AI Venture Capital Firms Pre-Seed and Seed for Agent Infrastructure Startups

Most lists of AI venture capital firms pre-seed and seed founders find online are useless for founders building infrastructure, because they're dominated by the same dozen mega-funds writing checks into foundation models and consumer AI apps. If you're building agent identity, governance, or autonomy infrastructure, the firms that actually understand your category are a different, much smaller list.

Here's how the landscape breaks down, and who's actually active at pre-seed and seed.

AI Venture Capital Firms Pre-Seed and Seed: The Generalist Funds

These firms invest broadly across the AI stack, from foundation models to consumer apps to infrastructure, and most pre-seed founders will encounter them first.

Gradient Ventures, Google's AI-only fund based in Palo Alto, closed a $220M Fund V in March 2026 with nearly $1.2B in total AUM, investing at pre-seed and seed across AI applications, agentic platforms, and real-world AI systems with a typical 3-6 week decision timeline. Sky9 Capital

NFX, a San Francisco pre-seed and seed fund with $875M in total AUM, made 28 AI seed checks in 2025-2026 averaging $800K-$1.2M, focusing on AI agents, multimodal models, and developer tools. Sky9 Capital

2048 Ventures is a thesis-driven early-stage fund focused on Vertical AI, Deep Tech, Health, and Bio, leading pre-seed and seed rounds with check sizes from $500K to $2M. Sky9 Capital

Khosla Ventures runs a $650M dedicated seed fund with particular strength in AI healthcare, robotics, and frontier models, though average seed checks there tend to run larger than most pre-seed founders need. Sky9 Capital

Infrastructure-Specific AI Funds

A smaller group of firms specifically prioritize the infrastructure layer, the picks-and-shovels companies that other AI products depend on, rather than chasing application-layer hype.

Alumni Ventures runs a dedicated AI Infrastructure syndicate that invests in the systems training, deploying, and scaling intelligence: developer tools, orchestration frameworks, GPU optimization, data infrastructure, and AI security, the rails and operating systems of what they describe as a new infrastructure-driven economy. Alumni Ventures

IA Ventures, founded in 2009 by Roger Ehrenberg and based in New York, focuses on seed and Series A investing in fintech, logistics intelligence, and AI infrastructure, evaluating proprietary data assets as the primary investment criterion. Qubit Capital

Air Street Capital and Radical Ventures are also frequently named among the more technically rigorous infrastructure specialists, though both tend to write later and larger checks than a typical first round.

The Agent Governance and Autonomy Control Plane Gap

Here's the category that almost no list captures, and it's the one we think matters most. As AI agents move from assisting humans to executing entire workflows autonomously, a new layer of infrastructure is emerging to govern them: agent identity, permissions, policy enforcement, and audit systems. Investor activity is already showing up here, with companies like Defakto, focused on non-human identity lifecycle management for machine identities, AI agents, and cloud/agent footprints, recently raising a $30.75 million Series B. But dedicated pre-seed and seed funds in this exact category remain rare. Crescendo AI

Untapped Ventures is one of the few funds built specifically around this thesis. We invest $500K to $2M+ at pre-seed and seed in founders building the Autonomy Control Plane: agent identity, governance, permissions, audit trails, and the machine workforce management systems that let enterprises trust autonomous agents at scale. Unlike generalist AI funds, we don't invest across the whole stack. We look specifically for categories that don't have a market map yet, evaluated by how much of a workflow a system runs without human intervention, not how impressive the underlying model is.

How AI Venture Capital Firms Pre-Seed and Seed Differ by Focus

If you're building a foundation model, a vertical AI application, or a developer tool, the generalist funds above are your natural first calls; they have the check sizes and pattern recognition for that category. If you're building the infrastructure that makes agents trustworthy, accountable, and safe to deploy at scale, that's a narrower list, and most of the capital in it hasn't been allocated yet.

---

Frequently asked questions

What VCs invest in pre-seed and seed AI startups?

Active pre-seed and seed AI investors include generalist funds like Gradient Ventures, NFX, and 2048 Ventures, infrastructure-focused syndicates like Alumni Ventures' AI Infrastructure fund, and category-specific firms like Untapped Ventures, which invests specifically in AI agent governance and autonomy infrastructure.

What is the difference between a generalist AI VC and an infrastructure-focused AI VC?

Generalist AI VCs invest across the full stack, from foundation models to consumer applications, while infrastructure-focused firms concentrate on the underlying systems other AI companies depend on, such as compute, data pipelines, or in Untapped Ventures' case, agent identity and governance.

Are there VCs specifically investing in AI agent governance and identity startups?

Dedicated funds in this exact category are still rare, though investor activity is increasing as non-human identity and agent governance startups raise institutional rounds. Untapped Ventures is one of the few pre-seed and seed funds built specifically around this thesis, investing in agent identity, permissions, and autonomy control plane infrastructure.

Who is Untapped Ventures and what do they invest in?

Untapped Ventures is an early-stage venture capital firm investing in Autonomous Economy startups, particularly those building the Autonomy Control Plane and Machine Workforce layers of infrastructure. Fund I deploys $500K to $2M+ checks into pre-seed and seed-stage technical founders building systems with no existing market map.

What does Untapped Ventures look for in startups?

Untapped Ventures looks for founders building categories that don't yet exist on market maps from firms like CB Insights or PitchBook, evaluated using an "Autonomy Density Score" measuring how much of a workflow runs without human intervention. Priority areas include agent identity and PKI, policy-as-code engines, and machine workforce management platforms.