AI dominated startup funding in the third quarter of 2026, with two inference companies each announcing $1 billion rounds. At the same time, the data center buildout's insatiable demand for memory pushed investors toward startups offering alternatives to HBM, which remains expensive and capacity-limited. These alternatives include building on commodity DRAM, holding models in on-chip SRAM, and stacking memory directly onto compute.
Data movement at all levels of the stack remains a critical bottleneck, making interconnect technologies, networking chips, and photonics the most active areas of the quarter. Startups are attacking the problem with a wide range of approaches, from manufacturing-ready co-packaged optics and millimeter-wave RF signals over plastic waveguides to ferroelectric glass modulators and coherent DSPs designed specifically for AI optics.
Nearly $7.5 billion of the quarter's total was directed to just 22 startups that raised $100 million or more, including eight that saw investors return within roughly a year. Notable deals include Xsight Labs' $300 million round for programmable DPUs and Ethernet switches, Celero Communications' $275 million Series C for AI-optimized coherent DSPs, and Cornelis' $205 million raise for scale-out networking. In total, 80 companies raised over $8.5 billion in Q3 2026.