7 gigawatts of compute this year and 14 gigawatts by 2027.

Representatives from Meta declined to comment.

In a Thursday note, Deutsche Bank analyst Benjamin Black wrote that the reported capacity additions imply the potential for roughly $24 billion in incremental third-party cloud revenue, significantly higher than his prior estimate of $17 billion. Meta's custom-chip business could also create "a credible path to lower blended cost per unit of compute," he said.

-Christine Ji

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

07-10-26 1825ET