AI venture funding hit a record $510 billion globally in the first half of 2026 — more than all of 2025's $440 billion — and OpenAI and Anthropic alone absorbed $217 billion of it, 43% of every startup dollar invested worldwide. In Q1, four companies took nearly 65% of global venture investment. That is the highest concentration of capital in a single technology sector in venture capital history, exceeding the dot-com peak of Q1 2000.
Which means the sentence "AI startup funding is booming" is true and almost completely useless. For the overwhelming majority of AI startups, 2026 has been a harder fundraising environment, not an easier one, because the record total and the money available to a Series A company are moving in opposite directions.
This article breaks down where the money actually went, what the concentration means if you are not one of four companies, why Europe's share keeps falling, and which categories are still funding normally.
Key Takeaways
- Global startup funding reached a record $510 billion in H1 2026, surpassing the $440 billion invested in all of 2025.
- OpenAI and Anthropic captured $217 billion — 43% of all H1 2026 startup investment between two companies.
- In Q1 2026, OpenAI ($122B), Anthropic ($30B), xAI ($20B) and Waymo ($16B) took $188 billion, nearly 65% of global venture investment.
- AI took roughly 80% of global venture funding in Q1 2026, up from about 55% a year earlier.
- The US absorbed $250 billion, 81% of global funding, and 88% of AI venture dollars — Europe's share fell for a third consecutive half-year.
How concentrated is AI funding in 2026?
More concentrated than any technology sector has ever been, including the internet at the height of the dot-com bubble. In Q1 2026, global venture capital reached $297 billion, and four companies took $188 billion of it.
| Company | Q1 2026 raise | Share of global Q1 VC |
|---|---|---|
| OpenAI | $122B | ~41% |
| Anthropic | $30B | ~10% |
| xAI | $20B | ~7% |
| Waymo | $16B | ~5% |
| Top four combined | $188B | ~65% |
| Everyone else on earth | ~$109B | ~35% |
Sit with the OpenAI line for a moment. A single company raised $122 billion in one quarter — more than the entire global venture market in a typical pre-2020 year. Crunchbase's analysis documented the concentration across three separate measures, all pointing the same direction.
The half-year picture is only marginally less extreme. $510 billion total, $217 billion to OpenAI and Anthropic, meaning 43% of global startup investment went to two companies. AI's overall share of venture funding climbed from roughly 55% in Q1 2025 to about 80% in Q1 2026.
We covered the competitive dynamics between the two largest recipients in Anthropic overtakes OpenAI on revenue and valuation, and one of the four, Waymo, is deep into a capital-intensive physical build-out we analysed in Waymo's 1 million weekly rides.
Is this a bubble?
Concentration this extreme is a genuine warning sign, but it is a structurally different pattern from 2000 — and the difference cuts both ways. The dot-com peak spread capital across hundreds of companies with no revenue. 2026 concentrates it in a handful of companies with enormous revenue and enormous costs.
The bull case is straightforward. OpenAI has a billion weekly active users, as we covered in ChatGPT's 1 billion weekly user milestone. Anthropic overtook OpenAI on revenue growth. These are not concepts with a pitch deck; they are among the fastest-scaling revenue lines in software history. Capital following them is rational.
The bear case is equally straightforward, and it is about what the money buys. A very large share of these raises funds compute and infrastructure — data centers, energy contracts, and silicon — rather than working capital. That spending is depreciating physical plant with a hard delivery schedule, and the schedule is slipping: 30-50% of data center capacity planned for 2026 is expected to arrive in 2028 or later, a problem we detailed in the AI data center power crunch.
The honest synthesis: this is not a bubble in the 2000 sense, where the assets were fictional. It is a bet that demand will arrive on the same timeline as capacity that is already known to be late. That is a real risk with a real mechanism, and it is a more specific worry than "valuations feel high."
What does concentration mean if you are a normal startup?
It means the fundraising market you are actually in is much smaller than the headline suggests, and you should plan against the residual, not the total.
Do the arithmetic from a founder's seat. Global venture in Q1 was $297 billion. Subtract the top four and roughly $109 billion remains for every other startup in every sector in every country — biotech, fintech, climate, enterprise software, and the several thousand AI application companies competing for attention. That residual pool is not obviously larger than it was in 2024, and it is being fought over by a much larger cohort.
Three practical consequences:
- "AI" is no longer a fundraising advantage. When 80% of capital goes to AI, the label is table stakes. Investors are discriminating on revenue quality and defensibility, not category.
- The barbell is real. Capital is available at the very top and at genuine seed, and unusually scarce in the middle. Series B for an AI application company with promising-but-not-exceptional metrics is the hardest raise in this market.
- Infrastructure adjacency is the exception. Capital is chasing compute, agent infrastructure, robotics, defense, and healthcare automation. If you sit next to a bottleneck, you are in a different market than if you sit on top of an API.
That last point has a clean illustration. London's OLIX Computing raised $312 million in Series B at a $3.3 billion valuation for photonic AI inference chips — a company solving the power and cost problem rather than consuming it. Investors funding a bottleneck-solver at that valuation while application companies struggle is the clearest signal in the data.
Why is Europe falling further behind?
Because AI capital concentrates geographically for the same reason it concentrates by company: it follows compute, and compute follows power and existing hyperscale infrastructure — both of which sit disproportionately in the US.
The numbers are stark. The US absorbed $250 billion in H1 2026, 81% of global venture funding, up from 55% a year earlier. Within AI specifically, 88% of venture dollars went to US startups, and Europe's share declined for a third consecutive half-year.
Three compounding factors explain most of it:
- Round size. European venture has never routinely written the multi-billion cheques a frontier lab requires. When the largest rounds define the total, an ecosystem that cannot write them loses share arithmetically even if it funds the same number of companies.
- Compute access. Frontier training requires energized data center capacity. The US has more of it and is building faster, and capital follows the ability to deploy it.
- Exit expectations. Investors price European exits lower, which raises the return threshold, which shrinks the set of fundable companies.
OLIX Computing is the counterexample worth noting — a London company raising a substantial round at a strong valuation, in hardware. Europe's remaining strength is deep tech where the moat is physics and patents rather than compute scale. That is a narrower lane than the US market, and it is a defensible one.
Where the money is going besides the frontier labs
Billion-dollar financings have expanded beyond foundation model developers into a consistent set of adjacent sectors. In rough order of activity:
- Compute and inference infrastructure — chips, power, cooling, and everything that reduces cost per token.
- Agent infrastructure — orchestration, sandboxing, observability, and the tooling layer around autonomous systems.
- Robotics and physical automation — where AI meets a factory floor or a warehouse.
- Defense technology — the fastest-growing category by percentage.
- Healthcare automation — administrative and diagnostic workflows with clear cost baselines.
- Industry-specific software — vertical applications with proprietary data.
The unifying logic is that all six sit adjacent to a bottleneck rather than on top of a commodity. An application built on a public model API has one supplier, no data moat, and a margin that moves whenever that supplier reprices. A company that makes inference cheaper, or safer, or physically embodied, owns something.
Frequently asked questions
How much venture funding did AI startups raise in 2026?
Global startup investment reached a record $510 billion in the first half of 2026, surpassing the $440 billion invested across all of 2025. AI captured roughly 80% of global venture funding in Q1 2026, up from about 55% a year earlier.
Which companies raised the most in 2026?
In Q1 2026, OpenAI raised $122 billion, Anthropic $30 billion, xAI $20 billion, and Waymo $16 billion — $188 billion combined, nearly 65% of all global venture investment that quarter. Across the first half, OpenAI and Anthropic together took $217 billion, or 43% of global startup funding.
Is AI funding concentration worse than the dot-com bubble?
By concentration, yes. The share of venture capital going to a single technology sector in 2026 exceeded the peak concentration in internet companies during Q1 2000. The structural difference is that today's recipients have very large real revenues, where many dot-com era recipients had none.
Is it harder to raise money for an AI startup in 2026?
For most companies, yes. Because a handful of firms absorb the majority of capital, the residual pool available to everyone else has not grown proportionally with the headline total. Seed and frontier-scale rounds are well funded; mid-stage rounds for AI application companies are the hardest part of the market.
Why is Europe losing AI funding share?
Europe's share of AI venture dollars fell for a third consecutive half-year, with 88% now going to US startups. The main drivers are round size — European funds rarely write multi-billion cheques — plus US advantages in energized data center capacity and higher expected exit valuations.
What AI sectors are still funding well outside the frontier labs?
Compute and inference infrastructure, agent infrastructure, robotics and physical automation, defense technology, healthcare automation, and vertical industry software. The common thread is proximity to a bottleneck rather than dependence on a commodity model API.
The verdict
The record $510 billion is real, and it describes a market almost nobody is actually participating in. Two companies took 43% of it. Four took 65% of the quarter. If you are raising money for anything else, the relevant number is the residual — and the residual has barely moved.
Plan against that. Being an "AI company" buys nothing in 2026; being adjacent to a bottleneck buys a lot. The investors funding photonic inference chips at a $3.3 billion valuation while application-layer Series Bs stall are telling you exactly where they think the durable value sits.
The AI funding boom is not a rising tide. It is a very tall wave, in a very small place.
Top comments (0)