NVIDIA RTX Spark Superchip drives AI mini‑PC rollout at Computex 2026

At Computex 2026 in Taipei, major chip makers shifted AI processing from data‑centres to edge devices. NVIDIA unveiled the RTX Spark superchip, a 20‑core ARM‑based processor with a Blackwell‑architecture GPU delivering 1 petaflop (a quadrillion operations per second). The chip is designed to run large language models of up to 120 billion parameters locally on mini‑PCs and even on displays, reducing latency and data‑transport costs.

ASUS, MSI, HP, Dell and Lenovo showcased compact AI‑enabled systems, with entry‑level pricing projected around €1,850 and higher‑end configurations up to €2,700. Mass delivery is slated for autumn 2026. Analysts note that while the hardware promises new enterprise use‑cases, the high component costs and an IDC forecast of an 11.3 % decline in global PC sales may confine early adoption to developers and creative professionals.

The move highlights a broader industry trend toward local inference, prompting enterprises to address new management, security and software‑update challenges as AI workloads migrate to the network edge.