Foxconn's AI Factories Cut Scheduling Hours by 25% as FoxBrain Hits 1 Trillion Tokens

Foxconn, the world's largest contract electronics maker, says its AI factory programme has cut production scheduling labour hours by 25 percent, according to a company update on 26 October. The programme also reduced engineering workload for injection-moulding changeovers by half, giving the world's most closely watched manufacturer a concrete measure of what industrial AI can actually deliver.
At the centre of the effort is FoxBrain, Foxconn's own large language model adapted with industrial data. The company says its NVIDIA DGX SuperPOD infrastructure trains 120-billion-parameter models 2.48 times faster than its previous H100-based systems, and processed one trillion tokens in just 12 days. The AI is being applied across manufacturing, electric vehicles and smart cities.
Foxconn is careful about what the 25 percent figure means: it applies to production scheduling, not to total staffing. Even so, the numbers matter beyond one company. As factories race to use AI for planning, quality and changeovers, Foxconn's results offer an early, measurable example of the gains — and the limits — of bringing large AI models onto the factory floor.
Source: read more
Comments
Post a Comment