Markets News
AI & Tech1 min read

Zuckerberg’s AI Bet Took Shape After a 44-Minute Call

Makkler Newsroom
October 17, 2026 · Based on reporting by Fortune
Photo: Anthony Quintano from Westminster, United States / Wikimedia Commons · CC BY 2.0

A call with Alexandr Wang and a memo two days later helped set up Meta’s costly AI hiring push.

Key takeaways

  • Meta expects 2026 capital spending of $130 billion to $145 billion.
  • Wang sent Zuckerberg a strategy memo two days after their 44-minute call.
  • Meta paid $14.3 billion for a 49% stake in Scale AI.
  • Wang said the lab’s core research team numbers closer to 100 people.

Meta expects to spend $130 billion to $145 billion on capital expenditures in 2026 as CEO Mark Zuckerberg backs an ambitious AI push led by Alexandr Wang. The forecast is nearly double the $72.2 billion Meta spent last year, according to Fortune.

A call and a memo

Wang said the partnership began after Meta’s Llama 4 models received a lukewarm reception. In April 2025, Zuckerberg sought Wang’s advice, and the two spoke for about 44 minutes, according to Zuckerberg’s account.

Wang sent a long memo two days later outlining what he would do in Zuckerberg’s position. Zuckerberg said Meta was unhappy with Llama 4’s direction and was willing to rebuild its team.

Zuckerberg said he had been impressed by Wang for years and wanted someone who could recruit leading researchers. Wang had cofounded Scale AI, but had not trained a frontier model, Fortune noted.

Building the team

Meta paid $14.3 billion for a 49% stake in Scale AI to bring Wang aboard. Zuckerberg described the recruitment drive as “supernova”; Fortune reported that Meta also dangled compensation offers of $100 million to OpenAI researchers.

Wang said the core research team at Meta Superintelligence Labs has fewer than 200 people, “closer to 100,” with larger infrastructure and product teams supporting it. He described the researchers’ culture as one of “insecure overachievers.”

Early model results

The lab’s first model, Muse Spark, arrived in April, followed by updates. Meta introduced Muse Wonder on Monday as its first model from the lab that the company calls frontier-class.

Wang said Muse Wonder is state-of-the-art for personal and professional agent tasks, based on the lab’s measurements. Artificial Analysis’s Intelligence Index placed its flagship version roughly level with leading OpenAI and Anthropic models, but a few points behind the leader, according to AI Weekly.

Topics
METAAlexandr WangScale AIMeta Superintelligence LabsMuse
Further reading

This article was produced with the help of AI technology. Based on reporting by the outlets listed above.

Comments (0)

Log in to join the discussion.Log in

No comments yet - be the first to weigh in.