The AI world was recently rocked by another major shakeup. On June 29, The Information reported that Meta (Facebook’s parent company) successfully poached four heavyweight researchers from OpenAI. These weren’t just junior staff—they were key contributors to OpenAI’s most important projects, including the development of top-tier models like o3 and the GPT-4 series. Now, they’re set to join Meta’s newly formed “Superintelligence” team.

Make no mistake—this isn’t just regular talent movement. It’s an all-out poaching war between AI giants.

The Bigger Game: Open Source vs. Closed Source

To understand why Meta and OpenAI are fighting so hard, you have to look at how fundamentally different their approaches to AI are.

OpenAI ignited the generative AI boom and built its brand on making money through a “black box” model. In other words, they keep their most powerful models—like GPT-4o—entirely closed-source. Instead of sharing them, they offer limited API access, charging by usage. It’s a smart move: it ensures their tech stays ahead of the pack and lets them profit from developers around the world. The advantage is clear—technical superiority and a lucrative, hard-to-replace business model.

But it comes with a huge risk: if any rival catches up or finds a workaround, the whole closed ecosystem could quickly collapse.

Meta has taken a fundamentally different path—embracing full transparency by open-sourcing their models and actively engaging the global developer community to expand their reach. Starting with the Llama series, Meta made its powerful models freely available to the world—academics, companies, developers alike. Zuckerberg’s strategy is calculated: instead of competing with OpenAI or Anthropic on chatbot products, Meta is betting on commoditizing the model layer itself. By making their foundation models free and widely accessible, they hope to weaken the high-priced, API-based dominance of closed companies.

In short, Meta’s game is “open-source empowerment”: build the world’s largest developer ecosystem by offering free and powerful tools, and shift the battlefield to infrastructure and adoption. This is classic asymmetric warfare—not trying to outprofit OpenAI, but to redefine the rules of the game.

Zuckerberg’s “Too Good to Refuse” Offer

To pull this off, Meta needs to solve a core issue: who’s going to build these models?

Developing large-scale AI models isn’t just about buying servers—it requires top-tier global talent. That’s why Meta has gone all-in on the talent war recently.

Reportedly, Zuckerberg has been personally emailing major AI researchers almost daily, trying to lure them to Meta with jaw-dropping offers. According to whispers in the industry, some signing bonuses have reportedly reached staggering amounts near $100 million. Ultimately, in this competition, the true prize isn’t merely the hardware like GPUs—but the expert minds capable of maximizing their potential.

OpenAI’s CEO, Sam Altman, previously minimized the risk posed by Meta’s recruitment efforts. He once claimed that although Meta was trying hard to poach their talent, “our best people are saying no.” Well, these four high-profile departures suggest otherwise—and show just how determined Zuckerberg is.

The Logic Behind This Talent War

Meta isn’t just winning over researchers with money—compute power is a major draw, too.

Conducting advanced AI research requires essential resources, much like an artist needs a canvas and colors. For AI scientists, this critical resource is computing power. Meta is investing heavily in this area, with Zuckerberg announcing plans to have hundreds of thousands of NVIDIA H100 GPUs operational by the end of 2024. For researchers who often feel constrained by limited resources, the promise of “unlimited top-tier compute” is hard to pass up.

On top of that, Meta offers highly competitive long-term equity packages. This means researchers don’t just get high salaries—they also get a shot at massive upside if Meta’s AI ecosystem takes off. For many elite AI scientists, this isn’t just about “working a job,” but about shaping the future and leaving a legacy.

The Real Battlefield of AI: People

This four-person poaching coup is just the tip of the iceberg. The AI industry has reached a heightened stage of intense rivalry, driven by the combined forces of technological innovation, financial investment, and top-tier expertise. GPUs can be bought, but the brains that design the best models and the experience to fine-tune them are much harder to replicate.

OpenAI’s closed model relies on maintaining a constant tech lead. Meta’s open-source push depends on continually attracting the world’s best minds to maintain model-building supremacy. In both cases, top AI talent is the lifeblood.

At the end of the day, this race isn’t really about how many GPUs you have or how big your model is—it’s about who you can get to build them. The future of AI will be shaped by human hands. Whoever can bring the brightest minds onboard will shape the direction of this revolution.

If we had to sum up the current Meta vs. OpenAI battle in one line, it would be:

“The moat of closed-source vs. the battering ram of open-source.”

And Zuckerberg is clearly set on recruiting the best “ram carriers” to break through.