The State of Early-Stage Capital 2026
2026 is the strangest year early-stage capital has had in a decade. Global venture volume is at a record. Early-stage founders have never found it harder to raise. Both of those statements are true at once, and the gap between them is the single most important fact in the market.
This is our reading of where early-stage capital stands, what the numbers actually say, and what changed.
Key takeaways
- Global venture hit a record in the first half of 2026, but the money concentrated: over 70% of Q2 capital went to AI companies, and the bulk of it went late.
- Early-stage selectivity got worse, not better. Roughly 15-20% of seed companies now reach a priced Series A within 24 months, against roughly 30% for the 2018 cohort.
- The gap between rounds stretched past two years. Plan 24 months of runway, not 18.
- Diligence cycles lengthened even as deal volume rose. Our reading: partners are reading more companies per decision, not fewer.
- The binding constraint has moved. It is no longer capital availability. It is evaluation throughput on the investor side, and evidence quality on the founder side.
1. The headline: a record that most founders never felt
Global startup investment reached roughly $510 billion in the first half of 2026 according to Crunchbase, with about $305 billion in Q1 and $205 billion in Q2. US and Canadian startups took around $392 billion of that, up sharply year over year.
Then the distribution:
| Signal | H1 2026 | What it means |
|---|---|---|
| Share of Q2 capital into AI companies (Crunchbase) | over 70% | Up from just under 50% a year earlier |
| Share of Q2 capital into US companies (Crunchbase) | about two thirds | Down from roughly 83% in Q1, as Europe and Asia caught up |
| Europe (Crunchbase) | Strongest venture quarter in four years | UK alone raised over $10 billion in Q2 |
| China (Crunchbase) | Roughly $30 billion, up sharply | Followed by Singapore at about $3.6 billion and India at about $3.3 billion |
A record year in aggregate can still be a famine at the bottom of the stack. Most of the increase is a small number of very large rounds into a single category. If you are raising a $1.5 million pre-seed for a vertical SaaS company, the $510 billion figure describes a market you are not in.
2. What a seed round actually looks like in 2026
The medians, from the two datasets founders most often cite:
| Metric | Figure | Source |
|---|---|---|
| Median seed round raised | $4.1M | Carta, July 2026 benchmark |
| Median seed post-money valuation | $24.3M | Carta, July 2026 |
| Median seed dilution | 18% | Carta, July 2026 |
| Median seed pre-money | $18.4M | PitchBook-NVCA |
| Median seed round size | $3.0M | PitchBook-NVCA |
| Median pre-seed pre-money | $7.7M | PitchBook-NVCA, Q3 2025 |
| Seed-stage capital deployed, Q1 2026 | $12B, up 31% YoY | PitchBook-NVCA |
The two sources disagree, and the disagreement is informative rather than a problem. PitchBook-NVCA tracks priced rounds across the US venture market. Carta sees the transactions that run through its cap-table platform, which skews differently. Cite the source with the number, always. A founder who quotes "the median seed is $4.1 million" without saying whose median invites a correction in the first meeting.
The sectoral spread matters more than the median anyway. AI companies tend to price well above the median; in our reading, most non-AI B2B SaaS lands below it. Benchmark against your category, not against the market.
We break this down in full in pre-seed and seed benchmarks 2026.
3. The graduation problem
This is the number that should drive your planning:
Roughly 15-20% of seed-funded companies raise a priced Series A within 24 months, per Carta cohort data. In Carta's figures the 2022 cohort managed 15.4% within two years; the 2018 cohort managed 30.6%.
The rate did not halve because founders got worse. Our reading is that the Series A bar moved and the definition of a seed round got looser at the same time: more companies are labelled "seed", while far fewer of them graduate on the old timeline.
Two practical consequences:
- Runway maths changed. The median gap between primary rounds ran to about 774 days in Carta's data, around 2.1 years. An 18-month plan puts your raise in the worst possible position: out of money, mid-process, with no leverage.
- The seed round has to buy proof, not just time. The companies that graduate are the ones that closed a specific evidence gap, not the ones that simply survived longer.
Full analysis in the graduation problem.
4. Diligence got slower while deal volume rose
As a rule of thumb from the processes we see, a seed round in 2026 runs roughly three to six months end to end. Inside that, the active window from first meeting to signed term sheet is often six to ten weeks, with a further few weeks from term sheet to wire. Formal diligence on a priced round adds time on its own.
The counter-intuitive part: diligence lengthened during a year of record deployment. The usual explanation is caution. The better explanation is throughput.
In our experience, partners are reading more companies per investment decision than they did a few years ago. Every company arrives in a different format: a deck here, a Notion page there, a Drive folder with inconsistent naming. The reading cost per company is high and nearly all of it is reassembly - working out where the number is, whether it is current, and what it was measured against.
That reassembly cost is paid by the founder in weeks of calendar time. It is also, notably, the one part of the process that structure fixes outright.
Detail in why diligence takes longer than it did in 2021.
5. Where the money is moving
Capital did not just concentrate by sector. It moved geographically.
The US share of global venture fell from roughly 83% in Q1 to about two thirds in Q2, not because US deployment fell but because everyone else accelerated. Europe posted its strongest quarter in four years. China's quarterly total rose sharply. Singapore and India both cleared $3 billion in a quarter. The Gulf continues to build out its own early-stage infrastructure rather than routing through London or New York.
For a founder, the operational implication is that a raise is now genuinely cross-border by default, and cross-border raises fail on documentation before they fail on conviction. An investor three time zones away, in a different legal regime, with no shared network to call, needs a complete and self-serve record. They will not chase you for it.
See the global map of early-stage capital.
6. What actually changed: the constraint moved
Put the four observations together.
- There is more capital than ever.
- Fewer early-stage companies convert.
- Decisions take longer.
- Every participant is reading more companies, from further away, in more formats.
The binding constraint on early-stage capital is no longer how much money exists. It is how many companies a given investor can properly evaluate per quarter - and how much evidence a given founder can present in a form that survives comparison.
That is an infrastructure problem, not a market-sentiment problem. It does not resolve when rates fall.
It is also why the evaluation layer is where the next decade of this market gets built. Public equities solved this a century ago with mandatory structured filings: every company reports the same fields, so any two can be compared by anyone. Private early-stage markets accept a PDF deck, an unverified spreadsheet and a warm introduction, and then express surprise that allocation is slow and noisy.
The standard is coming. The only open question is who sets it.
Method and sources
Figures are drawn from publicly published datasets as of September 2026: Crunchbase News (global and regional volumes, AI share), Carta (seed round size, valuation, dilution, graduation rates, inter-round gaps) and PitchBook-NVCA Venture Monitor (priced-round medians, pre-seed pre-money, quarterly seed deployment). The seed timeline ranges in section 4 are our own rule of thumb, not a published dataset.
Medians from different providers are not interchangeable: they measure different populations. Where two sources disagree we show both rather than averaging them. Figures are restated as published and are not adjusted.
Frequently Asked Questions
It is a good year to raise if you are in a favoured category and can evidence your claims quickly. It is a harder year than 2021 in every other case. Aggregate volume is at a record; early-stage conversion is near a decade low.
Carta's July 2026 benchmark puts the median at $4.1 million on a $24.3 million post-money with 18% dilution. PitchBook-NVCA, measuring priced rounds across the US market, puts the median round size nearer $3.0 million. Use the median for your sector, not the market median.
24 months, not 18. The median gap between primary rounds is now over two years.
Because the record is concentrated. Over 70% of Q2 2026 capital went to AI companies, and most of it went to large, late rounds. Aggregate volume and early-stage accessibility have decoupled.
Read next
Pre-Seed and Seed Benchmarks 2026: Round Sizes, Valuations, Dilution
The Graduation Problem: What Share of Seed Companies Reach Series A
How Long a Seed Round Actually Takes
The Global Map of Early-Stage Capital
The Startup Data Room, Field by Field: What Investors Actually Open in 2026If the constraint is evaluation throughput, the fix is structure. RaiseTalks gives founders a 143-field record investors can read in one pass, and gives investors a dealflow pipeline with one-click AI pre-scores instead of an inbox. See how it works.