AI startups raised $255.5 billion globally in Q1 2026, surpassing the entire 2025 full-year total for AI venture funding, according to PitchBook’s Q1 2026 AI VC Trends report. Three deals accounted for two-thirds of that capital: OpenAI’s $122 billion round, Anthropic’s $30 billion Series G, and xAI’s $20 billion raise. Forbes’ 2026 Midas List analysis, published today by TrueBridge Capital Partners, puts the concentration figure at 81% of all global venture funding flowing to AI in that quarter alone.
The numbers describe a venture capital market that has structurally bifurcated. On one side: a handful of frontier AI companies absorbing capital at a scale that dwarfs entire funding categories. On the other: everyone else.
Nine Out of Ten Are Still Private
The Forbes analysis reveals an unprecedented feature of the 2026 Midas List: nine of the ten companies that contributed most to investor performance remain privately held. Historically, the Midas List weighted realized outcomes like IPOs. This year, unrealized paper gains from private mega-rounds dominate the ranking.
OpenAI closed its $122 billion Series F in March at an $852 billion valuation, according to Forbes. The company reportedly topped $25 billion in annualized revenue by early March and is preparing a June 2026 IPO filing targeting a $1 trillion valuation. SpaceX, after absorbing xAI in an all-stock transaction valued at $1.25 trillion, is targeting a public offering north of $1.75 trillion.
Anthropic raised its $30 billion Series G at a $380 billion valuation after annualized revenue reportedly surged from roughly $1 billion in late 2024 to approximately $30 billion by April 2026, per VentureBeat. The company is reportedly in talks to raise again at a $900 billion valuation. Claude Code, its agentic coding platform, reached $2.5 billion in annualized billings within nine months of launch, according to Sacra.
The Two-Tier Market
The structural effect is stark. Nearly three-quarters of all U.S. venture investment in Q1 flowed into just five deals, per the Forbes analysis. Access to these rounds has become, in Forbes’ framing, “one of venture capital’s most valuable assets.” Sovereign wealth funds, crossover investors, and corporate giants writing multibillion-dollar checks now sit alongside traditional venture firms. The Midas List has become a ranking of who got into the room, not who found the company first.
Meanwhile, the broader startup ecosystem faces tighter conditions. PitchBook notes that while the headline number is record-breaking, the capital concentration is “extraordinarily concentrated at the top.” Remove the three largest deals and Q1 funding looks far more modest.
The Agent Infrastructure Question
For the agentic AI ecosystem specifically, the concentration creates a compounding dynamic. Frontier model development is extraordinarily expensive. Anthropic’s compute bill alone was reported at $1.25 billion per month on SpaceX infrastructure, a figure disclosed in pre-IPO filings. OpenAI has set a $600 billion compute target by 2030.
The infrastructure and tooling startups that agent builders depend on, runtime governance frameworks, evaluation platforms, memory systems, deployment tools, exist in the tier of the market that is not absorbing $30 billion rounds. This week’s launches tell the story: the Agent Control Standard is a volunteer-coordinated open standard; Custom Evals is a lightweight, backend-free framework; Pentest Agent Suite is an open-source security toolkit. These are critical infrastructure projects operating at seed-stage or community-funded economics while the model providers they orbit raise at nation-state scale.
The question is not whether the mega-rounds will produce returns. Public market validation, or correction, is coming: SpaceX, OpenAI, and Anthropic are all preparing for IPOs in 2026. The question is what happens to the middleware layer. Agent frameworks, security tooling, evaluation infrastructure, and governance systems need sustained investment to mature. If capital continues concentrating at the frontier model layer, the ecosystem risks building increasingly capable agents on top of underfunded infrastructure.
Forbes frames the Midas List as a measure of who invested best. The more useful measure for agent builders might be simpler: where is the next dollar going, and is any of it reaching the plumbing?