Adit Singh joins Mayfield to lead $20 million seed infrastructure rounds

The appointment of Adit Singh signals a shift toward capital-intensive seed rounds for the underlying hardware powering streaming and AI workloads. As infrastructure costs for large-scale video processing and physical AI rise, Mayfield is positioning itself to lead mega-seed rounds that smaller firms cannot afford. This move reinforces the trend of venture capital firms seeking deep technical expertise to vet complex semiconductor and data center technologies. The industry should monitor Mayfield's deployment of its $3 billion fund into startups like Lumilens and Upscale AI to see if these high-valuation seed entries yield sustainable returns in the competitive infrastructure market.

Cerebras, the AI chip startup that drew Singh's earliest institutional backing, has continued to scale its commercial footprint in the year leading up to his departure from Foundation Capital. The company's wafer-scale engine architecture targets training and inference workloads that underpin large-scale video generation and streaming recommendation systems, positioning it as a direct competitor to Nvidia in data center AI. Ericsson's Mobility Report found that gen AI traffic currently represents only 0.06% of total mobile network data but is expected to grow substantially as AI agents embed across devices, underscoring the infrastructure demand that firms like Mayfield are racing to fund at the seed stage.

Mayfield's $3 billion in assets under management places it among the largest early-stage firms targeting infrastructure, and Singh's mandate to lead $20 million seed rounds reflects a broader venture capital trend toward capital-intensive bets on physical AI and semiconductor startups. Ericsson's June 2025 Mobility Report projected that 5G FWA will contribute over 35% of new fixed broadband connections by 2030, with global connections expected to reach 350 million, a dynamic that increases the strategic value of U.S.-based venture firms funding the chip and networking layer. Singh's portfolio companies Lumilens and Upscale AI both operate at the intersection of optical hardware and AI processing, areas where streaming video workloads are increasingly demanding specialized silicon beyond general-purpose GPUs.

The technical thesis behind Singh's move aligns with measurable shifts in network traffic patterns driven by AI. Ericsson's research showed that AI traffic carries a 26% uplink ratio compared to the typical 10% for conventional mobile video, and that ChatGPT alone accounts for 60% of total AI traffic and 70% of all AI uplink traffic in measured networks, creating demand for new infrastructure architectures that current data center designs cannot efficiently serve. Ericsson's June 2026 Mobility Report edition includes feature articles examining AI's impact on mobile user behavior, uplink traffic patterns, and AI-driven enterprise transformation, signaling that the infrastructure buildout underpinning these workloads remains a multi-year investment thesis. These figures illustrate why infrastructure-focused VCs are betting that the next wave of AI value creation will accrue to companies building the physical layer rather than those building applications on top of it.