Robinhood Ventures Fund II, the brokerage’s venture arm, holds stakes in roughly 80 startups funded through $250K SAFE notes, with a pronounced Y Combinator tilt. The fund’s shares begin trading on the NYSE on August 13 at $25 apiece. Its portfolio reads like a map of where the agent economy is being built, and a large slice is devoted to dev tools and vertical AI: the runtime layer agents depend on, the compute behind them, and the industry-specific products that put them to work. The pattern mirrors the broader venture shift toward agentic software, but the fund’s SAFE-heavy structure keeps the checks small and the basket wide.
The fund cuts $250K SAFE checks into Y Combinator and early-stage companies rather than leading priced rounds. That structure keeps the bar low for entry and spreads risk across dozens of names, which is why a single fund can hold positions from agent runtimes to digital collectibles. It is a breadth-first approach that trades headline returns for exposure to the whole agent stack. The NYSE listing at $25 a share gives everyday investors a way to own the basket, and the August 13 debut puts a public price on a portfolio that most observers would otherwise only see in the prospectus.
The infrastructure tier starts with ReasonBlocks, which builds the runtime layer AI agents rely on for dependable execution, the software that sits between a model and the tools it calls. Carnot AI, operating as Jinba, sells enterprise workflow automation that carries multi-step processes to completion rather than stopping at a chat reply, a distinction that matters as agents move from demos to production. On the compute side, Cumulus Compute Labs runs a serverless GPU cloud built for AI workloads, giving developers on-demand capacity without the usual infrastructure overhead. The fund’s agent infrastructure names cluster around the same thesis: agents are only as useful as the systems underneath them, and those systems are where the durable value accrues.
A second cluster puts AI to work inside specific workflows. JigsawStack packages purpose-built AI models for deterministic tasks, a fit for teams that need consistent, repeatable outputs instead of open-ended generation. KelAI runs autonomous AI quant strategies, bringing agents to markets that historically leaned on manual research. Oxus applies AI to the internal audit function, automating review workflows for finance teams, while Caretta delivers real-time AI analysis during sales calls so reps get coaching as they talk. Crow brings AI to commercial real estate workflows, Veriad acts as an AI email and calendar assistant, and Autumn AI supplies prospect signals to sales teams. Each product owns a single workflow end to end, which keeps the models deterministic and the pricing clear.
The fund also backs agentic products for operations-heavy industries. Qomplement builds agentic ERP for supply chains, Arzana develops autonomous ERP for manufacturers, and Amboras runs AI-native e-commerce for brands that want a storefront that manages itself. In construction, Rudus handles AI takeoffs for concrete contractors, replacing manual quantity estimation with models trained on project plans. The takeoff process is notoriously slow and error-prone, which makes it a natural early target for automation, and contractors are a group that historically adopted software late. That lag is exactly what makes the takeoff market attractive to a fund looking for under-served verticals.
Services and consumer bets round out the cohort. Sarah AI is building a context layer for CPG brands, giving consumer goods teams a shared memory of their own data, while Sparkley uses AI agents to grow revenue for home services businesses by handling the selling and follow-up work. RMJ Labs deploys AI agents that build personal injury cases, automating intake and documentation for plaintiff firms, work that traditionally consumed paralegal hours. Unilabs runs AI voice agents for on-prem telephony systems, Formative Intelligence is testing an AI-original films subscription, and CatchBack Cards sells digital collectible packs aimed at sports fans. The mix is a reminder that the fund treats vertical AI as a long tail of niches rather than a single winner.
What links these names is a shared wager: that the winning AI companies are the ones that take on a specific job, not the ones that promise to do everything. The vertical AI bets each own a workflow, from audit to takeoffs to case building, and the dev-tools names sell them the parts. The fund’s NYSE listing arrives August 13, and its prospectus names all 80 holdings, with this cohort carrying the dev-tools and vertical AI thesis.
