Semiconductor Engineering's Q2 2026 funding roundup counted 80 semiconductor startups raising over $6B, with 18 companies at $100M or more. AI data center inference chips and their enablers -- faster data movement, higher-voltage power semiconductors -- took the bulk of the capital. Edge silicon re-emerged as a funded category after a quiet 2025: investors see inference moving to devices and are writing checks accordingly. Quantum had an unusually strong quarter, with 21 companies raising across superconducting, spin, neutral atom, and ion trap modalities, plus cryogenic control electronics and quantum-native test equipment.
Two specific rounds are worth tracking for what they signal about where bottlenecks are being priced in. SiFive's $400M Series G led by Atreides, with NVIDIA participating, is a bet that RISC-V processor IP -- not just cores, but the data-center-grade out-of-order and vector/matrix IP in SiFive's portfolio -- will be a structural input to AI chip teams who cannot afford to design everything from scratch. AttoTude's $52M Series C for THz radio-over-wire interconnect is more specific: the constraint they are attacking is the mid-range connectivity gap in AI infrastructure, where copper falls short at 200G/400G/800G per-lane rates and optical adds complexity most teams cannot absorb. Their approach -- ASIC signal generation combined with low-loss dielectric waveguides -- is an attempt to thread the needle between those two failure modes.
The pattern in these rounds is not random. It maps onto the parts of the loop that are still slow: inference throughput (compute and memory, being addressed by AI hardware startups), design IP reuse (addressed by SiFive-type rounds), and system interconnect (addressed by AttoTude). The $6B in Q2 is the market's current read on which constraints are close enough to breaking that a 5-7 year startup cycle can catch them.