A company had 100m shares, then repurchased 20m exactly halfway through its fiscal year. Why does the share count for its reported EPS differ from the count you'd use for equity value today?

EPS uses a time-WEIGHTED average: 100m for the first half and 80m for the second gives a weighted-average basic count of 90m for the year. Equity value is a point-in-time claim, so you use the shares outstanding NOW: 80m. If you multiplied today's price by the 90m EPS share count you'd overstate market cap by 12.5%. This is the buyback flavor of a general rule - weighted-average counts (and the diluted EPS count, which also uses average prices for TSM) are period measures built for earnings, while valuation needs the latest actual count, rebuilt for dilution at the current price. Big mid-year buybacks or issuances make the gap material.