US startups raised $412.7 billion in the first half of 2026, according to the Q2 2026 PitchBook-NVCA Venture Monitor (data as of June 30, 2026). That already tops the whole of 2025, and every other full year on record. Read one line further and the story flips: deals under $100 million, the size bracket a seed-to-Series-B AI-native company actually raises in, took just 12.5% of that money, down from 33.1% a year earlier. The record and the squeeze are the same report, and if you did not close one of the seven billion-dollar-plus rounds this year, the second number is the one that describes your market.
What the Venture Monitor actually measures
The PitchBook-NVCA Venture Monitor is a quarterly report jointly published by PitchBook and the National Venture Capital Association (NVCA), built from PitchBook's own US deal database rather than a survey of founders or investors after the fact. The Q2 2026 edition, dated as of June 30, 2026, covers dealmaking, fundraising and exit activity through the first half of the year, with added commentary from J.P. Morgan's commercial banking group. It is downloadable directly from nvca.org rather than paraphrased through a press release, which is the difference between a primary source and a summary of one.
The headline number, and the one underneath it
$412.7 billion sounds like an unambiguously good year to be raising. Almost all of the increase, though, sat in a shrinking number of very large checks. Megadeals of $100 million or more made up 87.5% of that total, and the share going to everything smaller has compressed hard, year over year:
| Year | Share of US VC dollars in deals under $100M |
|---|---|
| 2024 | 43.8% |
| 2025 | 33.1% |
| 2026 (H1) | 12.5% |
Source: PitchBook-NVCA Venture Monitor, Q2 2026 edition, data as of June 30, 2026.
Q2 2026 alone produced seven rounds at or above $1 billion, five of them for AI companies, totaling $87.2 billion between them. That is where the record actually lives. The remaining 12.5% of dollars is what a normal, non-mega round is competing over now, and that bracket has gotten smaller every single year since 2024.
AI's 86% share is not the same thing as your AI company doing fine
AI companies took $355.9 billion of the total, 86% of every US venture dollar deployed in the first half of 2026, per the same report. Read that as good news for AI-native founders generally and you would be reading it wrong. Ginger Chambless, J.P. Morgan's Head of Market Insights for Commercial Banking, describes venture capital splitting into two distinct strategies in the same edition: a small set of investors running concentrated, very large positions in a handful of frontier AI companies, and a separate group doing classic early-stage portfolio construction. Her line for everyone else: "For startups that do not align with one of these active VC strategies, a strong and differentiated AI story to tell, or a perceived category winner in their space, liquidity remains highly selective and below the 2021 peak." She also notes a rise in undisclosed deal terms in that middle ground, which the report ties to structured rounds built to give investors downside protection, meaning down rounds and flat rounds dressed in quieter language.
Valuations climbed too, and that is not free money
Median pre-money valuations have not just recovered from the 2022-2023 correction, they have pushed past 2021 levels at every stage, per the Q2 2026 Venture Monitor (as of June 30, 2026). Pre-seed, seed, and Series D+ valuations more than doubled relative to 2021; Series C came closest among the middle stages at 95% above 2021 levels, Series A at 89.9%, and Series B at 83.3%. The report is direct about the cost of that: "the time between rounds has compressed, especially for AI companies, so the margin for error is thin." A higher valuation is a higher bar the next raise has to clear, not a cushion, and a company that priced its last round against 2026 comparables is now on the clock to show growth that justifies it before the market reprices around it.
A worked example
Picture a 34-person Series A AI-native company that closed its round in February 2026 at a valuation the founders privately knew was priced for a strong 2027, not for where the product actually stood that month. The plan assumed 18 months of runway to hit the usage numbers a Series B would want to see. Read against the Q2 2026 Venture Monitor, that plan has less slack in it than the calendar suggests: the round sizes worth competing for outside the megadeal bracket keep shrinking, the time between rounds keeps compressing, and this company is not a frontier AI name that draws a concentrated bet on its own. The founders' real decision is not whether to hire. It is whether the next three senior engineers are permanent headcount that extends the burn for years, or capacity that can ramp inside two weeks and come off the books the moment the roadmap changes.
What actually fits a runway this tight
Staff augmentation exists for exactly that gap: senior engineering capacity added without the multi-year commitment a direct hire represents, useful specifically in the stretch between closing a round and proving the metrics the next one wants. It is the wrong tool when a company already has two-plus years of runway lined up and is building the differentiated system meant to be its own institutional knowledge; that work belongs to people who stay, not a bench that ramps down the moment a contract ends. The honest read of this report is that most funded AI-native companies are neither the frontier bet nor flush with runway. They are the middle the Venture Monitor describes as newly squeezed, and a hiring plan built for 2021 conditions does not match that position anymore.
If your team is working out which open reqs should be a direct hire and which should be staff augmentation before your next round gets priced, that is the scoping conversation our Silicon Valley team has before a sprint gets planned around funding that has not landed yet. It sits next to what the current senior-hiring market actually costs in time, and if the real gap is proving usage rather than headcount, building the eval harness that shows a feature is production-ready is usually the faster path to the metrics your next investor will ask for. Talk to us about where staff augmentation fits your specific timeline.
Sources
- National Venture Capital Association & PitchBook: Q2 2026 PitchBook-NVCA Venture Monitor (data as of June 30, 2026)
- National Venture Capital Association: Q2 2026 Venture Monitor report page