Crunchbase reports $65 billion in global venture funding for July 2026, up 100% year over year, with a record fourteen billion-dollar rounds but not a record dollar total in those rounds. The megaround has become an accessible size rather than a singular event, and it has crossed into capital-intensive hardware — space, defence, energy, semiconductors — which carries a different risk profile than 2021's software megadeals.
Global venture funding reached $65 billion in July 2026, up 100% year over year and 10% over June, making it the third-largest month of the year. Fourteen startups raised rounds of a billion dollars or more — the highest count in any single month on record.
Buried in Crunchbase's own write-up is the qualifier that changes what the month means: it was the highest count of billion-dollar rounds, but not the largest amount raised in such deals. A record number of megarounds that did not produce a record megaround total.
That gap is the story. It says the billion-dollar round has stopped being a singular event — one frontier lab absorbing an extraordinary sum in a given month — and has become a round size that a dozen-plus companies can access simultaneously. The megaround has been institutionalised.
All funding figures here are Crunchbase's own data as reported on August 4, 2026. Crunchbase notes that early-stage figures lag and are revised upward over time. The Safe Superintelligence round is described by Crunchbase as reportedly $5 billion; treat it as unconfirmed.
Nine of the fourteen were US companies. Two were German, two Chinese, one headquartered in Singapore.
The largest was $10 billion into Blue Origin — the first external financing in the company's history, which until now had been funded principally by Jeff Bezos. Behind it: Safe Superintelligence, reportedly at $5 billion from Nvidia; Moonshot AI at $3.5 billion; Kling AI at $2.8 billion. The rest of the cohort spans German defence technology, energy, industrial robotics and semiconductors alongside the frontier labs.
Read that list against the last comparable funding peak. In 2021, the megaround was overwhelmingly a software instrument — fintech, SaaS, marketplaces, delivery. Companies that could deploy a billion dollars into engineering headcount and customer acquisition and, if the thesis failed, wind down with the primary loss being the cash itself.
July's cohort is substantially physical. Rockets. Defence hardware. Energy infrastructure. Robots. Fabs. That is a different instrument wearing the same name.
Three things, and none of them are captured by the funding headline.
First, payback lengthens. A software company deploying a billion dollars can show revenue traction within the year. A launch company deploying the same amount is buying tooling, test infrastructure and a manufacturing base whose returns arrive across a decade. The capital is committed for far longer before anyone learns whether it worked, which means the fund holding it is exposed to a longer window of macro, regulatory and competitive change.
Second, failure modes become physical. Software fails by not being adopted — recoverable, and the team and code retain some value. Hardware fails by exploding on a pad, by a fab yielding badly, by a component supplier missing a spec. These failures are discrete, expensive and sometimes unrecoverable, and they are not correlated with the quality of the go-to-market motion.
Third, and least discussed, optionality collapses. A software company that discovers its thesis is wrong can pivot; the engineers, the codebase and the cash are largely redeployable. A company that has spent $10 billion on rocket manufacturing infrastructure cannot pivot into anything. The capital is embedded in physical assets specific to one thesis. If the thesis is wrong, the recovery value is scrap and real estate.
That is not an argument against funding hardware. Capital-intensive companies are precisely where venture returns have historically been largest when they work, and the sector's long famine was arguably a misallocation. It is an argument that a portfolio of fourteen billion-dollar hardware-weighted rounds carries a fundamentally different risk profile than fourteen software rounds of the same size, and that using the same word for both obscures it.
There is a strong case that the broadening is an illusion, and it deserves to be stated properly rather than dismissed.
About $35 billion of July's $65 billion — roughly 53% — went to AI-focused companies. US companies took $39 billion, roughly 59% of the global total. On that arithmetic, more than half the month is a single thesis funded in a single country.
And the apparent sector diversification may be the same thesis wearing different clothes. Semiconductors are AI's inputs. Energy is AI's binding constraint — data centres cannot run without generation and transmission. Robotics is frequently pitched as the embodiment of AI models. On this reading, capital is not diversifying away from AI; it is flowing down AI's own supply chain, which increases correlation rather than reducing it. If the AI capital-expenditure cycle slows, the semiconductor round, the energy round and the robotics round all deteriorate together with the model lab round.
Blue Origin and the German defence rounds are the clearest counter-evidence — those theses stand on their own regardless of what happens to inference demand. But they are a minority of the cohort, and anyone treating July as proof that venture has broadened beyond AI is reading the sector labels rather than the demand drivers.
The most underweighted number in the report is not on the funding side.
July saw more than $9 billion in venture-backed M&A across five billion-dollar-plus exits. Twelve venture-backed companies listed above $1 billion. The largest of those, ChangXin Memory Technologies, debuted at roughly $85 billion, up 466% on its first day.
This matters because the defining pathology of the 2021 cycle was that capital went in and did not come out. Valuations were marked up on paper, the IPO window closed, M&A stalled on antitrust posture, and limited partners spent three years funding capital calls against distributions that never arrived. Paper returns are not returns.
A functioning exit market changes the character of a boom. It recycles capital back to LPs, who can then commit to new funds, which is what makes a run sustainable rather than merely inflationary. It also imposes discipline: public markets reprice private marks, and companies that cannot survive that scrutiny stay private and eventually correct.
The caveat is the ChangXin figure itself. A 466% first-day pop is not a sign of a healthy pricing mechanism — it means the offering was mispriced by a very large margin, and first-day pops of that size have historically clustered in frothy windows rather than orderly ones. A reopened IPO window is good news. A reopened IPO window that prices this badly is a reopened window with a warning attached.
For a founder sizing a raise, the practical read is that the billion-dollar round is now a rung on a ladder rather than a lightning strike, and that the ladder extends into sectors that were unfundable at this scale eighteen months ago. For anyone assessing whether this is a bubble, the diagnostic is not the $65 billion. It is whether the $9 billion of M&A and twelve billion-dollar listings continue — because a boom with exits is a cycle, and a boom without them is an overhang.
Not on its own. The more informative signal is whether capital is coming back out. July 2026 saw over $9 billion in venture-backed M&A across five billion-dollar-plus exits and twelve listings above $1 billion, which distinguishes it from 2021 — when marks rose but distributions to limited partners did not follow. The warning sign in the same data is ChangXin Memory Technologies debuting up 466%, a pop that indicates the offering was badly mispriced.
DrafterDaily reads the venture data the way the people writing it do — including the footnotes.
More Business coverageThe headline was the 94% drawdown from a $3.44 billion peak. The more useful arithmetic is the cash sitting inside the thing being bought — and the fact that everyone except the angels was paid in another private company's paper.
The biography is the least interesting part. What the board actually did was unbundle the CEO job — keeping Tim Cook on with a defined policymaker-engagement remit — and that separation says more about Apple's next decade than the appointment does.
Every outlet reported the size of Meta's teen-safety settlement and the screen-time limits. Almost nobody read the payment structure — where roughly $5.3bn is released only if two competitors act.