Intel knew about its Pentium FDIV bug and decided the public didn't need to know

In mid-1994 Intel engineers discovered a flaw in the Pentium processor’s floating-point unit. A bug in the C script that generated the SRT division lookup table left five cells empty, so when the division instruction hit those entries it returned a value off by as much as 61 parts per million instead of the correct result. Intel had already begun a mask revision to fix the bug in future production runs — but it did not disclose the flaw to customers, and continued shipping existing inventory of the flawed chips.

Thomas Nicely, a mathematics professor at Lynchburg College, independently found the bug in June 1994 and confirmed it by October. When he posted about it online, the story spread. Intel acknowledged the flaw on November 7, 1994, but dismissed it as minor, claiming a typical spreadsheet user would encounter the error once in 27,000 years. Intel offered to replace chips only for customers who could prove they needed the accuracy — a condition that infuriated the market. IBM then announced it would stop shipping Pentium-based PCs, and CNN ran the story nationally.

The public backlash forced Intel to reverse course. On December 20, 1994, Intel agreed to replace every flawed Pentium on request, no questions asked. On January 17, 1995, it announced a $475 million pre-tax charge against earnings to cover the cost. The episode became a case study in product crisis mismanagement — a company knew of a defect, kept it quiet, let a third party find it, and then argued with the public about whether the defect mattered.