Token consumption surges, AI agents go live in batches: the AI race in listed banks' half-year reports

As listed banks' half-year reports for 2026 roll out, a new competitive front has come into view: AI. A report from China's Red Net published this week says the AI disclosures in these reports show token consumption surging and AI agents going live in batches, an AI race taking shape across the banking sector.

Token consumption is one of the most direct signals. The report notes that token volumes disclosed across the half-year reports grew sharply, indicating that large-model usage inside banks has moved from scattered pilots to genuine production workloads rather than demos and concepts.

Agent deployments are scaling up in parallel. AI agent is becoming a recurring theme in banks' digitalization disclosures, with multiple lenders saying agents have been rolled out in batches across different business lines.

The coverage frames the situation as an AI race among listed banks. With the industry under pressure to control costs and run more efficiently, banks increasingly treat AI as a way to cut costs and improve service efficiency at the same time, and executives are now putting AI usage metrics into regular investor communications.

The disclosure itself carries the stronger signal. When operational metrics such as token consumption and agent counts show up in half-year reports, banks' AI investment narrative is moving from storytelling to measurable usage.

What to watch next: how full-year AI spending and token costs evolve once the half-year window closes, how quickly the batch-deployed agents translate into business results, and whether the AI race reshapes banks' model procurement and compute planning.