When Robinhood Ventures Fund II starts trading on the New York Stock Exchange on August 13 under the ticker RVII, it will hand retail investors a rare view inside a seed-stage venture portfolio that has long been closed to everyday investors. The business development company’s roughly 80-company portfolio, assembled mostly through $250,000 SAFE investments, includes three bets that touch the semiconductor industry directly: Visibl Semiconductors, Matforge and Expanse Compute.
The fund’s N-2 registration statement, filed with the SEC on August 3, lays out a mandate of long-term capital appreciation through early-stage and growth-stage private companies, most of them illiquid holdings that retail investors could not otherwise reach. It is selling up to 7.6 million shares at $25.00 apiece, with a selling shareholder offering another 400,000, and the proceeds flow into a portfolio that skews heavily toward Y Combinator-linked startups across AI, fintech, enterprise software, consumer technology and robotics. Most positions take the form of SAFEs worth $250,000. The three semiconductor names are small pieces of a diversified spread, but together they map to three of the industry’s most stubborn bottlenecks: the cost of chip design, the pace of materials research and the utilization of GPU capacity.
AI agents for custom silicon
Chip design remains one of the most expensive and slowest stages of the silicon lifecycle, and Visibl Semiconductors is attacking it with AI agents. The San Francisco startup, which signed its SAFE on March 19, builds “AI agents for faster, lower-cost custom silicon.” The pitch is simple: teams that need application-specific chips should not have to staff a multi-year, multi-million-dollar design effort when software agents can automate the drudge work of architecture exploration, verification and implementation. Modern designs demand armies of verification engineers, and that headcount is precisely where agents can bite hardest. As demand for custom silicon spreads from data center accelerators to edge devices, tools that compress design cycles attack a cost structure that has kept bespoke chips out of reach for most product teams.
AI scientists for materials
Further upstream, Matforge applies a similar agentic playbook to semiconductor materials. The company, funded through a $250,000 SAFE on June 1, fields “AI scientists for semiconductor materials discovery,” using machine learning to screen candidate materials and predict their properties before a single wafer is processed. Materials research is a slow, trial-and-error discipline, and each new process node leans on substances that historically took years to qualify. A model that can simulate years of experiments in days could compress development timelines for interconnects, dielectrics and advanced packaging alike.
An intelligence layer for idle GPUs
Expanse Compute sits on the other side of the wafer, building an “intelligence layer that unlocks wasted GPU capacity.” The thesis tracks a well-documented industry reality: even amid the current AI buildout, a meaningful share of accelerator capacity sits idle or underutilized, fragmented across clouds and mismatched with workloads that burst and then go quiet. Reports from cloud providers and AI labs suggest utilization across the installed base remains surprisingly uneven. Expanse’s layer aims to route work toward available silicon the way a trading desk allocates capital, monetizing capacity that would otherwise go dark.
None of the three companies is a household name, and each is a $250,000 position in a fund that will carry the liquidity and valuation risks common to early-stage venture. For now the fund’s YC tilt means most of its names are early, pre-revenue or barely post-revenue, so the three silicon bets read as long-horizon options rather than near-term earnings plays. But the fact that all three sit inside one retail-accessible vehicle is itself a signal: the chip industry’s next wave of value creation is increasingly being driven by software that makes silicon cheaper to design, faster to develop and more fully utilized. When RVII opens on the NYSE on August 13, that thesis becomes available to any brokerage account.
