These are the three questions I think this dataset can answer best. The facts behind them come from the profiling queries I ran. **What the data covers.** The table holds 422,729 daily rows for 152 semiconductor tickers, from 2012-01-02 to 2025-08-04. Each ticker maps to one company. About 99,984 rows (23.7%) fall after ChatGPT's launch, which I treated as the AI-surge window. **1. Did the AI surge change the price level, volatility and trading activity of semiconductor stocks, and did it affect all companies or only a few?** - The dataset spans 11 years before and about 2.7 years after the surge, so a before/after comparison is possible for every ticker. - Dataset-wide, the average daily high-low range is 3.69% of the close. The median close is about $35.57. - The analysis would compare each company's price growth, range and volume before and after late 2022, then rank the companies with the biggest shifts. - This would show whether the gains were concentrated in a handful of AI leaders or spread across the sector. It bears directly on overweight/underweight decisions. **2. Which companies are the real risk drivers? How does volatility (daily range as a % of close) vary across companies and over time, and does it cluster in particular periods?** - The 3.69% average is only a sector-wide figure, so it hides differences between names. - The gold tables by company, year and month make it cheap to build a volatility league table and a timeline. That would separate stable large caps from high-risk small names. - It would also show whether risk rose during the AI run-up or around earlier shocks. - This is the input for position sizing and for deciding which names belong in a sector basket. **3. When does trading volume signal unusual interest, and are those spikes followed by price moves?** - Volume is extremely skewed. The median is 898,500 shares, but the mean is about 8.47M and the maximum is about 3.69B. - That gap means a few names or days dominate total volume, so "unusual" has to be defined relative to each stock's own baseline, for example a multiple of its rolling median. - The analysis would flag spike days per ticker and test what happened to price over the following days. It would also check whether spikes became more frequent after 2022. - Data quality needs handling first. There are 4,888 rows with zero volume, which are probably halted days or bad records, and they should be excluded or investigated before the baselines are built. **Caveats.** - The card lists about 149 companies but the data has 152, so check the company list before building a benchmark basket. - Nothing here establishes that the AI surge caused any change, because the data is observational and has no market-wide control. - The "post-ChatGPT" cutoff is my own choice, not a column in the data.