Mark Zuckerberg was furious after Mira Murati refused his Rs 8,500 cr offer; Here's what he did next

A few months ago, Mark Zuckerberg tried to buy Mira Murati’s startup, Thinking Machines Lab. She declined. Despite a $1 billion offer, she wouldn’t part with it. Murati had just left OpenAI, where she had spent six years and risen to Chief Technology Officer. She wasn’t about to hand over her new venture to Meta.

Zuckerberg didn’t take the rejection quietly. According to The Wall Street Journal, he followed up with what they described as a "full-scale raid". Over the following weeks, he and Meta’s senior team approached more than a dozen of Murati’s 50 employees. One person he focused on was Andrew Tulloch, a highly regarded AI researcher and co-founder of the startup.

The $1.5 billion man

To tempt Tulloch, Meta reportedly offered him a compensation package worth up to $1.5 billion over six years, depending on bonuses and Meta’s stock performance. A figure so large it raised eyebrows even in Silicon Valley. He still turned it down.


Meta spokesperson Andy Stone denied the report’s accuracy, calling it “inaccurate and ridiculous,” and insisted the company wasn’t trying to acquire Thinking Machines. He claimed the compensation would have been entirely dependent on stock price performance.

But whether or not the number is exact, it tells you something important: Meta is prepared to spend huge sums to pull the best minds into its orbit. And that effort isn’t going unnoticed.

Not just a paycheque

Andrew Tulloch is not an easy man to poach. Born and educated in Australia, he studied mathematics at the University of Sydney, earning a University Medal and graduating top of his class. He went on to Cambridge for a Master’s in statistics and completed his PhD at UC Berkeley.

He spent over a decade at Meta (then Facebook), where he helped develop PyTorch, one of the most widely used tools in machine learning today. His influence on Meta’s AI infrastructure was massive. But loyalty, it turns out, doesn’t only go one way.

OpenAI’s Greg Brockman tried to recruit Tulloch back in 2016. At the time, Brockman told Elon Musk via email that Tulloch was earning $800,000 at Facebook and was hesitant to take a steep pay cut to join OpenAI’s leaner operation. Tulloch didn’t join then.

He waited seven years, and finally moved to OpenAI in 2023, by which time ChatGPT had already exploded onto the global stage. There, he helped with GPT-4’s pretraining and reasoning models. In 2025, he co-founded Thinking Machines with Murati.

Zuckerberg’s team wasn’t the only one reaching out. Alexandr Wang, now leading Meta’s new superintelligence lab, also tried to woo Tulloch back. Neither succeeded.

What makes people stay

There’s a pattern here. Meta has reached out to more than 100 OpenAI employees. At least 10 have joined. But not from the circles that matter most. Most of the top researchers, especially those closest to the mission of building artificial general intelligence, have stayed put.

Why? Because for many of them, this isn’t just a job. It’s a calling.

At OpenAI and its offshoots, like Dario Amodei’s Anthropic and now Murati’s Thinking Machines, the culture is unusually tight-knit. Many early employees shared a belief in “effective altruism,” a social movement focused on doing measurable good, and a shared concern that AI, if not handled carefully, could cause catastrophic harm.

Some lived together in shared houses in San Francisco. They debated ethics and long-term risks. They weren’t there for advertising revenue or status. They were there to steer the future.

That’s hard to buy, no matter how big your offer is.

Inside the Thinking Machines lab walls

Mira Murati’s Thinking Machines is still operating in stealth. Even its investors, who’ve poured $2 billion into the company, don’t fully know what’s being built.

What we do know: the goal is to make “AI systems more widely understood, customisable and generally capable.” Murati herself said the team is building “multimodal AI that works with how you naturally interact with the world.” A product announcement is expected soon.

Murati has recreated the low-ego, high-trust culture she once nurtured at OpenAI. The titles are flat. Even senior engineers are listed as “Member of Technical Staff,” a nod to the egalitarian approach inspired by Bell Labs. This kind of atmosphere breeds strong loyalty.

When she founded the company in February, more than 20 OpenAI colleagues joined her. One was John Schulman, a key figure behind ChatGPT. He had briefly left for Anthropic, only to return to work with Murati.

The Superintelligence arms race

Meta has been playing catch-up. The company recently launched a superintelligence team led by Shengjia Zhao, an ex-OpenAI researcher. It’s hired at least two employees from Anthropic, Joel Pobar and Anton Bakhtin, both of whom had previously worked at Meta.

But there are limits to how many people Meta can lure back.

Anthropic’s seven co-founders, including CEO Dario Amodei, are still at the company. Most of them met over a decade ago through the effective altruism community. They’re still there. Still committed. Still cautious about where their work ends up.

Even Ilya Sutskever, one of OpenAI’s original co-founders, has built a startup, Safe Superintelligence (SSI), in a way that makes it nearly impossible to poach from. He’s deliberately hiring unknowns, mentoring them quietly, and asking employees to leave the company off their LinkedIn profiles.

Meta tried to buy SSI earlier this year. Sutskever said no.

This moment in tech isn’t just about building better AI models. It’s about who builds them, why, and for whom. You can’t just throw money at that and expect it to stick.

Zuckerberg can offer more zeroes than almost anyone. But what he’s finding is that belief, loyalty and culture often outweigh cash. You can’t just raid your way to the top.

Some engineers may jump for a better offer. Others don’t. Tulloch didn’t. Murati didn’t. Their teams stayed. That says more than any dollar figure ever could.