In March 2026, Paris-based Advanced Machine Intelligence raised $1.03 billion — in a seed round. Three months earlier, San Francisco’s Unconventional AI closed a $475 million seed to build energy-efficient silicon. In January, Humans& raised $480 million at a $4.48 billion valuation, also labeled seed. None of these companies has meaningful revenue. All three are earlier, by any operating measure, than a typical Series A.

The Data

Crunchbase counted at least 27 seed rounds of $100 million or more announced globally since the start of 2025 — a size class that was “once exceedingly rare.” The share of seed deals at $10 million or above has climbed from 2% of all seed deals in 2018 to 9% today, per the same Crunchbase dataset. Meanwhile the deal-count floor hasn’t moved: the majority of seed rounds are still under $5 million, and that share has been trending down only slowly.

This is not the seed market getting bigger across the board. It’s the seed market splitting into two markets that happen to share a label. At one end: physical-AI and frontier-model labs (Advanced Machine Intelligence’s world-model research, Periodic Labs’ $300 million round for AI-driven materials science) raising sums that would have been considered large Series B checks three years ago, priced on team pedigree and compute roadmap rather than traction. At the other end: the median founder is still raising a few million dollars against early usage signals, competing in a pool where seed rounds keep skewing larger and more competitive even without a mega-check.

Why It Matters

For founders outside frontier AI or deep physical-world compute, the jumbo-seed headlines are noise, not benchmark. A founder building an applied AI product should not anchor pricing expectations on Humans&’s $4.48 billion valuation — that capital is chasing a different risk profile (unproven science, multi-year time horizons, small teams of research talent) than a product company chasing initial revenue. Conflating the two markets leads to founders over-asking on unproven traction, and to operators misreading “AI is flush with seed capital” as a signal that applies to their category.

For investors, the barbell creates a structural problem: capital concentrated at the top 27 deals is capital not available to fund the next cohort of applied, revenue-near AI companies at reasonable terms — even as seed-stage deal counts fall and odds of reaching a Series A drop for the broader pool. The two markets aren’t just differently sized; they’re pulling attention and follow-on capital away from each other.

The Charaka View

Our knowledge graph tracks funding-round structure alongside traction signals for every company we assess, and the pattern is consistent with what Crunchbase’s aggregate data shows: the highest-conviction seed checks in 2026 are going to teams with research-lab pedigree and a science bet, not to teams with early revenue. That’s a rational allocation for the labs writing $300M–$1B checks — but it means most founders should stop reading “seed funding is booming” as a market-wide signal. It describes roughly two dozen deals, not the market.

This analysis draws on Crunchbase News: The Largest Recent Seed Rounds Are All For AI Companies, Crunchbase’s Europe’s Largest Seed Round Ever coverage, and Crunchbase’s Humans& seed round report. Human editorial oversight applied.

This analysis is informational and does not constitute investment advice, a research report, or a recommendation to buy, sell, or hold any security.

Charaka Notes by Manthan Intelligence. Subscribe