IBM's Four New AI Models Forecast 100,000 Data Series Overnight
The early-access build reaches AWS first, run through two Flink SQL commands, AI_FORECAST and AI_DETECT_ANOMALIES, no separate ML team needed.
- TTM, the tiniest of the four, forecasts more than 100,000 series nightly on CPU alone.
- PatchTST-FM predicts probabilities, FlowState tracks uneven time gaps, TSPulse flags anomalies and fills data gaps.
- Confluent Platform support follows this AWS cloud release, after tests with cement, steel and food firms.
Why it matters: The company that already carries a factory's data now also sells the model predicting it.