Meta shares rally as investors reassess AI push

Meta shares climbed about 6% on Friday, taking the stock’s weekly advance to roughly 15% as investors responded to a series of artificial intelligence announcements and reports on the company’s chip plans. CNBC reported that the move put Meta on track for its strongest weekly performance since early 2024 and erased its losses for 2026.

The rally leaves Meta up more than 2% for the year, according to CNBC, although the stock continues to trail the Nasdaq Composite, which has gained 13% over the same period. The share move reflects renewed market attention on chief executive Mark Zuckerberg’s effort to build a broader AI business around models, paid services and computing infrastructure.

Meta announced two AI products this week. On Tuesday, the company released Muse Image, a model designed to generate images. CNBC reported that the product is part of Meta’s effort to appeal to creators and advertisers through new subscription offerings.

On Thursday, Meta unveiled Muse Spark 1.1, a version aimed at agentic and coding workloads. Agentic AI generally refers to systems built to carry out tasks through multiple steps with less direct human prompting than a conventional chatbot. Coding tools are a contested market within generative AI, with developers using models to write, test and modify software.

The new releases follow Meta’s April introduction of Muse Spark, which CNBC described as the company’s first proprietary foundation AI model. A foundation model is a large system trained on broad data sets that can be adapted for a range of uses, rather than a single narrow task. For Meta, such models are central to its attempt to compete with companies that established earlier positions in generative AI.

CNBC reported that Meta is seeking to challenge OpenAI, Anthropic and Google, all of which entered the market with a lead in high-profile AI models. The effort also fits Meta’s push to find revenue sources outside its core advertising business, according to CNBC.

The product announcements point to activity inside Meta Superintelligence Labs, the AI unit led by Alexandr Wang, CNBC reported. Wang joined Meta as Zuckerberg placed greater emphasis on building advanced AI systems within the company.

Meta also benefited this week from reports about its custom AI chips. Reuters reported that Meta expects to begin manufacturing its first in-house chip, code-named Iris, in September. The chip plan is tied to Meta’s data center expansion and to its goal of reaching 14 gigawatts of computing power next year, according to Reuters.

In-house chips can affect AI economics because model training and deployment require large amounts of computing capacity. If a company can use more of its own silicon, it may reduce reliance on external suppliers and potentially lower the cost of capacity, depending on performance, production scale and deployment timing.

Bank of America analyst Justin Post wrote after the Reuters report that Meta may have found material savings in the cost of capacity per megawatt, potentially below both his estimates and broader Wall Street expectations.

The market reaction remains tied to execution. CNBC reported that Meta’s recent announcements show an aggressive effort to establish itself in AI models, subscriptions and infrastructure, while the stock’s year-to-date performance still trails the broader technology-heavy Nasdaq benchmark.

This story draws on original reporting from CNBC.