— HAWK EYE • ISSUE 1 • JULY 13, 2026
The 81% Illusion
The most quoted number in venture capital this year is technically accurate and practically misleading. Founders are making raise decisions on the incomplete reading of it. Here is the correction, with the data.
01
Next ⟶
Coming Soon
$188 bn
of Q1 2026 venture capital went to just four companies (Cunchbase. Q1 2026)
73.1%
of Q1 VC fundraising captured by only five fund managers (Pitchbook, Q2 2026)
43%
of all H1 2026 global venture funding absorbed by two AI labs (Crunchbase News, H1 2026)
You have seen the statistic. In the first quarter of 2026, 81% of all global venture capital went to artificial intelligence companies. Crunchbase counted roughly $297bn deployed across some 6,000 startups, the largest venture quarter ever recorded, and about $242bn of it carried an AI label. Commentators have called it the highest capital concentration in a single technology sector in the history of venture investing, exceeding internet companies at the peak of the dot-com bubble in early 2000.
Read one way, that number says: build anything with AI in the name and the capital will find you. Read the other way, it says: if you are not an AI company, four in five venture dollars are closed to you. Both readings are circulating in founder communities right now. Both are wrong, and the error is not small. It changes what you should build, how you should price your round, and which investors you should be speaking to.
01 Where the money actually went
Decompose the quarter and the story changes shape entirely. Of the $297bn raised globally in Q1, four rounds accounted for roughly $188bn: OpenAI at $122bn, Anthropic at $30bn, xAI at $20bn and Waymo at $16bn. That is 63% of every venture dollar on the planet, in three months, into four names. The pattern held into the second quarter. As Crunchbase News reported, OpenAI and Anthropic alone absorbed $217bn of the $510bn raised worldwide in the first half, roughly 43% of everything.
Figure 1. Global venture capital, Q1 2026, $297bn total. The "81% to AI" headline conceals a 63/18 split within it. Source: Crunchbase, TechCrunch, PitchBook, Q1–Q2 2026.
Strip out the four mega-rounds and the residual AI pool, spread across thousands of AI-labelled startups, is roughly $54bn, or 18% of the quarter. PitchBook data reported via SiliconANGLE puts the funding outside the five largest deals at about $72bn across roughly 4,600 deals, and describes it as broadly consistent with a normal venture quarter. The mega-rounds are a parallel universe, not a rising tide.
"Five fund managers captured 73.1% of first-quarter fundraising. That is not a broad market. That is a small number of gatekeepers."
PitchBook Q2 2026 analyst note, via Angel Investors Network
02 The Long tail is where the damage is concentrating
Your instinct might be that the thousands of small AI companies splitting that $54bn are the next generation of category winners. The operating data says most of them are something else: thin software layers over somebody else's model, holding neither the asset nor the customer.
The market has a name for them. An AI wrapper is a product whose core function is to call a foundation model API and format the output. Industry aggregates put 60–70% of AI wrapper startups at zero revenue, with only 3–5% ever reaching $10k in monthly recurring revenue. RevenueCat's State of Subscription Apps 2026 report, drawn from a platform managing over $11bn in annual subscription revenue, found that users cancel AI-powered app subscriptions roughly 30% faster than non-AI apps, with annual retention of 21.1% against 30.7%. ChartMogul's retention data is harsher still: AI-native products priced under $50 per month show gross revenue retention around 23%.
Figure 2. Retention economics, AI-native applications against traditional SaaS. Sources: RevenueCat State of Subscription Apps 2026; ChartMogul SaaS Retention Report; industry churn aggregates.
Enterprise demand is not rescuing these companies either. MIT's widely circulated study, "The GenAI Divide: State of AI in Business 2025", found that 95% of enterprise generative AI pilots showed no measurable P&L impact despite $30–40bn of spending. We would flag, as few who quote it do, that the methodology has been credibly challenged: the headline rests on 52 interviews and an undisclosed synthesis of 300 public deployments, and it excludes efficiency gains that never reach a P&L line. Treat it as directional, not gospel. Directionally, though, it agrees with everything else: Gartner reports over 40% of agentic AI projects stalling, and a February 2026 NBER study found 90% of firms reporting no current productivity impact from AI at all.
The platforms themselves are saying the quiet part aloud. Google Cloud's Darren Mowry told TechCrunch's Equity podcast in February that companies counting on the back-end model to do all the work are exposed. The proof arrived the same month: when Anthropic published a blog post on using Claude Code to modernise COBOL systems, IBM's stock fell 13% in a single session. One blog post from a model lab repriced a Fortune 50 consulting franchise overnight. Now ask what the same post does to a twelve-person startup whose entire product is a thinner version of that capability.
"If your entire product is one clever prompt over someone else's model, you are one feature release away from irrelevance."
The consensus now forming across Google Cloud, Sequoia and the churn data
03 The backdrop: a $600bn question nobody has answered
All of this sits inside a capital cycle that is itself under strain. Sequoia Capital's David Cahn posed what he called "AI's $600B Question" in 2024: the annual revenue the AI industry needs to generate to justify its infrastructure spending. He first ran the numbers in September 2023 and got $200bn, at a time when OpenAI's annualised revenue was $1.6bn. Nine months later the figure had tripled. By July 2026, with infrastructure spend for the year estimated at $1.5tn, TechCrunch reports Cahn's requirement at $3tn, in his own words likely an underestimate. JPMorgan's parallel calculation puts the revenue needed for a bare 10% return on current AI infrastructure at $650bn a year. The combined run rate of the major model providers is roughly $75bn.
Figure 3. Revenue required to justify AI infrastructure spend against actual industry revenue. Sources: David Cahn, Sequoia Capital ("AI's $200B Question", Sept 2023; "AI's $600B Question", June 2024); TechCrunch, "Can AI answer the $3 trillion question?", July 2026.
Allianz Research calculates the divergence between AI capital expenditure and revenue growth at roughly 46%, already wider than the 32% recorded during the 2001 telecom overbuild. Ray Dalio told Bloomberg at the Forbes Iconoclast Summit in June that the AI bubble will burst eventually, and Bridgewater's founder is not alone; Ed Zitron argued on Bloomberg Radio the same week that AI companies which have never reported a profit should not be permitted to IPO. Meanwhile the cost side is worsening: the Iran conflict removed roughly 30% of semiconductor-grade helium supply, DRAM and HBM memory prices nearly doubled in Q1 2026 per Counterpoint Research, and Goldman Sachs tracked circuit board prices up as much as 40% in April alone. Fortune reported in April that big tech has announced $740bn of capital expenditure this year, up 69%, even as Nvidia's own VP of applied deep learning conceded that for his team the cost of compute exceeds the cost of the employees.
We do not think the lesson here is "AI is a bubble, stay away". The dot-com correction destroyed the tourists and left the infrastructure for Amazon and Google. The lesson is narrower and more useful: capital is concentrated because the asset that matters is scarce, and that asset is not access to a model. Everyone has access to a model. The scarce asset is a defensible position a model release cannot erase.
"The divergence between AI capital expenditure and revenue growth now exceeds the 2001 telecom cycle. The market has started to notice."
From Allianz Research, via Forbes, June 2026
04 The moat test
Which brings us to the practical question, the one an investment committee will ask about your company whether or not you have prepared an answer. We put every AI-adjacent business through three questions before we let a client take it to investors.
Three questions before your next raise
One. Does your product accumulate proprietary data with usage, data a competitor cannot buy, scrape or reconstruct? Stored chat logs do not count; the model labs hold more conversation data than you ever will.
Two. Are you embedded in a workflow with genuine switching costs, a system of record, an approval chain, a compliance layer? Or are you a chat window a customer can close?
Three. Do you own distribution into a niche that a horizontal platform cannot justify chasing?
Zero of three is a feature awaiting absorption. One is a speed bump. Two or more is a company. Private-market data supports the arithmetic: vertical AI businesses with genuine data assets or workflow depth are still raising at 15–30x ARR in 2026, while thin wrappers struggle to raise at all.
05 The Hawk Eye
First, we expect the AI label to become a pricing liability for undifferentiated companies within twelve months. The valuation premium for seed-stage AI, recently measured at around 42% over non-AI peers, is a premium on a word. As wrapper churn data normalises across investment committees, that premium inverts: an AI claim without a moat answer will price below an equivalent non-AI business, because the diligence cost of separating substance from costume gets charged to the founder. The repricing has already begun in traditional software, where forward revenue multiples have compressed from roughly 7x to 3.3x.
Second, the binding constraint for everyone outside the frontier is shifting from capital to distribution, and fundraising strategy should be rebuilt around that. Sixteen companies took 53% of everything raised in Q2. For the several thousand companies splitting the remainder, the differentiating asset in diligence is no longer the model, the demo, or the team slide. It is evidence of one repeatable acquisition channel and revenue that survives a frontier lab's next release. Founders who spend the next two quarters manufacturing that evidence will raise; founders who spend them polishing the deck will not.
Where does your business sit in this cycle?
The moat test above takes three minutes and most founders fail it on the first pass, which is precisely the right time to find out, before an investor does it for you. Our business health check scores your company across capital readiness, structure, and defensibility, and returns a written diagnostic.
Next ⟶
Coming Soon
Sources. Crunchbase / TechCrunch, Q1 2026 global venture data; Crunchbase News, H1 2026 funding review; PitchBook Q2 2026 analyst note via Angel Investors Network; SiliconANGLE, Q1 2026 deal analysis; David Cahn, Sequoia Capital, "AI's $200B Question" (2023) and "AI's $600B Question" (2024); TechCrunch, "Can AI answer the $3 trillion question?" (July 2026); JPMorgan infrastructure return analysis; Forbes, "AI Costs More Than The People It Replaced" (July 2026); Forbes, "The AI Capex-To-Revenue Gap Is Widening" (June 2026), citing Allianz Research; Fortune, "The cost of compute is far beyond the costs of the employees" (April 2026), citing Morgan Stanley; MIT NANDA, "The GenAI Divide: State of AI in Business 2025", with methodology caveats per Marketing AI Institute; RevenueCat, State of Subscription Apps 2026; ChartMogul, SaaS Retention Report; Gartner agentic AI project data; NBER productivity study (February 2026); Bloomberg Radio interviews with Ray Dalio and Ed Zitron (June 2026); CNBC and Counterpoint Research, semiconductor supply chain coverage (March–May 2026); Goldman Sachs component price tracking; TechCrunch Equity podcast (February 2026).
BlackHawk Capital Advisors Ltd is a UK-incorporated strategic finance advisory firm. This publication is general commentary, not investment advice, and BlackHawk does not arrange or broker securities transactions. Figures are drawn from the public sources listed and were current at the time of writing.
BlackHawk Capital Advisors is a boutique management consulting firm, helping businesses solve complex strategic, financial, operational, and growth challenges.
EXPLORE
Strategy. Structure. Scale.
CONTACT
First Floor Office,
3 Hornton Place,
London, W8 4LZ
United Kingdom.
+44 7887 159089
© 2026 BlackHawk Capital Advisors Ltd. All rights reserved.