There are two aspects to this question on risk. One is the probability of failure, the other is the potential harm that could result from failure or degraded performance. I'll focus on the first.

Wikipedia provides some interesting statistics on computer power supplies.

Life span is usually specified in mean time between failures (MTBF), where higher MTBF ratings indicate longer device life and better reliability. Using higher quality electrical components at less than their maximum ratings or providing better cooling can contribute to a higher MTBF rating because lower stress and lower operating temperatures decrease component failure rates.

An estimated MTBF value of 100,000 hours (roughly, 140 months) at 25 °C and under full load is fairly common.

**Understanding the number**

This number, by itself, is a little misleading because it doesn't mean what most people think it means. Ignore, for a minute, differences in manufacturing quality, usage patterns, etc. This is a big pool of various models, tested under good operating conditions (so your mileage may vary).

It is an average time, but the failure times are not a bell curve. Some units last an extremely long time, which distorts the average. It is **not** that half of the units last 100,000 hours; rather, 63% of units fail by then, and 23% of units are expected to fail within three years of operation.

So if your PSU has already survived four years, is "average quality" and has been used under good conditions, then statistically speaking, it has roughly an even chance of lasting another eight years, and in the neighborhood of a 75% chance of lasting another four years.

**Is it realistic?**

Realistically speaking, I seriously doubt those statistics. But even much worse statistics yield surprising results.

For example, suppose we accept the 23% failure rate within three years, but assume that only 30% last more than eight years (just a number that sounds reasonable based on experience). Intuitively, you might think that the average life under those assumptions is 6 years; and since your existing unit is already 4 years old, you would expect only another 2 years from it. But that's not the case. There would be an even chance that it would last another 3 1/2 years, or roughly what some people would consider the "technology" life of the next computer.

Take that a step further. Under these assumptions, we have a 50% chance that your old one will die within 3 1/2 years. However, if you buy a new one, there is roughly a 27% chance that it would die during that same time. So the **increased** risk of failure for using the old unit is only 1 chance in 4, even under these assumptions.

Just some numbers to give you some perspective.