True but this isn't specific to the tech bubble. It's a feature of capitalism. Competition forces firms to adopt shorter term horizons. If a firm has significant profit to make by focusing on the short term and it does not, its competitor would. If the profit possoble within this period is significant, having the competitor collect it runs the risk of the current firm failing, or the competitor accumulating enough for hostile takeover, among other failures. That would stop the current firm onwer from collecting profits in the future. Even if focusing on the long term is more profitable over time, firms may not survive in a competitive environment to realize long term profits. These are some of the fundamental processes that drive firms into short term horizons. With liquid asset markets there are even more immediate processes driving firms into short term planning.

Add to that planning based mainly on prices, which don't capture a ton of reality and you get situations like a water hungry datacenter in the desert, cause the price of water does not capture its long term availability for example.

All of this has happened in the past, even a century ago. It's happened and keeps happening in other industries too. For example the fossil fuel industry.

They're locusts. They don't think about anything past the next fiscal quarter.

Because closed loops are more expensive in the short term, making it a non-starter

Closed loop watercooling is really just air cooling with extra steps. The water is heated by the devices and cooled by a large radiator with fans. Or it's cooled with a chiller which in turn is cooled by a radiator with fans.

Replacing the water is the most effective (yet wasteful) way to remove the heat.

Cheap land, dry air is good for evaporative cooling, and many arid areas have a surprising amount of ground water. It ultimately comes down to being the cheapest option, not the smartest or best option.

In addition to the other answers;

America's deserts are tectonically stable and don't experience natural disasters. If you want your data and/or compute running in two regions for redundancy, somewhere in the desert is a good choice for one of your DCs.

Two more reasons not yet mentioned:

It is close to a population center (Phoenix) keeping latency low to customers. Getting customers off the public Internet quickly and into your private network fast is best for a lot of reasons.

Cheap and abundant solar power. Data centers are extremely power hungry and power lines are expensive so companies like Amazon almost always secure abundant power rights before building. Google built their first data center in The Dalles Oregon because an aluminum smelter had gone belly up and left a bunch of capacity unclaimed in a local hydroelectric dam.

Because the local and state governments in those deserts keep promising them unlimited water for nearly free

Demand, ain't much of it in the desert. Also, easy to manipulate governments.