Square Launches AI-Powered Kiosk With Real-Time Conversational Ordering

Square has introduced the Square Kiosk, a 21.5-inch touchscreen terminal shaped like a large pill that integrates an agentic, conversational voice layer. The system uses presence detection to greet customers as they approach and allows them to place orders by speaking naturally, just as they would with a human cashier.

Real-Time Menu Understanding

The AI agent fully comprehends the seller's menu in real time, enabling customers to ask questions, request substitutions, and modify items mid-conversation. This level of contextual awareness requires continuous, low-latency language processing — a workload well suited for GPU-accelerated inference either on-device or via a nearby edge node.

Payment is handled through a built-in contactless reader and card slot. The hardware sells for $1,999 for businesses on Square Plus or Premium plans, plus a $99 monthly software fee per device.

Why It Matters for GPU / AI Infrastructure

Conversational kiosks like this depend on streaming speech recognition, large language model inference, and text-to-speech synthesis — all running with sub-second response times. Deploying such pipelines at scale creates steady demand for GPU cloud capacity, especially for providers offering low-latency edge inference or managed model serving. AiGpu's infrastructure is positioned to support exactly these real-time, voice-first AI workloads.

As retailers and quick-service chains adopt voice-enabled self-service, the underlying compute footprint will shift further toward accelerated hardware. This kiosk is a tangible example of AI moving from backend analytics to frontline customer interaction, driving sustained GPU utilization.