IBM Research explored quantum hardware with AI running costs 11.5 times lower than a fully proprietary setup.

Up to 3,000 coding agents were served at once; at current cloud rates, token costs would be 5 to 10 times lower than equivalent API pricing.