SVB sees a polarised market: large rounds at the top and widespread closures at the bottom

Two opposing markets at once

At the top, the venture market looks overheated: individual AI companies are raising hundreds of millions of dollars at multibillion-dollar valuations. At the bottom, conditions remain harsh: SVB expects 2,345 venture-backed companies to close in 2026, with more than a third coming from startups founded during the zero-interest-rate era. These developments do not contradict each other: capital is no longer broadly distributed and has concentrated in a small number of winners. [1 · SVB]

The strongest demand is for companies that have demonstrated access to enterprise workflows, a proprietary data feedback loop or physical infrastructure. For others, investors assess not only the potential market size but also the number of months until the next objective milestone: repeatable sales, production deployment, an audit-ready security system or proven unit economics for a sale or operation. [1 · SVB]

AI claims need verification

SVB also reports that, among 9,000 venture-backed companies analysed that use AI terminology in their descriptions, 42% had a weak connection between AI and their core technology. The methodology is an analytical assessment by SVB/PitchBook, rather than a technical audit of each company. The figure therefore cannot be interpreted as a proven share of sham AI startups. [1 · SVB]

Nevertheless, the report points to a change in investment due diligence. AI vocabulary is not enough for investors and enterprise buyers: they want to see the share of tasks completed without manual rework, time to the first production result, activity in mature cohorts, cost per accepted action, data rights and dependence on a particular model provider. [1 · SVB]

What a polarised market means

Large rounds and high closure rates can continue simultaneously. At one end of the market are companies with proven execution and access to a scarce layer of technology or infrastructure; at the other are many projects without sufficient revenue or their next financing round. The report reflects data as of 30 June 2026 and should be treated as an indicative market signal, rather than a precise forecast of each company's fate. [1 · SVB]