Why is the temperature range of industrial and military products so high?

The maximum temperature the silicon experiences can be much more than ambient. 50 °C ambient certainly happens. That's only 122 °F. I've personally experienced that in the Kofa Wildlife refuge north of Yuma Arizona. You need to design to worst case, not wishful case. So let's say ambient can be 60 °C (140 °F).

That by itself isn't much of a problem, but you don't get that by itself. Take the same thermometer that reads 60 °C in open air and put it in a metal box sitting on the ground in the sun. It's going to get much hotter.

I've seen someone fry a egg on the hood of a car in the sun in Phoenix AZ. Granted, this was a stunt deliberately set up for this purpose. The car was parked at the right angle, the piece of hood was tilted at the right angle, and painted flat black. However, it still shows that just a piece of metal sitting in the sun can get really hot.

I once left a car parked at the Las Vegas airport for a few days. I had left one of those cheap "stick" ballpoint pens on the dashboard, partly sticking out over the side. When I got back the pen was bent at 90° over the lip of the dashboard. I don't know what temperature such pens melt at, but clearly it gets a lot hotter than ambient under common enough conditions in a enclosed box.

If you left some cheap piece of consumer electronics on the dashboard in the sun and it didn't work, you'd probably be a little irritated, then toss it and replace it. If the controller for your oil pump stopped working in the summer because it got too hot, you'd lose a lot of money, be pretty upset, and probably buy the replacement from a different company that takes quality more seriously. If your missile defense system stopped working because you deployed it in the desert of Iraq instead of some nice comfy test range in Massachusetts where it was developed, you'd be dead. The procurement officers that don't get fired will be extra careful to require all electronics to work at high temperature, and insist it get tested under those conditions.

First of all, military equipment is expensive. You can afford to actually test things for high temperatures only if your customer is willing to pay. Military customers tend to have budgets that normal people can only dream of.

Then, obviously, if you put an IC into a missile, you might not want that thing to fail if your missile gets hot from its burning end or from its air-frictioning end. Same goes for things that might be put in a satellite, intercontinental rocket etc: as soon as you hit space, and are in the shadow of earth, things can get really cold. Military and Aerospace (which typically are mostly the same companies) are the typical place where you'd expect a device to withstand a lot of G of acceleration, be hot-cold-hot-cold-hot within seconds, still need to be extremely well integrated and lightweight, and where costs just don't really matter much compared to risk:

The main difference (aside from how temperature management is done physically), though, is simply that these three groups of applications do a different kind of risk assessment:

• consumer/commercial grade device: 1/5000 of your TVs fail within the five years because some IC baked in heat too long. Bad thing. Many customers will just get a new one. For the remaining 1/10,000 customers, you'll have to do service (calculate that into your product cost) or live with a degraded image (which you don't really have to, because your competitors do the same). So, having more security margin in your designs doesn't make much sense, as little as testing components to the edge of the assumable environmental conditions. You're in a market where price is most important, and failure rate is mainly a concern for the manufacturer's finances.
• industrial grade device: Your customer is someone who's hinging a possibly very expensive production line on your product. Let's say Volkswagen's production line stands still for 8h because your IC failed to function. That's a very solid amount of loss you've just caused. VW will be willing to pay extra just to make sure that its suppliers make sure you tested the components for all the environments that are likely to occur, and quite a bit beyond, to keep that risk manageable.
• automotive grade device: Human lives are at stake. That's not as important as the fact that cars vibrate like hell, are complex as hell, get partially hot as hell, and are rolled out in millions, meaning that figuring out that whatever component gets a little hot to work reliably (even if it's just something non-critical to safety) means you might need to service a lot of cars, which is really expensive, and you actually risk your brand image. Every country has its own prejudices against "that car manufacturer with shoddy reliability and bad electronics", and it's seriously hurting their sales.
• military grade device: Well, the promise of military is to be ready whenever for whatever. They will not risk anything failing just because they didn't ask all suppliers to fulfill extreme environmental specs. That's how they roll – don't leave anything to risk, especially if your application is expensive as hell anyway (think fighter jets) or gets deployed in tens of thousands and is still life- and mission-critical (think military communication equipment).

Military (and aerospace in general) equipment is often:

1. In an unpresserised bay which means cooling the equipment is by conduction. Convection cooling loses meaning at 30,000 feet as there are very few air molecules to transfer heat by convection. It is much more difficult to effectively transfer heat by conduction only.

2. In a glare zone (think just under the canopy in a fighter aircraft) and this area can be very hot.

3. In a bay where the ambient temperature may be in excess of 70C.

4. In the leading edge of a wing, which can range in temperature from icing conditions (well below zero) to very hot (at Mach 2 or so, the friction of even the few molecules available is still very high; that is why the space shuttle had elaborate heat management for re-entry).

It is not unusual to have a card edge temperature requirement of 85C for short periods (30 minutes typically) and it does not take much processor (to name but one device type) activity to raise the junction temperature to 120C or more.

In summary, military and aerospace environments are really harsh (as are down hole applications incidentally).

As noted by others, fully qualified military grade parts can be expensive (as much as 10x the cost of the commercial equivalent and in some cases more); in response to that some manufacturers have instituted screening programmes for plastic parts which still have a premium, but not as much as the previous solutions.

[Update]

In response to the comment on card edge temperatures, here is a typical conduction cooled chassis:

The outer part of the chassis is known as a cold wall (where we can know the temperature) and it may simply be metal or have other methods of maintaining a reasonably well known temperature.

Now here is a typical card, with heat ladders:

These are often made of aluminium (it is cheap and has decent thermal parameters) and the ladders are in contact with the side edges of the enclosure above; as there will be some heat differential between the outside and inside of the box, the temperature withstanding requirement for the PCB is set at this internal heat ladder, which is, as you can see at the card edge.

As the heat must get from the components to this point, it is not unusual for the PCB at a hot component (such as a processor or GPU) to get to 95C or more with a card edge temperature of 85C (which is often a specific requirement).

The thermal resistance of most flavours of FR-4 is \$0.4 \frac {W} {mK}\$ so lots of internal metal layers will exist in this type of card.

In some situations, we may need to use thermally clad PCBs which although expensive may be the only way of getting the heat out.