Instacart co-founder Apoorva Mehta on the DoorDash deal and his new AI-native investment firm Abundance

Apoorva Mehta on the DoorDash deal and Abundance

Apoorva Mehta built Instacart in what he describes as the lowest-margin category in retail, then added complex last-mile logistics on top. VCs walked out of pitches. Jeff Bezos went on stage to call the model unlikely, then bought Whole Foods two years later. When Amazon, Target, and others shut down during COVID, Instacart stayed online and doubled and tripled its growth. The company's most recent quarter, under CEO Chris Rogers, was its best as a public company. The DoorDash deal, which values Instacart at $17 billion, is the outcome of fourteen years of staying alive when the conventional wisdom said not to bother.

“Instacart is a business in the lowest margin category of retail... and then you're like, let's add really complex last mile logistics to that already low margin category... I wanted to be... to have the potential to be a trillion dollar company... with Abundance, I believe that I found that. But I guess like, probably, what's even the coolest part is how fun it is.”

Agents and the grocery problem

Mehta is careful not to dismiss AI agents, but he's honest about the ceiling. Instacart already saved consumers hours every week — the original 10x improvement. Agents can save users the remaining taps, which is a much smaller delta. The more interesting agent opportunity, in his view, is proactive and continuous: an agent that understands a family's dietary needs, anticipates what they want to eat, and handles ordering on an ongoing basis. That's different from sending a bot to press the order button. Instacart, he argues, is effectively the original agentic company — it already sends a human agent to the grocery store on your behalf. The interface layer can improve; the hard infrastructure problem underneath is already solved.

Abundance

Mehta's new company, Abundance, applies long-running AI agents to capital allocation. The distinction he draws from prior algorithmic trading is the shift from statistical arbitrage — pattern recognition and repetition — to something closer to fundamental research. Reasoning models can now evaluate moats, durability of growth, and qualitative competitive dynamics in ways that stat arb never could. Renaissance-style pattern matching was the first generation; Mehta is betting the next generation looks more like Warren Buffett's analytical process, run continuously by agents that get better and cheaper over time.

The consumer monetization problem for AI agents, he acknowledges, remains unsolved — he points to Elon Musk's public comments about margins not justifying the compute spend. Enterprise is where the economics work, because companies will pay significantly more for faster, better information that creates competitive advantage.

Abundance operates in public markets but also allocates capital in other areas Mehta declines to detail for now. The two criteria he set when leaving Instacart were scale — potential to reach a trillion dollars in value, citing Anthropic's trajectory as an inspiration — and benefit to humanity. He frames the firm less as a traditional hedge fund and more as a new category: AI-native holding company structures where the old SaaS-versus-fund lines no longer apply.

He's staying quiet on specifics. Expect a clearer picture in a couple of years.