Analysis
Bison Ventures, an early-stage venture firm, closed its second fund at $160 million to back founders building what it calls "Real World Intelligence," following its $135 million debut fund, according to a Business Wire release carried by Morningstar. With Fund II, the firm says it will continue investing from pre-seed through Series B in founders building what it describes as non-incremental technology.
Bison's existing portfolio gives some shape to the "Real World Intelligence" thesis: Zyphra, which is building frontier AI models and infrastructure on AMD's full-stack compute platform, is among the names the firm points to, alongside Cobot, Vivodyne, Thinking Machines, Grid Aero, Great Sky and Inner Logic. That mix spans AI infrastructure, robotics and biotech simulation under a single fund mandate rather than a narrow sector focus.
“That mix spans AI infrastructure, robotics and biotech simulation under a single fund mandate rather than a narrow sector focus.”
A modest step-up, not a re-rating
A jump from $135 million to $160 million -- about 19% -- is a meaningfully smaller step-up than the fund-size increases some AI-focused early-stage firms have posted this year, suggesting Bison's LPs are backing continuity in the firm's existing thesis and portfolio rather than betting on a dramatically larger fund deploying into a hotter market. Investing across pre-seed through Series B with a single fund is also a wider stage mandate than many specialist seed shops take on, which raises a real diligence question for LPs: how much of Fund II's capital is earmarked for new pre-seed bets versus follow-on checks into standouts like Thinking Machines, where reserves can quietly consume most of a fund's dry powder if a portfolio company breaks out.
The Zyphra bet is also specific and checkable: building frontier AI infrastructure on AMD's compute platform is a different wager than the Nvidia-centric stack most AI infrastructure startups and their backers have defaulted to, and its payoff depends heavily on whether AMD closes the gap with Nvidia on training-workload performance and software tooling over the life of this fund.
