Exclusive: Quantum without quantum computers — Quinas bets on quantum tunneling to break AI's memory bottleneck

Quinas is betting on a "quantum-adjacent classical" device category—exploiting tunneling physics without quantum coherence, mirroring how MRAM leveraged spin without qubits. Upstream, this demands sub-angstrom barrier engineering via ALD/MBE and a full overhaul of tunneling-aware TCAD in EDA toolchains. Downstream, a 10x reduction in per-bit access energy would fundamentally rearchitect GPU memory subsystems currently locked into HBM roadmaps by NVIDIA and AMD. On compliance, if process nodes dip below 5nm, BIS export controls and advanced packaging capacity allocation in Taiwan, China and mainland China become de facto barriers; the 2023 gallium-germanium export restrictions already demonstrated how material supply chains get weaponized. Strategically, Samsung's HBM4 (2026) and SK Hynix's PIM are the visible competitors, but Quinas' real disruption is reframing "quantum" from a compute narrative to a storage one—potentially invalidating the entire HBM investment thesis. Within 12–24 months, expect IP licensing or acquisition by a memory major post-lab validation, and "quantum-adjacent devices" entering JEDEC standardization drafts as a new product taxonomy.