Meta accelerates in the race for artificial intelligence (AI) to avoid falling behind OpenAI and Google. After the investment in Scale AI, Zuckerberg aims for the acquisition of elite startups like Perplexity and Thinking Machines Lab, but encounters several obstacles.
Artificial intelligence (AI) at the center of Meta’s strategy
In recent months, Meta has intensified its efforts to establish itself as one of the key players in the generative artificial intelligence landscape.
After a promising start with the LLaMA models, the Menlo Park company seems to be struggling to keep up with the industry leaders, primarily OpenAI and Google DeepMind.
To fill this gap, Mark Zuckerberg has embarked on a true “hunt for talent” and emerging startups. The goal?
Strengthen the company’s AI division and give new impetus to the development of advanced models, capable of competing with ChatGPT, Gemini, and the other giants of AI.
According to authoritative sources such as Bloomberg and The Verge, Meta is said to have initiated preliminary contacts with some of the most promising startups in the sector. Among these:
Perplexity.ai, the AI-based search engine that is gaining popularity thanks to a fast and efficient conversational interface.
Thinking Machines Lab, laboratory specialized in the development of large-scale models and in the design of distributed intelligent systems.
Safe Superintelligence (SSI), the new venture by former OpenAI executives, Ilya Sutskever and Mira Murati, focused on the development of safe and controllable artificial intelligence.
However, the negotiations do not seem to have yielded the hoped-for results. The startups have indeed declined Meta’s offers due to differences in strategic vision, company structure, and, above all, governance.
Zuckerberg’s Plan B: Hiring the Best Talents
In the face of rejection by the startups, Zuckerberg did not give up. According to reports, Meta allegedly attempted to recruit directly the CEOs and the main developers of some of these companies. In particular, the names have emerged of:
Daniel Gross, CEO of Safe Superintelligence and well-known investor in the tech sector;
Nat Friedman, former CEO of GitHub and a central figure in the open source ecosystem.
The offers made to potential candidates would have been extremely generous, in some cases even exceeding 100 million dollars, but so far no negotiation has been successful.
The reason? Probably a mix between project independence, personal ambition, and little trust in Meta’s AI roadmap.
The pressing of Meta on startups and AI talents reflects a certain state of alert within the company.
While on one hand the presence of AI in the group’s social networks (such as Instagram, Facebook, and WhatsApp) has achieved good penetration among users – over 1 billion reached by May, on the other hand, the performance of the internal language models is still disappointing.
A clear example is the Behemoth project, initially scheduled for the first half of 2025 and now postponed to the end of the year due to issues related to the quality of responses and system stability.
In the same way, LLaMA 4 did not achieve the expected results, prompting Zuckerberg himself to revise the entire organization of the AI team and create a new “Superintelligence” unit under his direction.
Despite its efforts, Meta’s position in the global AI landscape is starting to resemble more and more that of Apple or xAI, the company founded by Elon Musk.
Apple also bets on Perplexity: a two-horse race?
Great resources, an immense user base, and solid infrastructure… but inability to truly compete on frontier models.
OpenAI, with ChatGPT-4o, and Google, with Gemini 1.5, continue to dominate the scene thanks to a combination of technical innovation, rapid release, and active community.
In this context, Meta risks becoming a secondary player, especially if it fails to bridge the quality gap between its models and those of the competition.
As a demonstration of the strategic value of certain AI startups, even Apple is said to have expressed interest in Perplexity, likely to enhance Siri and AI-based voice search and assistance services.
The move highlights how AI startups are today the most contested asset among big tech, and how difficult it is to emerge without high-level proprietary technology. The race for AI is more heated than ever, and Meta finds itself at a strategic crossroads.
After the rejections from the main emerging startups, Zuckerberg will have to decide whether to continue with targeted acquisitions, invest even more in internal development, or adopt a new collaborative approach, perhaps open source, to remain competitive.
What is certain is that time is running out, and OpenAI and Google wait for no one.