Instacart co founder Apoorva Mehta launched Abundance, a hedge fund that operates with only 10 employees supported by thousands of autonomous AI agents handling everything from trade idea generation to execution. The $100M seed backed fund, inspired by OpenAI's o3 model's…
Apoorva Mehta has a theory about what a hedge fund looks like when you rebuild it from scratch around large language models. The theory involves ten people.
Mehta, the billionaire co-founder of Instacart, quietly launched Abundance last year. The firm raised $100 million in seed capital and currently trades its own money, though it plans to open to outside investors eventually. The team is ten people: quantitative researchers, engineers, AI specialists. The investment staff is considerably larger, if harder to count.
"We employ thousands of bots," Mehta said in an interview. They search the internet for trade ideas, conduct research, select stocks for long and short positions, determine bet sizes, and execute trades. Some strategies already run autonomously. Others still have a degree of human involvement, for now.
The organizational chart is the story. It is not a product claim. Nobody is arguing that better software makes better investment decisions in the short term. The argument is structural: ten people with thousands of specialized agents can do what used to require a hundred. If that math holds at scale, the fee model that funds the modern hedge fund industry is under pressure.
Abundance was inspired by OpenAI's o3 model, which demonstrated a more advanced ability to reason through complex tasks. Mehta saw it as evidence that generative AI could handle more consequential decisions, including capital allocation. "Even for the exceptional investor, the process is locked inside their mind. AI changes that entirely."
The counterargument is not hard to find. Ken Griffin of Citadel argued late last year that generative AI is not yet meaningfully affecting the hedge fund industry, and is not helping funds beat the market in any measurable way. Two Sigma's chief AI innovation officer has framed the technology differently: not as replacement but as operating system, making humans who use it well more valuable, not less.
Mehta declined to specify which benchmarks his fund has outperformed, citing competitive reasons. The fund currently trades its own capital. The question that matters — whether this model works with outside money, in a down market, under the scrutiny of institutional allocators — is unanswered.
There are things Abundance has not disclosed. The architecture is not publicly described in technical detail. No named limited partners have come forward. The "thousands of bots" claim has not been independently verified. It is a real number or it is marketing language for a conventional automation stack. The difference matters enormously. If Abundance is a next-generation quant fund with a novel agentic architecture, it is a data point in a real shift. If it is a well-funded quant shop with good PR, the story is thinner than the headline suggests.
Hedge funds have pursued automation for decades. D.E. Shaw built its reputation on systematic, computer-driven trading. Renaissance Technologies is still the reference point for mathematical investing at scale. What changes with large language models is not the ambition but the granularity of tasks that can be delegated. A quant fund has always needed humans to encode the rules. An agentic system can, in principle, let the rules emerge from the data and revise themselves. That is a different kind of firm, or it is a pitch for one.
The ten-person number is what makes this worth watching. It is specific and quotable and it answers a question every builder and investor in the AI-native space is asking: what does your team actually look like? Mehta is not the first person to say that fewer humans plus more software equals a different kind of company. He is one of the first to put a concrete number on it and a $100 million seed behind the claim.
Whether it holds is a separate question. The fund is live. The architecture is not audited. The performance data is selective. What Abundance has done is state a theory of the firm clearly enough that the next twelve to twenty-four months will either support or refute it. For an industry that charges fees based on the assumption that human judgment is scarce and expensive, that is not a comfortable position to be in.