Instacart Co-Founder Apoorva Mehta Launches AI Hedge Fund Abundance
Instacart co-founder Apoorva Mehta’s AI hedge fund Abundance uses agent swarms for full investment decisions. Thousands of bots replace PMs; raised $100M seed.
Apoorva Mehta’s Abundance hedge fund replaces human portfolio managers with thousands of AI agents handling research, stock picking, bet sizing, and trade execution—claiming benchmark-beating returns while trading its own capital.
AI Agents Run Everything
Abundance (Palo Alto, launched 2025) deploys thousands of AI agents that independently:
-
Scour internet for trade ideas
-
Conduct fundamental research
-
Pick long/short stocks
-
Size positions
-
Execute trades
10-person team (quants, engineers, AI experts) builds/sustains models. Currently trades own capital; external money planned. Raised $100M seed equity.
Post-Instacart Vision
Mehta (39), Instacart co-founder (2012-IPO exit), was inspired by OpenAI o3‘s reasoning capabilities. Hypothesis: Generative AI can handle “consequential decisions” like capital allocation. Abundance tests this at scale—fully AI-native hedge fund replacing fundamental PMs.
India Quant Opportunity
Mumbai/Bangalore quants: Abundance model validates agentic finance for India’s 1M+ fintech engineers. Local funds can deploy similar swarms using GPT-5.5 Terminal-Bench leaders (82.7%) or DeepSeek-V4 (1M context) at 1/10th Western costs. JioCloud → NSE trading agents now viable.
Current Status & Expansion
✅ Live AI stock-picking strategies (no human PMs)
✅ Trading own capital (benchmark-beating returns)
✅ $100M seed funding secured
🔜 External investor capital
🔜 Multi-asset class expansionSome strategies retain limited human oversight during development, but endgame remains zero human discretion.
Competitive Edge Claims
Mehta claims multiple index outperformance (specifics undisclosed). Key differentiator vs traditional quant funds:
-
Agent swarms vs single models
-
End-to-end automation (research→execution)
-
Zero fundamental PMs (vs AI-supporting humans)
India angle: Mumbai HF teams can replicate at lower costs using open-source DeepSeek-V4 + NSE data feeds.
Technical Architecture
AI Agent Workflow:
1. Data ingestion (web, filings, news)
2. Research synthesis (o3-level reasoning)
3. Thesis generation (long/short pairs)
4. Position sizing (risk models)
5. Execution (market making optimized)Post-GPT-5.5/DeepSeek-V4 era: Terminal-Bench 82.7% + 1M context enables exactly this stack. Bangalore engineers build identical systems profitably.
Global Fintech Shift
Abundance validates what Indian startups suspected: AI agents > human PMs for alpha generation. With GPT-5.5 Terminal-Bench victory and DeepSeek-V4 universal access, Mumbai funds gain Western hedge fund capabilities at startup costs. The quant revolution just became agentic.
Apoorva Mehta’s Abundance proves AI hedge funds work—thousands of agents replace portfolio managers entirely. $100M seed + benchmark wins confirm viability. Indian quants: Build your version now using GPT-5.5 + NSE data. Finance’s future runs on agent swarms.
