On putting your own people as suppliers, your distributors and customers, etc. to lower costs

A big company like Amazon sometimes gets into crazy ventures to "lower costs". They believe they can do everything, and therefore cut out all the profits and sales which they have to give to other companies. Sounds like  a groovy train of free profits, right?This is because of a poor understanding of Economics. Here's what promoted me to write this piece.
Amazon is in talks to lease 20 cargo planes to build its own overnight air operations

The summary is that Amazon wants to eliminate Fedex and UPS eventually by setting up it's own distribution and logistical network.

What's wrong? Why doesn't Amazon go even further and plant it's own wheat, that way it can feed it workers cheaply, and therefore lower costs further? Or build houses and then make the workers live in them, and therefore keep the "rent" these workers give as it's own profit?

This can't work because your capital is finite. Whatever you put in a new venture, you take away from your existing venture. In other words, if Amazon will lease these planes to make deliveries, it will necessarily do it at the expense of it's core retail operations (warehouses, inventory of items, etc). Since capital is not infinite, by putting your capital in new ventures, you are always taking away capital from your existing ones. Since the existing ones are your competitive advantage, you are almost always making a mistake.

This was explained well by Adam Smith. Wholesalers can't become retailers just because they are envious of them. They will need to take away the capital they have in wholesale goods to supply the retailers, and that means that if they put retail stores, their wholesale operations necessarily shrink, or are reduced in size (grow at a smaller rate).

You normally want to allocate more capital to your core competency, and move to newer ventures only if you fear that you have saturated out your core businesses. I doubt this is the case even for Amazon.

A few years ago there was a company called Molycorp (it went bankrupt) which was extracting rare earth minerals from Nevada. Since the allocation of capital was never clear to these people, they went ahead and bought out their own customer Neo Materials in Canada (Neo made rare earth magnets. Molycorp supplied them with the raw materials, the rare earth minerals). They thought they could keep the profit of the magnet maker too! Well, they took away capital from their core operation of extracting rare earth minerals from the ground, and note that you take away capital not just in the form of money, but by diverting the attention of Management on another business, or sending your employees to the new business, etc. So instead of one problem, they had two problems now. Sure enough, the company couldn't handle all this; and they went bankrupt in a a couple of years.

So remember this next time someone wants to cut costs or increase profits by buying their own suppliers or customers; they are in almost all cases making a terrible mistake. The part of capital allocation and the that capital is finite is often ignored by most businessmen and economists; when they are going into new ventures, means necessarily less capital for existing ones.

Amazon has as good a chance in building a successful logistics and distribution network as Fedex has of building an online retailer. Both are moving away from their core-competency; and in doing so, probably making a mistake, and will cause losses of this capital.

Update: This is also true for companies trying to move sideways into new businesses....whatever capital they use to buy the new company, they necessarily take out from their existing business. This is what Tesla seems to be doing with Solarcity-which they offered to acquire today (22 June 2016). Following that logic, Tesla should buy a wheat producer too, because certainly the family who wants to buy solar cells and Teslas needs to buy wheat as well, and some narrative of synergies and cross-selling wheat to Tesla owners can be proposed.