A Crunchbase report published this week put a number on something that practitioners in the AI space have been feeling for a while: capital has not just flowed into AI — it has flooded into two companies specifically.

Global startup funding hit a record $510 billion in the first half of 2026, already outpacing the $440 billion deployed across all of 2025. That alone would be a remarkable headline. The more striking figure is buried inside it: OpenAI and Anthropic together raised $217 billion in H1 2026. That is 43 percent of every dollar invested in every startup on earth during that period.

Two companies. Forty-three cents of every venture dollar. In six months.

What the Numbers Actually Say

The H1 2026 record was driven by a small number of very large rounds. OpenAI and Anthropic led the pack, but xAI and Waymo also contributed to the mega-round concentration that pushed the headline figure higher. Strip out those four deals and the rest of the market tracked roughly in line with 2024-25 levels.

That pattern matters because it tells you something about the structure of this AI investment cycle. This is not broad-based enthusiasm — it is concentrated conviction around a handful of companies deemed essential to the AI future. Venture funds that missed early positions in the leading labs are now paying frontier valuations to get in before the IPOs.

The public markets opened back up for AI in the first half of 2026, with exits accelerating. That has created a self-reinforcing cycle: capital flows in at record pace, valuations rise, IPO windows open, investors get liquidity, and more capital is available for the next round.

Why This Matters for Businesses Choosing AI Vendors

For businesses evaluating which AI platform to build on, this capital concentration carries practical meaning.

Model availability is not at risk. OpenAI and Anthropic together have enough runway to sustain model development for years. Businesses choosing either as a core platform are not betting on a startup that might disappear — they are betting on what are now effectively AI infrastructure companies. The $217 billion raised in six months is not burning; it is being deployed into compute, talent, and infrastructure at a scale that cements these platforms’ positions.

Pricing pressure is building from the outside. The concentration at the top has drawn challengers. Meituan’s LongCat-2.0, open-sourced under MIT this week and trained entirely on domestic Chinese chips, adds a 1.6-trillion-parameter option to the open-source ecosystem. Meta’s Llama family continues to close the gap with proprietary models. The leading labs know this, which is why pricing on capable mid-tier models has dropped substantially over the past year.

Vendor lock-in risk is real but overstated. The dominant narrative in enterprise AI procurement right now is worry about dependency on one lab. That concern is legitimate — your workflows, fine-tuning, and institutional knowledge will accrue around whichever platform you choose. But the concern is tempered by the fact that model-switching costs are falling as standard interfaces (like the MCP protocol) become more widely adopted.

The gap between labs and everyone else is getting bigger. Strip out the mega-rounds and the rest of the AI startup ecosystem is seeing normal funding levels. That means the companies trying to build point solutions and niche tools are not getting the same gravitational pull. Some will succeed. Many will consolidate or fail. For businesses buying AI tools, the implication is to weight vendor stability more heavily in procurement decisions.

The Open Question: What Comes After the Concentration?

When this much capital concentrates in a few hands, two things tend to happen. First, those companies become infrastructure — not just products, but platforms that the rest of the market builds on. We are already seeing that: more companies build on Claude or GPT-4o via API than compete with them directly.

Second, the concentration invites political attention. OpenAI’s proposal this week to hand the US government a 5% equity stake worth roughly $42 billion is a direct response to this dynamic. When a private company holds that much AI infrastructure and capital, it becomes a policy question, not just a business one.

For enterprise buyers, neither of these outcomes changes the immediate calculus. The platforms you can build on today are better than they were six months ago, cheaper per token than they were a year ago, and backed by more capital than any technology sector has seen in a comparable window. The question for your business is not whether to engage with AI — that window has closed. The question is whether you are building the organisational foundations to capture value from it.

What This Means for Business

The $510 billion headline is a signal, not a prescription. Here is how to read it practically:

Pick platforms with staying power. The record funding makes it less likely, not more likely, that the leading labs will stumble. Build your AI workflows on platforms that have demonstrated staying power, even if that means accepting premium pricing in the short term.

Do not outsource your AI judgment. Capital concentration produces marketing at scale. Every AI vendor in 2026 has money to spend on positioning. Your team’s ability to evaluate claims independently — to ask “does this actually work in our context?” — is worth more than any vendor’s sales pitch.

Invest in your data foundations now. Gartner’s April 2026 research found that organisations with successful AI initiatives invest up to four times more in data quality, governance, and AI-ready infrastructure than those that struggle. The companies raising $217 billion are building the tools. Whether those tools deliver for your business depends almost entirely on what data you give them to work with.

If you’re deciding where to start with agents, start here. The free Working With Claude field guide walks through the ecosystem, Claude Code, and a real rollout plan. Get your copy.