We publish a directory of venture funds raised since 2022. It holds 441 of them, every entry with a source link. When I ran the size distribution last week, the shape was not what I expected.
106 funds at a billion dollars or more. 126 between $100 million and $500 million. Six below $25 million.
Six.
The obvious reading is that small funds are rare. The correct reading is that our directory cannot see them, and neither can anyone else's, because a $60 million first close does not generate a TechCrunch headline, a Bloomberg terminal alert, or a press release anyone syndicates. The dataset is built from what gets reported, so it inherits the reporting bias exactly.
That gap is worth sitting inside for a minute, because the funds it hides are the ones most people reading this are trying to raise.
The barbell is real, and the middle is the worst place to be
The structural picture is well documented even where the individual funds are not.
2025 was the trough. Global VC fundraising came in at $118.4 billion, described across the industry as exceptionally weak. Around 537 US funds closed, raising roughly $66.1 billion, which is about 30% of the 2021 peak. The capital that did move went almost entirely to established, top-quartile managers, and first-time managers faced conditions that several sources describe as near-impossible.
Underneath that, the market has split into a barbell. At one end, micro-funds of $1 million to $10 million, mostly solo GPs, increasingly viable because the operating stack got cheap. At the other, platform funds of $500 million and up, raised by firms with institutional relationships already in place.
Between them, roughly $50 million to $200 million, is the hardest money in venture to raise.
The reason is structural rather than about quality. A fund in that band is too large to be filled by friends, angels and a handful of family offices writing $250,000 cheques. It is also too small for most institutional LPs, whose minimum cheque and concentration limits mean they cannot write less than $10 million and cannot be more than 10% of your fund. Do that arithmetic and a $100 million fund needs an LP willing to write $10 million, which is an LP who mostly does not look at first-time managers.
So the middle fund has to be assembled from a category of investor that barely exists in size: allocators large enough to write meaningful cheques and flexible enough to back an unproven manager.
The clock nobody plans for
The second thing the data says clearly: it takes far longer than anyone budgets.
The average venture fundraise now runs about 20 months, roughly double the pre-pandemic figure. 38% of funds take more than two years to close.
Two years is not a fundraising timeline. It is a full economic cycle, an entire vintage, and long enough that the thesis you opened with may not be the thesis you close on. It also means the personal-runway question is the real gating factor for most first-time GPs, not the pitch. If you cannot fund yourself for two years while raising, you do not have a fundraising problem, you have a sequencing problem, and it needs solving before the deck does.
This is where GP seeding has started to matter. It is maturing as its own asset class precisely because somebody noticed that the constraint on new managers is survival during the raise rather than the raise itself.
Specialisation stopped being an edge and became a filter
The share of generalist funds among new launches has fallen from 22% to about 5%.
That is a striking number, and it is easy to misread as "specialists win." The more useful reading is that specialisation no longer differentiates you, it merely qualifies you. Turning up as a generalist now reads as not having decided.
The bar has moved further than most decks have. Saying you invest in AI is not a thesis when several hundred other funds say the identical sentence. What gets read now is a layer: inference economics, memory architecture, hardware-software convergence, vertical applications with a defensible data position.
There is a real opening in deep tech for exactly this reason. Global deep tech venture funding reached $48 billion in 2025, up from $18 billion in 2020, and now represents over 20% of all venture globally. Most generalist funds pass after one meeting, not out of discipline but because they cannot evaluate a cost curve, a pilot reliability figure or a supply chain assumption. Technical depth is one of the few remaining moats that an emerging manager can actually hold against a larger firm.
Where the middle is being raised
If the missing middle is hard everywhere, it is measurably less hard in some places than others.
65% of new venture firms are launching outside the United States. That is not a marginal shift, it is most of the formation activity happening outside the market that gets covered.
MENA is the clearest case. Startup funding in the region hit $6.6 billion year to date by Q3 2025, already ahead of most full years since 2021, with AI investment up 134%. Abu Dhabi is positioning deliberately as a regional AI capital through ADIO and Hub71, and DIFC's Innovation Hub is doing the equivalent work in Dubai.
The more important structural point is about who the LPs are. Family offices now account for around 31% of all capital going into startups. They are pulling back from public markets and increasing private allocations, and critically they are not bound by the concentration rules that make institutional LPs structurally unable to fund the middle. A family office can write $10 million into a $100 million fund and does not need a committee to agree it fits a mandate.
That is the population that makes a middle-sized fund raiseable, and it is disproportionately concentrated in the Gulf.
What this means if you are raising one
Three things follow from the numbers rather than from anyone's opinion.
Pick a side of the barbell deliberately. A $60 million target is not a smaller version of a $200 million fund, it is a different LP base, a different cheque size, and a different set of conversations. Drifting into the middle because it sounded reasonable is the most common way to end up eighteen months in with a first close that never closes.
Budget the calendar honestly. Twenty months average, two years for 38%. Solve your own runway first, then build the deck.
And accept that nobody is going to write about it. The absence of coverage for funds in this band is not a signal about their quality. It is a reporting artifact, and it is visible in our own directory, where six funds under $25 million sit alongside 106 above a billion. The billion-dollar funds are not 17 times more numerous. They are 17 times more legible.
The quiet part of this market is most of it.
Sources: fund distribution from the Venture Protocol fund directory, 441 funds raised 2022 to 2026, each with a source link. Market figures from Crunchbase, Carta's Fund Economics Report 2025, SVB emerging manager research, VC Lab, Wamda and the CFA Institute. The directory is compiled from public announcements and inherits their bias toward larger funds, which is the subject of this piece.
