Meta tried poaching OpenAI staff with $100M signing bonuses

Meta CEO Mark Zuckerberg is reportedly writing nine-figure checks in a desperate bid to win the AI talent war.

Speaking on his brother's podcast Uncapped, OpenAI CEO Sam Altman revealed that Meta has been trying to aggressively poach OpenAI researchers with signing bonuses as high as $100 million, alongside massive annual compensation packages.

Altman sounded unfazed by the offensive, claiming that despite Meta's best efforts to recruit a lot of people, none of OpenAI's best talent had taken the bait.

Zuckerberg's talent offensive

Altman did not mince words about Meta's AI standing. He noted that Meta views OpenAI as its biggest rival, but added that their current AI efforts have simply not worked as well as they hoped.

That frustration appears real inside Meta. The Wall Street Journal recently reported that Meta delayed the release of its flagship AI model once again over performance concerns, prompting Zuckerberg to get personally involved in recruitment.

Meta has already shown it will spend whatever it takes. Last week, Meta acquired a 49% stake in Scale AI for $14.3 billion, bringing founder Alexandr Wang in-house to lead a new superintelligence lab. Bloomberg also reported that Meta snatched Jack Rae, a top principal researcher from Google DeepMind.

Altman argued that throwing absurd, upfront guaranteed money at people is a terrible way to build a winning culture. In his view, trying to copy OpenAI means Meta will always be running toward where OpenAI used to be, rather than learning how to innovate.

Yet massive paydays are becoming standard operating procedure. Just last month, OpenAI completed a $6.4 billion all-equity deal to buy Jony Ive's startup io, bringing the former Apple design chief into its ranks.

Why it matters

We think this $100 million signing bonus rumor exposes just how desperate the frontier model race has become.

Compute is no longer the sole bottleneck. The real scarcity is the tiny cohort of researchers who actually know how to push pre-training and reasoning models past their current plateaus. Zuckerberg's willingness to hand out lottery-sized bonuses shows that Meta feels dangerously behind with its next-generation models and is trying to buy an instant fix.

However, building a frontier research lab like a mercenary squad rarely works. High-performance AI teams rely on shared intuition, long-term alignment, and tight-knit collaboration. Massive guaranteed compensation packages can alienate existing engineers and create cultural friction rather than breakthrough science.

For the broader industry, Meta's spending spree is still a net positive. Zuckerberg's obsession with closing the gap ensures that billions of dollars will continue pouring into open-source models, data pipelines, and foundational research. Even if Meta fails to clone OpenAI's culture, developers and startups will likely benefit from the sheer volume of open tooling and models spilling out of Menlo Park.