AI Funding Hits $510B in Six Months, Reshaping Venture Capital

Global venture capital investment reached $510 billion in the first half of 2026, a record for any six-month period in the asset class's history, according to Crunchbase data published on July 2. More than $200 billion of that total landed in the second quarter alone, making Q2 the second-largest quarter on record. The force driving the surge is not diffuse. Eighty percent of total global venture funding in the first quarter flowed into artificial intelligence startups, CRN reported, citing Crunchbase figures. The money is accumulating from Series A through pre-IPO at a scale that has no precedent in the software era. What is less discussed is what the lead investors are getting in return, and who gets squeezed when the music stops.

The distribution of that capital tells a more textured story than the topline suggests. It is not merely the frontier-model labs hoovering up billion-dollar cheques. A broad class of companies spanning sovereign AI, private-infrastructure deployment, agentic coding, and custom silicon is pulling in rounds that would have been outlier events two years ago. Indian startup Sarvam closed a $234 million Series B at a $1.5 billion valuation in June, TechCrunch reported, led by HCLTech with a $150 million commitment. Together AI raised $800 million in a Series C at an $8 billion valuation, led by Aramco Ventures, The Next Web noted, as demand for open-source AI compute infrastructure continued to outstrip supply. These are not anomaly rounds. They are the new baseline.

The Series A end of the pipeline is heating up in ways that suggest the boom has another gear. Swiss startup Prem AI is raising $100 million in a Series A, Bloomberg reported in June. The company, which helps hedge funds and law firms run AI models on their own private infrastructure, sits at the intersection of two of the hottest sub-themes in enterprise AI: sovereignty and security. A $100 million Series A for a company that does not build foundation models is a signal. It means investors are underwriting the picks-and-shovels thesis with chequebooks that used to be reserved for the model layer itself. The question is whether the revenue multiples implied by these early-stage prices can be sustained when the companies face their Series B and Series C marks.

The semiconductor category is a parallel proof point. So far in 2026, investors have poured roughly $10 billion into seed-through-pre-IPO rounds for chip startups, Crunchbase News reported in a sector snapshot on June 10. Companies like Cerebras, which filed confidentially for an IPO earlier this year, and MatX, which is designing chips purpose-built for transformer workloads, are riding a thesis that Nvidia's dominance creates an addressable market for challengers, not just a moat. The capital is not cheap. These rounds are pricing in revenue trajectories that assume hyperscaler capex budgets will keep expanding through the end of the decade. If cloud providers tighten spending in a downturn, the semiconductor startups that raised at 50-times forward revenue will be the first to reprice.

At the growth stage, the numbers get harder to anchor in fundamentals. Cognition AI, the maker of the autonomous coding agent Devin, raised $1 billion at a $26 billion valuation in May, according to CRN's ranking of the hottest AI startups of 2026. CEO Scott Wu said more than 90 percent of the company's internal code is now written by Devin itself, a statistic that functions as both a product demo and a pitch to enterprises wondering whether agentic coding tools can replace headcount. The round's structure was not disclosed, but a $26 billion post-money for a company whose primary product is a developer tool implies the lead is betting on a market-size expansion that takes Devin from tens of thousands of users to a platform powering a material fraction of global software production. That is a big bet on a single product category.

The infrastructure layer is attracting a different class of capital altogether. Helix Digital, the AI infrastructure startup unveiled by former AWS CEO Adam Selipsky, launched with $20 billion in committed capital from a consortium that includes KKR, Nvidia, Vistra, and the Kuwait Investment Authority. "The AI era will require a new generation of infrastructure. We're building it," Selipsky said, according to CRN. The startup's proposition is to integrate data centres, power generation, and connectivity into a single delivery model for hyperscalers. Nvidia will deploy its DSX AI factory infrastructure inside Helix's facilities. The presence of Vistra, a utility-scale power provider operating 44,000 megawatts including nuclear, signals that the binding constraint in AI infrastructure is no longer just the chips. It is the grid.

Cohere, the Canadian LLM company, took a different path to scale. In April it acquired and merged with Germany's Aleph Alpha, creating a combined entity valued at roughly $20 billion, CRN reported. The merger was designed to combine Cohere's enterprise AI distribution with Aleph Alpha's European sovereign deployment capabilities. In May, Cohere doubled down on the strategy by acquiring Reliant AI to deepen its sovereign cloud credentials. The play is a bet that the market for foundation models will fragment along geopolitical lines, and that a company explicitly positioned as the non-American, non-Chinese alternative can command a valuation premium that pure-play LLM startups cannot. It is also a bet that European regulators will continue to tilt the playing field in favour of local providers.

Mistral AI, the French foundation-model company, raised $830 million earlier this year for a data centre equipped with 13,800 Nvidia GPUs, CRN reported. The round was structured around physical infrastructure rather than model development, a pattern that is becoming standard for the model-layer companies. The capital does not flow primarily to research headcount. It goes to compute capacity, which means the investors are underwriting hard assets with a depreciation schedule, not just intellectual property with an uncertain monetisation timeline. That shift changes the risk profile of these rounds in ways that traditional venture frameworks are still catching up to.

The exit environment has finally unfrozen in a way that validates the marks. Crunchbase data showed that IPOs and acquisitions returned in force during the second quarter, with venture-backed exits notching one of the strongest periods in years. For limited partners, the return of distributions changes the allocation calculus. After two years of net-negative cash flows, the venture asset class is once again returning capital, enabling fund-of-funds managers to re-up into new vintages. That in turn has allowed GPs to raise megafunds at a pace that would have been unthinkable in 2023. The risk, familiar to anyone who lived through the 2021 cycle, is that the distributions are concentrated in a handful of AI names, and that a correction in the sector would freeze the fundraising environment for the rest of the industry.

The Trillion-Dollar Question

No company embodies the scale and the contradictions of this cycle more acutely than Anthropic. The maker of the Claude family of AI models confidentially filed its S-1 registration statement with the SEC on June 1, 2026, Goodreturns reported, setting the stage for an October Nasdaq debut that could cross the trillion-dollar mark on its first day of trading. The filing came days after Anthropic closed a $65 billion Series H round at a $965 billion post-money valuation, led by Altimeter, Dragoneer, and Sequoia Capital, with institutional investors including Baillie Gifford, Blackstone, and Fidelity also committing. Goldman Sachs, JPMorgan, and Morgan Stanley are the lead underwriters. If the offering raises more than $60 billion, as expected, it would be the largest technology IPO since Facebook's 2012 listing.

The revenue trajectory is what the underwriters are selling. Anthropic's annualized revenue crossed $47 billion by May 2026, up from $1 billion in December 2024, CRN confirmed. No software company in history has grown at that pace over a comparable period. The engine is Claude Code, the company's agentic coding product, which reached $2.5 billion in annualized revenue within nine months of launch and now accounts for an estimated 4 percent of all public GitHub commits. Claude is the first frontier model available on all three of the world's largest cloud platforms: AWS, Google Cloud, and Microsoft Azure. Eight of the Fortune 10 are Claude customers. The company has been acquiring strategically, picking up Stainless in May, Coefficient Bio in April, and Bun in December, among others, expanding both its technical capabilities and its surface area for enterprise distribution.

It will help us cure cancer. It may help us eradicate tropical diseases. It will help us understand the universe. But there are these immense and grave risks., Dario Amodei, Founder and CEO of Anthropic, at the World Economic Forum, Davos 2026

Anthropic's public narrative rests on a paradox that its S-1 will have to reconcile. The company was founded in 2021 as a Public Benefit Corporation by Dario Amodei and six other former OpenAI researchers who believed their previous employer was moving too fast on commercialization at the expense of safety. The founding team raised $124 million at a $623 million valuation, modest by the standards of what followed. Today, Anthropic is simultaneously the most commercially successful AI safety company in history and the subject of intensifying criticism from researchers who argue its safety commitments are being eroded by the imperatives of a public listing. The company has formally dropped a conditional pause commitment, under which it had pledged to halt development if its own models became too dangerous, and has opened conversations around defense contracts, Goodreturns reported.

The Series H pricing at $965 billion implies a set of assumptions that deserve scrutiny. At $47 billion in annualized revenue, the round valued Anthropic at roughly 20.5 times forward revenue. That multiple is not unreasonable by the standards of high-growth software companies, but Anthropic is not a software company in the traditional sense. Its cost structure is dominated by compute infrastructure, which the S-1 is expected to disclose at roughly $19 billion in annualized commitments across Amazon, Google, and other providers. The company lost $5.6 billion in 2024, according to Goodreturns. In a market where interest rates remain elevated, investors are betting that revenue growth will outrun infrastructure costs faster than any comparable company has managed. If it does not, the post-IPO shareholders are the ones who absorb the compression.

Anthropic's IPO is not a standalone event. It is the capstone of a cycle in which the Series A through pre-IPO pipeline has collectively absorbed more than half a trillion dollars in six months, with AI and adjacent infrastructure capturing the overwhelming majority. The exits that the Crunchbase data show returning in Q2, including a strong pipeline of semiconductor IPOs from Cerebras and others, are validating the marks that late-stage investors have been carrying on their books since the 2021 vintage. But the speed of the repricing is itself a risk. When rounds move from $100 million Series As to $65 billion pre-IPO rounds within the same calendar year, the distance between a mark and a realised return can collapse faster than any distribution waterfall can accommodate. LPs who were patient through the drought of 2023 and 2024 are about to find out whether the returns are as good as the marks suggested.

The second half of 2026 will test whether the public markets can absorb what the private markets have produced. Anthropic's October listing is the main event, but the undercard is crowded: semiconductor IPOs, European sovereign AI plays, and a wave of agentic-software companies will all be jostling for allocations from the same institutional buyers. The $510 billion that went into startups in the first half has created a cohort of companies that need to deliver public-market returns, and soon. For a venture industry that spent 2023 and 2024 in a liquidity drought, the flood of 2026 is welcome. The question nobody can answer yet is whether the buyers at the other end of the trade will agree with the prices the VCs have already set. The roadshow decks are printed. The bookbuild begins in September.