Back to the AI front line? Meta surged 15% in a single week

Author: Bao Yilong; Source: Wall Street Journal

Meta logged its strongest weekly performance since February 2024 thanks to a series of AI strategic breakthroughs this week, and market confidence in its AI capabilities and commercialization path is rebounding significantly.

On Friday this week, Meta’s stock rose 6% in a single day, with a cumulative gain of 14.8% for the whole week, marking at least its best weekly performance since February 2024, and lifting its year-to-date gain back into positive territory to around 1.4%.

The last time Meta posted similar strong weekly performance was in February 2024, when investors reacted positively to early results of the company’s “Year of Efficiency” cost-cutting plan, which aimed to reverse the negative image brought by its previous massive bet on the metaverse through financial discipline.

The stock breakout this time suggests Meta may be gradually shedding the market label of an “AI laggard,” which would also create room for it to further step up its AI strategy.

Wall Street Journal noted that on July 9, Meta launched its flagship model Muse Spark 1.1, which has already surpassed Google’s Gemini across multiple test projects including Agent capabilities, programming, and multimodal tasks.

Meanwhile, Reuters reported that Meta is pushing forward with its own chip mass-production plan and has significantly expanded its compute infrastructure. On that basis, Deutsche Bank analyst Benjamin Black raised Meta’s expected potential incremental revenue from third-party cloud services from $17 billion to $24 billion.

Research firm SemiAnalysis released a report estimating that Meta’s Meta Superintelligence (MSL) could surpass Google in leading frontier AI capability rankings within the next six months, shifting the AI competitive landscape from the current duopoly dominance of Google and OpenAI to a three-way standoff among Meta, OpenAI, and Anthropic.

Muse Spark 1.1 low-price strategy: targets a price war

The Muse Spark 1.1 model launched by Meta this week is its first commercialized model with agent programming capabilities at nearly state-of-the-art level, and it comes with a paid API interface.

On Thursday, CEO Mark Zuckerberg posted on the X platform, stressing that the model’s pricing is “very affordable,” sparking widespread speculation that Meta will proactively initiate an AI inference pricing war to pressure competitors.

Meta Model API provides a $20 free credit to each account and charges on a pay-as-you-go basis. The input price is $1.25 per million tokens, and the output price is $4.25 per million tokens.

Richard Windsor, founder of Radio Free Mobile, wrote in a research note on Friday that the launch of Muse Spark 1.1 confirms recent reports that Meta plans to launch a new business selling compute capacity to third parties. Windsor wrote:

There is growing evidence that, given the current attractive return rates, Meta will roll out a new business offering compute capacity for sale to third parties.

He added that Muse Spark is already close to top-tier models in AI programming capability, “but priced at only 25% of the latter,” which would make it highly attractive to the mass market.

Chip in-house development and compute expansion boost cloud revenue potential

Wall Street Journal noted that Meta has planned to start mass production of its in-house AI chip codenamed “Iris” this September. The chip is co-designed by Broadcom, manufactured by TSMC, completes testing in just six weeks, and has already signed multi-year supply agreements with Samsung, SanDisk, and Sumitomo Electric.

In terms of compute scale, Meta plans to deploy 7 gigawatts of compute this year, and double that to 14 gigawatts in 2027.

Supporting these goals are five gigawatt-class “titan” hyperscale data center clusters Meta is building in parallel, as well as its in-house “AI-Backbone” network architecture, which allows Meta to asynchronously expand complex training tasks across geographic distances of thousands of kilometers.

In a research report on Thursday, Deutsche Bank analyst Benjamin Black said the above compute expansion plan implies Meta’s potential incremental revenue from third-party cloud services of about $24 billion, significantly higher than the previous $17 billion forecast.

He also noted that Meta’s in-house chips could open up a realistic path for the company to reduce costs and improve efficiency.

SemiAnalysis: Meta AI could surpass Google within six months

Wall Street Journal noted that SemiAnalysis believes that after a year of aggressive capital spending and architectural restructuring, MSL is expected to surpass Google in the ranking of frontier AI capabilities within the next six months.

The report states that the current duopoly landscape of Google and OpenAI would be rewritten into a three-way setup of Meta, OpenAI, and Anthropic.

SemiAnalysis’ key judgment hinges on the speed of compute expansion: Meta’s growth trajectory in AI compute scale would enable it to surpass the combined compute totals of OpenAI and Anthropic by the end of the year.

Citing an internal memo, Reuters reported that Meta’s cap on capital expenditures for AI infrastructure this year could be as high as $145 billion.

On the talent side, Meta has redeployed 3,000 engineers to internal reinforcement learning environment factories, building proprietary data that is hard for commercial data providers to replicate, and it invested $14.3 billion in Scale AI to bring in large numbers of top researchers from institutions such as OpenAI and Anthropic.

SemiAnalysis believes that evaluating MSL solely based on current benchmark test results is “seeing the trees but not the forest.” The truly crucial factor is the future momentum, not the height of the starting point today.

The report notes that if Zuckerberg maintains the current pace of capital spending, Google could be permanently pushed out of the first-tier group of global hyperscalers in AI.