Jun 10, 2025 5 min read Explainer

What's Behind Meta's Big Bet on Superintelligence?

Sign up for ARPU: Stay ahead of the curve on tech business trends.

Mark Zuckerberg, frustrated with the performance of Meta’s latest AI models, is personally recruiting a new secretive team with a singular, audacious goal: to build artificial general intelligence, or AGI. His plan to assemble a "superintelligence group" and sink billions into partner Scale AI is the latest and most explicit sign that the AI race is entering a new, far more ambitious phase. The battle is no longer just about building a better chatbot; it's about creating a machine that can reason and perform on par with, or better than, a human.

What is AGI and why is it the new focus?

Artificial General Intelligence refers to a theoretical form of AI where a machine possesses the ability to understand, learn, and apply its intelligence to solve any problem a human being can. While today's Large Language Models (LLMs) are powerful, they are a form of narrow AI, trained on specific tasks. AGI is the industry's holy grail.

The shift in focus is driven by both ambition and necessity. As the performance of foundational models from Google, OpenAI, DeepSeek, and others begins to converge, some analysts argue they are becoming "commodities." Here's tech analyst Benedict Evans on the dynamic:

OpenAI and all the other foundation model labs have no moat or defensibility except access to capital, they don’t have product-market fit outside of coding and marketing, and they don’t really have products either, just text boxes – and APIs for other people to build products.

With the basic technology becoming more accessible, the new competitive frontier is the creation of a true, reasoning intelligence—a breakthrough that would represent an unassailable technological lead and redefine every industry.

Who are the main players in this race?

The pursuit of AGI is a high-stakes game reserved for a handful of players with immense resources.

OpenAI, backed by Microsoft, has AGI as its stated founding mission. Its CEO, Sam Altman, is spearheading multi-hundred-billion-dollar infrastructure projects like "Stargate" specifically to secure the massive compute power he believes is necessary.
Google, with its DeepMind lab, has been a research powerhouse in this area for over a decade. Its latest Gemini models are pushing the boundaries of reasoning and multimodal understanding.
Meta's new "superintelligence" group is Zuckerberg's direct entry into the fray, backed by the formidable cash flow of its advertising empire. He is pitching recruits on the idea that Meta can out-spend its rivals without needing constant fundraising.
Other players include Elon Musk's xAI, which has reportedly secured tens of billions in funding, and Anthropic, which is focused on building AGI safely.

What does it take to compete?

The price of admission to the AGI race is astronomical, requiring an unprecedented combination of capital, talent, and raw power.

First is the sheer cost. A recent McKinsey report projects that the global investment needed for data center infrastructure to support AI demand could reach nearly $7 trillion by 2030. Zuckerberg’s pitch to potential hires—that Meta has the cash flow to fund multi-gigawatt data centers—underscores that only a few entities can afford to play at this level.

Second is talent. The race has ignited a fierce "talent war" for a few dozen to a thousand "superstar" researchers who can make foundational breakthroughs. Companies are offering compensation packages exceeding $10 million to $20 million a year and retention bonuses in the millions to lure or keep these key individuals. CEOs like Zuckerberg are now acting as chief recruiters, personally courting talent in a way typically reserved for star athletes.

Finally, there is the insatiable need for compute power. This drives the massive demand for NVIDIA's GPUs and has created a new set of challenges around physical infrastructure. The power required for a single AI data center can be equivalent to that of a small city, straining local electricity grids and water supplies. This has led to novel solutions, like the Stargate project in Abilene, Texas, building its own on-site natural gas plant to get around grid limitations.

What are the risks and challenges?

Despite the flood of investment, the path to AGI is far from certain. There are signs that the "scaling laws"—the idea that simply adding more data and compute power will inevitably lead to smarter AI—are yielding diminishing returns. A recent report noted that top labs like OpenAI, Google, and Anthropic are all hitting stumbling blocks and seeing performance gains plateau with their newest models, struggling to find enough high-quality training data to justify the immense cost of training.

Furthermore, the very nature of these powerful models creates profound safety risks. Anthropic CEO Dario Amodei has publicly stated that "we do not understand how our own AI creations work," a reality that makes it difficult to prevent harmful or unintended behaviors. As these systems become more autonomous in the pursuit of AGI, the challenge of ensuring they remain aligned with human values becomes one of the most critical, and perhaps unsolved, problems in technology today.