AI megarounds, private-market liquidity, and contracted compute reshape venture capital and growth financing

AI Mega-Rounds Concentrate Venture Capital Around Category Leaders

This week’s largest U.S. startup financings were dominated by AI, with Instinct raising $1.0 billion, Temporal Technologies $550 million, EliseAI $350 million, GMI Cloud $223 million, General Intuition $220 million, Factory $200 million, Arena $200 million, and Profound $180 million — about $2.7 billion in disclosed capital across that cohort. The concentration reached the top of the market as OpenAI reportedly closed a $122 billion financing on March 31, 2026 at an $852 billion post-money valuation, with Amazon, Nvidia, and SoftBank as core backers; Bloomberg said the syndicate included $50 billion from Amazon, $30 billion from Nvidia, and $30 billion from SoftBank, plus more than $3 billion from individuals and other investors.

The signal is a barbell shift in venture capital: capital is flowing to a small number of AI leaders that can absorb massive rounds, rather than being spread across many smaller bets. That concentration is visible beyond applications, extending into the infrastructure stack where Broadcom’s reported roughly 75% share of the AI ASIC market and its near-duopoly with Marvell in custom AI ASIC co-design point to spending clustering around the hardware layer too. For operators and vendors, the bar now favors category leadership and direct relevance to frontier AI ecosystems; for investors, access, scale, and infrastructure adjacency are becoming the main sources of advantage.

Private-Market Liquidity Moves Into Retirement and Secondary Infrastructure

August 7, 2025 marked a policy break for private markets: an Executive Order told the DOL and SEC to remove barriers to “alternative assets” in retirement plans, the DOL then withdrew its December 2021 guidance discouraging private-market exposure, and in March/April 2026 it proposed a fiduciary safe harbor to reduce sponsor litigation risk through principles-based review of performance, fees, liquidity, valuation, benchmarks, and complexity. The language explicitly covers private equity and, in some descriptions, venture capital.

At the same time, Ondo’s 24/7 private-notes market shows how liquidity is being productized outside the fund structure: eligible non-U.S. investors can trade tokenized notes linked to private-company common stock on an onchain secondary market, with pricing set by buyers and sellers rather than the last financing round. The strategic implication is clear: retirement-plan access could expand the buyer base for private assets while secondary rails make those positions more tradable, reinforcing demand for pricing, transfer, and liquidity infrastructure. The US VC secondary market was projected to reach $94.9 billion by Q3 2025, and GP-led transactions already represented 8% of secondary activity.

OpenAI’s Amazon Expansion Pushes Compute Into Contracted Capacity

OpenAI’s reported expansion with Amazon, cited at up to roughly $50 billion of support plus broader compute access and additional NVIDIA capacity tied to Vera Rubin systems, pushed the market further from component procurement toward contracted capacity platforms. HPE’s $1.2 billion Vultr deal points the same way from the supply side: AMD Helios AI Rack deployments are being sold with networking, liquid cooling, and rollout across Vultr’s U.S. data centers, not as standalone silicon. Reported commitments around Tencent, Anthropic, and Nebius reinforce the shift to multi-year cloud and data-center capacity instead of one-off GPU purchases.

The change is not just deal size; it is what can now be financed. Once compute is bundled with facilities, cooling, operations, and long-duration customer commitments, it starts to resemble infrastructure revenue. That matters because the binding constraint is increasingly power and grid access: research cited this week points to long interconnection queues, 72% of developers calling grid capacity very or extremely challenging, and Morgan Stanley estimating a roughly 32 GW U.S. power shortfall through 2028 even after mitigation. Building on last week’s financing structures, site selection, PPAs, and utilization are now product decisions. For vendors and investors, value is concentrating in integrated capacity contracts and the ability to underwrite power, financing durability, and delivery risk.

Growth Capital Is Concentrating in AI and Cross-Border Hard Tech

Q3 2026 growth-stage funding in North America fell sharply, but the drop was driven by the absence of a few AI megarounds, not a broad collapse in venture demand. Roughly two-thirds of total Q3 funding went to AI-focused companies, and PitchBook-NVCA said AI represented 82.7% of deal value; most of the decline came from venture-growth rounds rather than early-stage activity.

India moved in the opposite direction. Growth-stage capital surged across AI, cleantech/EV, and deeptech/spacetech, led by Emergent’s $130 million round, River Mobility’s $120 million, Navi’s $100 million, Pixxel’s $100 million, Yulu’s $93 million, and Ultraviolette’s $85 million. Sector totals were similarly concentrated: AI drew $438 million across 35 deals, cleantech/EV $433 million across 23 deals, and deeptech/spacetech $290 million. The largest checks came from a mix of cross-border and domestic investors, including SoftBank, Khosla Ventures, Temasek, Radical Ventures, Battery Ventures, Prosus, GEF Capital Partners, Qualcomm Ventures, Alpha Wave, Creaegis, Lightspeed, Nexus Venture Partners, Accel, 360 ONE Asset, Anicut Capital, Alteria Capital, and InnoVen Capital.

The strategic read is clear: late-stage capital is becoming more selective, more global, and more concentrated in AI and hard-tech winners. For investors, edge now comes from cross-border sourcing and sector-specific underwriting; for operators and vendors, fundraising and enterprise sales will be easier in favored categories and materially tighter elsewhere.