OpenAI and Anthropic absorbed over 40% of global venture funding in H1 2026

Global venture capital investment reached $510 billion in the first half of 2026, a record that surpasses the entire $440 billion invested across all of 2025. The headline number suggests a booming startup ecosystem. The breakdown tells a different story: OpenAI and Anthropic alone accounted for more than 40% of that total, a concentration of capital in two companies that has no real precedent in venture capital history.

The scale is worth stating plainly. OpenAI closed a $122 billion round in March 2026, led by Amazon ($50 billion), Nvidia ($30 billion), and SoftBank ($30 billion), pushing its valuation to $852 billion. Anthropic followed close behind, raising a round that pushed its valuation toward $965 billion — near enough to a trillion dollars that the distinction barely matters for what it signals about investor conviction. Together with xAI's $20 billion round and Waymo's $16 billion raise, four companies absorbed $188 billion in a single quarter, or roughly 65% of all global venture investment in that period.

What this concentration actually does to the rest of the market

The obvious question is whether this is a healthy market with an unusually large winner, or a market where capital allocation has stopped functioning as a discovery mechanism for the next generation of important companies. The early evidence points toward the second answer, at least in two specific places.

First, second-tier foundation model labs — companies like Reka and AI21 that were building genuinely competitive models eighteen months ago — have reportedly seen institutional check sizes shrink as limited partners concentrate their AI exposure into the three or four labs with the clearest path to dominance. When an LP can get AI exposure by writing one check to a fund with an OpenAI or Anthropic position, the case for taking a bet on a smaller, higher-risk lab weakens considerably. This isn't a judgment on those companies' technology — it's a structural effect of capital concentration that has nothing to do with product quality.

Second, application-layer AI startups — the companies building products on top of frontier models rather than training their own — are finding it harder to raise at premium multiples specifically because so much capital is locked up at the infrastructure layer. Industry analysts expect at least three significant M&A transactions in the $1 billion to $5 billion range during 2026 as application-layer companies that can't raise their next round get acquired by better-capitalized competitors instead of continuing as independent companies.

The part of the story that is not collapse

It's worth resisting the framing that non-AI startups are being starved entirely. Roughly $58 billion went to non-AI startups in Q1 2026 alone — a figure that would have been the largest venture quarter in history before 2018. In absolute terms, capital available to good non-AI companies remains substantial. The honest read is not that the rest of the startup ecosystem is dying; it's that the bar for what counts as fundable has quietly moved, and it moved without most founders outside the frontier AI space getting the memo.

There's also a geographic dimension worth noting: U.S.-based companies captured 83% of global venture capital in Q1 2026, up sharply from 71% a year earlier. That's a second concentration effect stacking on top of the AI-company concentration — capital isn't just piling into a few companies, it's piling into a few companies in one country, which has implications for where the next generation of globally competitive tech companies gets built.

What founders outside the frontier-AI tier should actually do with this

If you're raising outside the top four AI labs right now, the practical implication is that generic "AI-powered" positioning is no longer sufficient to access premium valuations — that framing worked when AI capital was seeking any credible home; it stops working when AI capital has three or four obvious homes that dwarf everything else. What still raises well: genuine data or distribution moats unrelated to model quality, vertical-specific applications where domain expertise matters more than raw model capability, and infrastructure plays that remain useful regardless of which frontier lab wins. Founders should also expect slower fundraising timelines outside the AI tier as LPs and funds spend more diligence cycles per non-frontier check — the money hasn't disappeared, but the attention allocated to finding it has.