Venture capital's AI investment spree has reached a fever pitch, with the second quarter of 2026 generating the most billion-dollar startup exits since the 2021 market peak, according to Crunchbase data. This momentum reflects genuine value creation in the AI infrastructure space, exemplified by strategic acquisitions like Qualcomm's purchase of Modular—a transaction that Dave Munichiello, managing partner at Google Ventures (GV), argues signals a critical industry shift toward AI software platforms as essential infrastructure rather than commodity tools. The deal underscores how technology giants are attempting to manage soaring computational costs and complexity by consolidating software layers that optimize hardware efficiency. Yet beneath this headline-grabbing activity lies a troubling reality: the same venture capital firms fueling these mega-rounds are relying on inadequate data infrastructure to source and evaluate opportunities.
Henrik Landgren, founder of Gilion, has identified a systematic blind spot in how venture capital firms approach due diligence and startup identification. Rather than connecting directly to authoritative sources like financial systems, payment platforms, and accounting records, VCs are increasingly deploying AI against poorly structured, secondhand data—creating a feedback loop that reinforces existing biases and causes investors to systematically overlook promising startups. This data problem disproportionately affects founders outside traditional venture networks. In a parallel investigation, six startup investors acknowledged that rethinking sourcing patterns and broadening network effects remains the primary barrier to funding Black-founded startups, suggesting that systemic data gaps extend beyond infrastructure into human capital sourcing. Landgren's thesis directly contradicts the venture industry's current posture: firms are spending billions on AI-powered sourcing tools while simultaneously weakening their ability to identify overlooked opportunities through inadequate foundational data.
The concentration of capital into mega-deals—with AI driving "another spree of megadeals" across the sector—masks a fragmentation problem at the market's edges. Landgren argues that better data infrastructure could democratize deal sourcing and surface overlooked startups that current venture methodologies dismiss as immeasurable risks. Without direct connections to financial and operational data sources, VCs remain dependent on founder networks, warm introductions, and pattern-matching against historical successes, systematically excluding underrepresented founders and non-obvious opportunities. As Munichiello's comments on Modular suggest, sophisticated investors recognize that infrastructure consolidation creates durable value—yet most venture capital remains trapped in outdated sourcing methodologies. The paradox is stark: the industry investing most heavily in AI is simultaneously most vulnerable to missing transformative opportunities simply because it lacks the data hygiene required to see them.