Uncapped with Jack AltmanMike Volpi on Why AI Breaks Traditional Venture Capital | Ep. 52
CHAPTERS
Why AI creates an opening to build a new kind of venture firm
Volpi explains that breaking into venture is difficult unless a major macro shift is underway—and he argues AI is that shift. He lays out why new firms must be narrowly focused on the wave, staffed by “native speakers,” and careful not to overfit to prior-era success patterns.
How AI changes the assumptions behind software—and therefore VC strategy
The conversation shifts to how AI collapses the cost and time to build software, which breaks many classic VC heuristics. Volpi argues this impacts everything from company formation to go-to-market, and makes many incumbent investment playbooks less reliable.
Designing Hanabi: people-first firm design and true AI fluency
Volpi describes starting firm design with talent: investors must be able to speak the technical language founders speak. He notes AI-native experience is still scarce, but catch-up is possible if someone has the right technical foundation and curiosity.
Why ‘stage focus’ matters less in the AI era
Volpi argues stage-based identities (seed vs growth) are less useful because AI companies can produce venture-like outcomes even from very high valuations. He emphasizes opportunity magnitude over stage, and suggests late-stage entries can still yield outsized returns.
Access and reputation: the real constraint at later stages
While valuation and analysis can be learned, Volpi says relationship proximity becomes the gating factor as company seniority rises. He contrasts building trust with very young founders vs. established leaders, and explains how reputation influences allocation in competitive rounds.
Venture brand in 2026: organic reputation beats traditional marketing
Volpi views brand as essential, especially for younger founders with less capital-market context, but says it must be built differently now. Instead of sponsorships and glossy marketing, brand is conveyed through high-trust references, insider help, and founder-to-founder signal.
Board seats vs. high-bandwidth support: evolving engagement models
Given a small fund size, Hanabi often leads seeds and selectively co-leads Series A’s, while also investing into later-stage winners. Volpi downplays board seats as a status symbol and prefers frequent 1:1 check-ins that match the founder’s real-time problems.
What makes an enduring VC firm: the four-job model and lean teams
Volpi reduces venture to four core jobs: sourcing, judgment, selling the founder, and helping the company succeed. He argues many modern VC “decorations” bloat firms, and predicts smaller, well-rounded teams will outperform highly specialized, stitched-together models.
The hardest problem in venture: generational transition and incentives
They discuss why firm longevity is uniquely difficult because venture feedback cycles are long. Volpi argues many firms fail when legacy partners capture economics while contributing less, and that enduring firms must “bet on the future” by handing opportunity to the next generation.
The frontier lab landscape: ‘the winners are the winners’
Volpi frames AI as a stack and argues the central model layer is dominated by a small set of labs whose advantage is compute scale tied to capital. He believes new entrants are structurally disadvantaged unless architectures change meaningfully.
Open source models: important phenomenon, weak business at the frontier
Volpi says open source matters but mostly commoditizes the lagging edge, not the monetizable frontier where most prompting and dollars concentrate. He argues the cost to train near-frontier models pushes even open-source-aligned teams toward closure over time.
Neolabs and where they can win: proprietary data pockets (robotics, science)
Volpi is skeptical of neolabs built on generic algorithmic claims (e.g., “we do RL better”) because top labs are already investing heavily there. He sees opportunity where proprietary data is defensible and not broadly available—highlighting robotics embodiment data and lab-generated scientific datasets.
The compute race: inference explosion, supply constraints, and post-NVIDIA specialization
Compute demand is rising from both training and inference, while supply is constrained by foundry capacity. Volpi expects a move away from NVIDIA-only dependence toward specialized inference silicon (e.g., Cerebras, ASIC-style designs), alongside geopolitical and domestic manufacturing considerations.
The future of software: data + workflows as moats, and the return of ‘services’
Volpi argues durable AI applications will own proprietary business data and embed deeply into workflows—especially in verticals the big labs won’t prioritize. He predicts SaaS models will split between low-cost commodity software and high-value, customized deployments that blend agents with humans (FDE-like roles).
Defense tech as a non-AI tailwind—and why the sector is reopening
Though Hanabi is mostly AI-focused, Volpi highlights defense as a structural opportunity driven by geopolitics and rising European spend. He argues Anduril/Palantir normalized venture-backed defense adoption and changed DoD openness, making the category more investable for new entrants.
From operator to investor: differentiation, empathy, and ‘beginner’s mind’ with founders
Volpi reflects on how operating experience (Cisco’s hypergrowth) became a differentiator in a crowded VC market, though he notes it’s not required to be great. He closes with observations on today’s unusually mature young founders and the mindset older investors need: treat founders as equals and stay willing to relearn.
