It's not easy to keep an old process running for over 10 years. Capital equipment breaks down and suppliers no longer support old systems. All are computerized and running some (ancient) OS be it Unix, SunOS, Linux, or Windows. One macro defect inspection system I helped develop ran on Windows NT back in 2000-04. Another was running multiple Unix and Linux flavors on various subsystems.
Now imagine a chip plant (a fab as it is known in the industry) with hundreds or thousands of such systems all running on a massive assembly line in service of one or a handful of production processes (combinations of chemistry, semi-conductor physics, optics, mechanics) to make wafers with tens to many hundreds of dies on each. The older the equipment the more breakdowns which negatively impact throughput and yield, both of which impact profit margins.
At some point either the process needs to be ported to newer equipment (major effort and expense), new "equivalent" processes must be developed for new equipment, or a whole new fab is built (say moving from older 150mm wafers and equipment to 200mm or 300mm) and new processes developed to make the older products. Huge expense and effort (years and many billions of dollars). Or the demand no longer justifies any of these options and the product(s) is (are) EOL and the fab shuttered and not upgraded or replaced.