Google puts TabFM inside BigQuery

Google put TabFM into BigQuery preview, turning tabular classification and regression into 1 SQL statement while claiming distributed inference can process millions of rows in minutes. Its new AI.PREDICT and AI.EVALUATE functions remove separate training and deployment steps, but Google still says traditional models win when teams need very large datasets, custom tuning, or feature-importance explanations. For data teams connecting governed enterprise data to agents, file TabFM as a low-friction baseline—not proof that zero-shot forecasts fit production.