Positron AI Enters Nvidia Turf With Oracle Deal

Positron AI, an emerging AI chip designer, is entering Nvidia's core domain in AI data centers through a partnership with Oracle. According to CEO Mitesh Agrawal, Positron has already deployed systems... Read original →

Positron’s Oracle-backed entry targets NVIDIA’s inference stronghold not through raw performance, but architectural frugality—eschewing HBM for optimized LPDDR on TSMC’s N3P, with air cooling to sidestep power constraints. This forces EDA and OSAT players to accelerate chiplet integration while revaluing LPDDR5X supply chains. Yet reliance on TSMC amid U.S. fab restrictions exposes acute geopolitical fragility. NVIDIA will likely counter by deepening Rubin’s software moat rather than price wars; Intel and SambaNova may fast-track MoE-optimized low-power ASICs. If Positron demonstrates superior watts-per-inference versus A100-class GPUs within 18 months, it could capture green data center and edge AI niches—but scaling remains bottlenecked by cloud providers’ entrenched procurement cycles and capex inertia.