SaIn August 2025, Mark Zuckerberg sent Andrew Tulloch a personal message on LinkedIn.
Tulloch was a co-founder at Thinking Machines Lab , the AI startup built by former OpenAI CTO Mira Murati. Zuckerberg had already tried to buy the whole company for $1 billion. Murati refused.
So Zuckerberg went after the engineers individually. The offer to Tulloch: $1.5 billion in total compensation over six years.
Tulloch said no.
Two months later, in October 2025, he said yes. The deal is now the most expensive individual hire in the history of the technology industry.
And almost nobody has explained what it means for the rest of us.
The numbers that broke the labor market
Tulloch's $1.5 billion package is the headline. The numbers behind it are more interesting.
Implied annual comp
$250M
Per year, one session
Multiplier
1250X
Tulloch vs typical engineer
For context. A typical engineer at OpenAI makes roughly $200,000 per year. Jamie Dimon, CEO of JPMorgan Chase, made $39 million in 2024. Tim Cook made $74.6 million the same year. Tulloch's $250 million annual package puts him above the highest-paid CEO of any S&P 500 company.
And he is not the CEO. He is a researcher.
Annual compensation comparison ($M per year)
The market mechanism
Meta is not paying $250M because Tulloch produces $250M of work per year. They are paying $250M because keeping him at Thinking Machines costs them more, in slower research, in falling behind OpenAI and Anthropic, in losing the AGI race. Talent has stopped being priced on output. It is now priced on counter-factual loss.
Presented by Atoms.dev
Atoms.dev gives you a full team of AI employees, PM, engineer, SEO specialist, data analyst that builds your product, deploys it, and finds your first customers. 165K GitHub stars. #1 Product Hunt of the Week. Trusted by Amazon, Microsoft, and NVIDIA.
How we got here in eight moves
Jun 2025: Meta invests $14.3B in Scale AI for a 49% stake. Alexandr Wang joins Meta to lead the new Superintelligence Labs. The playbook is set, buy the company, take the founder.
Jul 2025: Meta offers $100M signing bonuses to senior OpenAI researchers. Sam Altman calls the strategy "distasteful". Eight OpenAI researchers accept by end of July.
Aug 2025: Meta tries to acquire Thinking Machines Lab for $1B. Mira Murati refuses. Zuckerberg pivots to personal outreach.
Aug 2025: Tulloch offered $1.5B over six years. He turns it down. Meta calls the figure "inaccurate and ridiculous."
Oct 2025: Tulloch joins Meta. Multiple outlets confirm the original $1.5B framework remained the basis.
Mar 2026: Five of Thinking Machines' nine founding members have joined Meta. The startup is functionally dismantled.
Apr 2026: Meta launches Muse Spark, its first major AI model since the Wang acquisition. Reception is mixed. Llama 5 is delayed.
Jun 2026: Compensation ceiling now $250M/year for top AI researchers. Anthropic, Google, OpenAI all forced to match. Mid-tier AI talent compensation rises 60% year-over-year.
The compensation inflation has cascaded down. When the top of the market re-prices, the rest follows.
AI researcher compensation by tier, 2023 vs 2026 (log scale)
When the ceiling moves from $2M to $50M for top talent, every salary below it has to move too. Mid-tier AI researchers now cost what senior engineers cost two years ago. Senior engineers now cost what executives cost.
What this actually means for your business
You are not competing with Meta for Andrew Tulloch. But you are competing in the same labor market and the rules have changed everywhere.
Your AI hire just got 60% more expensive. Mid-tier AI researcher comp rose roughly 60% year-over-year as the ceiling moved. If your hiring plan budgets 2024-level compensation for 2026-level talent, you will lose every candidate to a better offer. We covered the other end of this market last week, the data shows the middle is hollowing fastest.
Buying talent is now cheaper than poaching it. Meta paid $14.3B for Scale AI to get Wang. That works out to roughly the same per-head as poaching one $1.5B engineer, except you also get a company, an infrastructure layer, and a customer base. Acqui-hires are back, structured differently than the 2010s but at much larger scale.
The talent that matters is not the talent that's expensive. The Stanford payroll data we covered last week showed entry-level employment in AI-exposed jobs dropped 13%. This week's Meta story shows top-of-market compensation tripled. Both are true. The middle is hollowing out. The real question for your business: which tier of talent do you actually depend on, and is the right move to outbid Meta, or bypass that market entirely?
Related reading from our archive
AI Ate the Entry-Level Job : the Stanford data on what AI is doing to the bottom of the labor market
Karpathy's Real Job at Anthropic : why the most important AI hire of the year is not what people think
OpenAI Blinked First : how Anthropic became the most valuable private AI company
The Startup With No Employees : how solo founders are hitting $5M ARR without hiring anyone
Presented by Semrush
Struggling to understand why your competitors rank higher or where your next customers will come from?
Use Semrush - the powerful platform that brings clarity to your entire digital marketing strategy.
Keyword & topic research to target the right audience
Comprehensive site audits to identify and fix SEO issues
Backlink analysis with valuable link-building opportunities
Content optimization tools to boost rankings and engagement
Advertising & PPC insights to maximize campaign performance
🔮 The Bottom Line
Mark Zuckerberg offered one person $1.5 billion to switch jobs.
That number is not really about Andrew Tulloch. It is about what the AI labs believe they are racing toward, and what they think losing the race would cost them.
If the most powerful companies in the world are willing to pay $250 million per year for a single researcher, they are not playing the same game everyone else is. They are not optimising for productivity. They are not optimising for ROI on that hire.
They are buying time on a timeline they believe is closing fast. Everyone else is hiring at salaries that assume there is still time. The gap between those two assumptions is the most important business question of this decade.
For the readers
Reply to this email with the engineer you would pay $1.5B for if you had it. Most interesting answer gets featured in next week's issue. Or browse the full hiPreneurs archive for more issues like this one.
Sources used in this issue: WSJ on Tulloch's package · TechCrunch on his Meta move · Bloomberg on the Wang acquisition · CNBC on $100M bonuses · TNW on the Thinking Machines raid · Levels.fyi AI compensation data. Meta has disputed specific compensation figures as "contingent on stock performance."
📧 Forward this to 3 entrepreneur friends who need to see this opportunity
