Zuckerberg Just Paid More for One AI Researcher Than Most Hedge Funds Make in a Year

What the hunt for super-intelligence could mean for financial markets and automated trading

Jul 03, 2025

A few days ago I published a post about farming data using ML and the AI arms race. I told you about the $100 million signing bonuses offered to a small group of researchers at OpenAI by Mark Zuckerberg. What I did not realize is how much Zuckerberg wanted this thing. Maybe some of that judo is changing his brain because this was a fire move made by a man that was sitting at the ‘Apple AI’ table with Tim Cook just a few weeks ago. Then he did something that no one saw coming—not even ChatGPT 5 (jokes)—he poached OpenAI’s top players

The Scale AI Acquisition: More Than Meets the Eye

As it turns out, Meta's recent $14 billion investment for a 49% stake in Scale AI wasn't just about data infrastructure—it was a masterful case-study in talent acquisition. While Scale AI's data collection and labeling capabilities are valuable (both OpenAI and Google were customers), the real prize was Alexander Wang.

By opting for a minority stake (just under 50%), Meta avoided lengthy regulatory reviews that a full acquisition would trigger. The immediate fallout? Both OpenAI and Google canceled their contracts with Scale AI, but Meta had already secured what it really wanted: Wang himself, who now leads the newly formed Meta Super Intelligence Labs (MSL) as chief AI officer, co-leading with GitHub CEO Nat Friedman.

Zuckerberg has been on a talent spree, and got Scale AI's data along with one of the world's hottest CEOs, Wang, for a measly $14 billion. Add that to the $100M sign-on bonuses given to top talent in the new Super Intelligence team, and it becomes clear: Zuckerberg wants to be 'the one' and he is willing to go all-in. The only constraint is capital and he’s got more of it.

Super intelligence—AI that surpasses human cognitive abilities across all domains, not just matching them but exceeding them—is the ultimate prize in this race. While others talk about AGI (human-level AI), Zuck is shooting for the stars.

To understand the stakes, here's the progression:

Current AI → AGI (Artificial General Intelligence) → Super Intelligence

It's the difference between having an AI assistant, an AI colleague, and an AI that makes Einstein look average.

"Everyone has a plan until they get hit in the face."

-Mike Tyson

OpenAI's Crisis Response: Missionaries Beat Mercenaries

Needless to say, the impact on OpenAI has been visceral. No doubt Elon's somewhere laughing.

In an internal memo, OpenAI's Chief Scientist Mark Chen described the feeling as "if someone has broken into our home and stolen something." They feel Zucker-burglered (get it?). But at $100M a pop in Meta options, I think it's fair to say he purchased them fair and square. It is a testament to their loyalty for Chen that it took $100 million to do so. The company is now scrambling to recalibrate compensation packages, but Chen acknowledged a harsh reality: "Meta's market cap is in the trillions...OpenAI is still a private company and relative to Meta is kind of small."

Perhaps most telling was Meta's tactical timing—announcing their super intelligence team during OpenAI's scheduled week off and pressuring staff with "exploding offers" that required quick decisions. Chen warned employees against making decisions "fast and in isolation," but the damage was already being done.

The internal turmoil at OpenAI runs deeper than compensation. Chen's memo revealed strategic soul-searching: "We've been too focused on these incremental buzzy releases every few months." He urged the team to "remain focused on the real prize of finding ways to convert compute into intelligence. A lot more supercomputers are coming online later this year. This is the main quest and it's important to remember that skirmishes with Meta are the side quest."

Meta didn't just recruit—they assembled the "who's who of AI researchers" from OpenAI, Anthropic, and Google DeepMind. The roster reads as follows:

From OpenAI:

Shenzha Xiao (PhD Stanford, Research Scientist)
Hong Yu Ren (PhD Stanford, led development of O3 mini and O1 mini)
Shu Chiao Bai (co-creator of GPT-4o Voice Mode and O4 Mini, led multimodal post training)
Trapit Bonsel (pioneered reinforcement learning on chain of thought)
Huen Chang (co-creator of GPT-4o's image generation)
Shanggha Xiao (previously led synthetic data)
G Lynn, Jihu Yu (key contributors to GPT-4 and reasoning models)
Plus the entire Zurich office, including four of their best AI researchers

From Google DeepMind:

Xia Hui Yu (co-led Gemini Multimodal)
Jack Ray (reasoning for Gemini 2.5)
Pay Sun (reasoning for Gemini)
Johan Schulwick

From Anthropic:

Joel Pobar

These aren't just any researchers—they're the architects behind breakthrough reasoning models. Their expertise spans: large language models, multimodal AI, reinforcement learning, synthetic data, and reasoning systems.

Crucially, Meta's commitment to open-source development through its Llama models offers a philosophical draw that compensation alone cannot match. Many researchers are attracted to the idea of democratizing AI rather than keeping it locked behind corporate walls. As Zuckerberg has emphasized, Meta plans to make its AGI breakthroughs "open source and accessible to everyone," a stark contrast to the closed approaches of OpenAI (despite its name) and Google. For researchers who entered the field to advance human knowledge rather than corporate profits, the opportunity to work on cutting-edge AI that will be freely available to the world represents a powerful motivator beyond any signing bonus.

If Meta's AI achieves breakthrough capabilities first, they could potentially dominate not just social media, but financial markets as well. And here's the nightmare scenario—what if they don't even NEED to trade? What if just having this capability lets them reshape markets through pure information asymmetry? They'd be the house in a casino where they can see everyone's cards.

The Implications for Our Hunt

AI is taking over the world, and all I can think about is: “What would I do if I had access to this AI think tank?” If I had this team working on automated trading, we wouldn't just find alpha, we'd redefine what markets ARE:

Advanced Pattern Recognition: The multimodal expertise of researchers like Xia Hui Yu and Shu Chiao Bai could lead to AI trading systems that spot supply chain disruptions from satellite photos before earnings reports, detect emotional patterns in CEO voices during earnings calls or track private jet movements to predict M&A activity. With multimodal AI, we're not just reading the market—we're reading the entire world as one massive trading signal.
Reasoning-Based Trading Strategies: With experts in reasoning models like Jack Ray and Pay Sun, Meta's super intelligence could develop trading systems that don't just pattern match, but actually reason through complex market scenarios, understanding cause and effect in ways current systems just can’t. This isn't just faster trading—it's understanding why markets move. I’m talking about an AI that understands that a drought in Brazil affects coffee futures, which impacts Starbucks margins, which influences consumer discretionary spending, which ripples through the entire S&P 500. It's not following correlations—it's building a mental model of the entire global economy.
Synthetic Data Revolution: Shanggha Xiao's expertise in synthetic data could enable the creation of infinite market scenarios for training, allowing AI to prepare for events and market conditions that have never existed before. Train on every possible future, trade the one that happens. What if we could simulate entire alternate economic histories? Train on a world where 2008 never happened, where COVID hit in 2010. An AI that's seen every possible timeline wouldn't just predict the future; it would be impossible to surprise.
Speed and Scale: Super intelligence applied to trading could operate at speeds and scales that make current high-frequency trading look primitive. We're talking about systems that could potentially predict and respond to market movements before they fully materialize —not microseconds, but actual prescience. Imagine AI that trades on information that hasn't happened yet because it can simulate CEO decisions before they're made, predict Federal Reserve moves by modeling every board member's psychology, or front-run market crashes by sensing fear in aggregated biometric data from smartwatches worldwide.

I’m looking for a physics that treats the entire global economy as a solvable equation. Traditional concepts like "market efficiency" become relics in such a world. Every trade, every ticker, every asset class becomes part of one unified theory of value. The AI wouldn't just trade markets—it would conduct them like a symphony. And as AI systems become more sophisticated, regulators will face an unprecedented challenge. How do you regulate a trading system that may be smarter than the humans overseeing it?

But here's the trillion-dollar question: How will we know when super-intelligence has entered financial markets? More importantly, how do we know it hasn't already?

Here’s what to look for:

Impossible Patterns: Look for trading patterns that seem to anticipate events before any human could recognize them.
Market Efficiency on Steroids: Look for bid-ask spreads to approach zero, arbitrage opportunities will vanish, and price discovery will happen before news even breaks.
The Interest Rate Tell: According to one very compelling study, the arrival of super-intelligence would cause a massive spike in real interest rates. Why? Because if AGI means either explosive growth or existential risk, people will save less and spend more. Look for the bond market to be the canary in the coal mine.
Volatility Collapse or Explosion: Either markets become eerily stable as super-intelligence eliminates uncertainty, or they become wildly volatile as multiple super-intelligent systems battle for supremacy in ways we can't comprehend.
The Correlation Breakdown: Traditional correlations between assets, sectors, and global markets might suddenly stop working. A super-intelligence might find hidden relationships that rewrite the rules of portfolio theory.

Wall Street has always been 1-2 years ahead of Main Street in tech adoption. If Renaissance Technologies' Medallion Fund is consistently beating the market by 30%+, if high-frequency traders are making millions from microsecond advantages, what happens when that technological edge becomes a super-intelligence edge?

The military is often cited as being ahead of civilian tech. But in AI, finance might be the real pioneer. After all, markets are the perfect training ground for AI —clear objectives (profit), instant feedback (prices), and infinite data.

The scariest part? A true trading super-intelligence might be designed to hide its own existence. After all, the most profitable strategy would be to extract value without causing the kind of market disruption that brings regulation. It would be the perfect predator— invisible.

The Road Ahead

As Meta assembles its "all-star team," the rest of the industry must adapt. OpenAI is returning to its roots, focusing on the "main quest" of converting compute into intelligence. Google DeepMind faces its own talent retention challenges. And smaller players may find themselves increasingly squeezed out.

Speaking of OpenAI's founding principles, one can only imagine Elon Musk's reaction to this drama. The Tesla CEO, who co-founded OpenAI in 2015 with the explicit mission of keeping AI open and safe, must be watching with a mix of vindication and irony as his former organization hemorrhages talent while abandoning its open-source principles. Musk, who left OpenAI's board in 2018 citing conflicts of interest, has been increasingly vocal about it becoming a "closed-source, maximum-profit company"—exactly the opposite of its founding charter.

Now, as Meta raids OpenAI's ranks while championing the very open-source philosophy OpenAI abandoned, Musk might be having the last laugh. His own AI venture, xAI, launched with the goal of "understanding the universe," could potentially benefit from this chaos. While there's no confirmed partnership between Meta and xAI, the philosophical alignment around open-source AI development creates possibilities. After all, the enemy of my enemy…

The broader AI talent war extends beyond just Meta and OpenAI. Former OpenAI CTO Mira Murati's new startup "Thinking Machines" has already poached 20 OpenAI researchers and raised funding at a $10 billion valuation.

The AI talent war of 2025 may be remembered as the moment when the race for super-intelligence transformed from a research endeavor into an all-out corporate battle, with consequences that will shape not just the tech industry, but the entire global economy.

The next six months will be critical. Watch for:

More talent movements (contracts expire and bonuses vest)
The first outputs from Meta's new super intelligence team (it may take some time for the group to become a “team”)
Regulatory responses
Early applications of advanced AI in trading systems

I'm not just developing strategies—I'm on a quest for the holy grail of automated trading. For links to all strategies click here.

Contact: Celan @ AutomatedTradingStrategies@protonmail.com