OpenAI is the undisputed poster child of the AI revolution, the company that forced the world to pay attention with the launch of ChatGPT. But behind the scenes, a desperate and wildly expensive battle is raging, and the cost of keeping the company’s geniuses in-house is becoming astronomical.

According to a recent report from The Information, OpenAI revealed to investors that its stock-based compensation for employees surged more than fivefold last year to an astonishing $4.4 billion. That figure isn’t just large; it’s more than the company’s entire revenue for the year, accounting for a staggering 119% of its $3.7 billion in total revenue.

This is an unheard-of figure, even for Silicon Valley. For comparison, Google’s stock compensation was just 16% of its revenue the year before its IPO. For Facebook, it was 6%.

So what’s going on? In short, OpenAI is fighting for its life in an unprecedented talent war, and its chief rival, Meta, is on the offensive. Mark Zuckerberg has been personally courting top AI researchers with massive compensation packages, successfully poaching several key minds from OpenAI’s core teams. This has reportedly prompted a crisis at OpenAI, forcing it to “recalibrate compensation” and promise even more rewarding pay packages to prevent a catastrophic brain drain.

While stock-based compensation doesn’t immediately burn through a company’s cash reserves, it creates a major risk by diluting the value of shares held by investors. Every billion dollars in stock handed to employees means the slices of the pie owned by major backers like Microsoft and other venture capital firms get smaller.

OpenAI is trying to sell this strategy as a long-term vision. The company projects that this massive expense will fall to 45% of revenue this year, and below 10% by 2030. Furthermore, OpenAI has reportedly discussed a future plan where its employees would collectively own roughly one-third of the restructured company, with Microsoft also owning another third. The goal is to turn employees into deeply invested partners who have a massive incentive to stay and build.

But the “Meta effect” is throwing a wrench in those neat projections. The aggressive poaching and the ensuing pay bumps mean OpenAI’s costs are likely to remain sky-high.

The Stakes for OpenAI

This high-stakes financial strategy puts OpenAI in a precarious position. The company is already spending billions of dollars a year as it spends heavily on the computing power needed to run its models. Adding billions more in stock compensation puts immense pressure on the company to dramatically increase revenue and find a path to profitability before its investors get spooked.

While Microsoft seems locked in for the long haul, other investors may grow weary of having their ownership diluted so heavily. It forces a countdown timer on the company to deliver a massive financial return to justify the cost.

OpenAI was founded with a mission to build artificial general intelligence (AGI) that “benefits all of humanity.” This costly talent war, fueled by capitalist competition, puts immense pressure on that founding ideal. It becomes harder to prioritize safety and ethics when you’re burning billions to keep your top minds from joining the competition.

Ultimately, OpenAI is betting these billions to ensure it has the best talent to win the race to create the world’s first true superintelligence. If they succeed, the financial cost will seem trivial. If they fail, or if a competitor gets there first, they will have spent themselves into a hole for nothing.

OpenAI did not immediately respond to a request for comment.