wonder how that is solved in practice?

(also wondering about the armed services and paying for medical school)

The armed forces are probably not the best example. There tend to be pretty severe consequences to just deciding you don’t want to be in the Army any longer.

What is the usual term of an enlistment?

I believe 4 years active duty and 4 years in reserves.

The investment into training isn't merely money and can't usually be neatly calculated. It's usually paid in the form of time, motivation, distraction and lost opportunities by more senior staff. If you, personally, sink a year into training someone intensively, and then they leave, would you perceive it as merely lost opportunity costs? Probably not. Mentors and mentees are supposed to have a more than purely transactional relationship in most understandings.

As to why not just pay more, probably the reason you were trying to train someone to begin with is because you can't afford to pay the higher wages to just poach the most experienced people? Otherwise why not just pay the higher wages to start with and skip the risky and slow training? Also, the software world has the unique problem that everyone is competing against VC backed startups that are willing to spend more up front to minimize time to market, so these firms generally don't train but can afford to poach anyone who is trained at a perhaps unsustainably high salary.

> why not just pay the higher wages to start with and skip the risky and slow training?

It’s not like walking around with sacks of money and hiring people interested in that money is low-risk. I’ve found training to be lower risk/uncertainty than hiring in terms of fewer outright failures and negative surprises.

Yet, companies just love to hire instead of retain. Take an employee who just got trained/experience and is now worth $N, who is currently making $N-2. Most companies I've worked with would piss and moan about giving that employee a raise to $N-1. They'd rather let him leave. Then, they suddenly have no problem hiring a relatively unknown employee and paying him $N+2. It's never made sense to me in all my decades in this business. Obviously value-destroying behavior that's routinely done by most companies.

You talk as if figuring out what an employee is worth is easy, and that worth is absolute (same to all employers). If you have competitors who are willing to pay a premium over what the skills are worth to the first company not to train someone, then this situation can arise.

The issue is that a company that (all else equal) doesn't spend on training has more budget for salaries. Employees who already have skills generally value salary over additional training.

Also there's the obvious culture / coordination angle: if enough companies train their employees, then the people they hire to replace ones that left were already trained by their competitor. It boosts average competence level for everyone. And, to the extent it builds goodwill with the (ex)employees, you may discover they're sending quality candidates your way.

Unfortunately, this is a coordination problem, and those are what humanity sucks at.

> This mentality has never made sense to me. Why would a business invest so much money into training someone, then fail to spend the money needed to retain the (more valuable) trained employee??

It didn't and doesn't to me either. It's very similar to the disturbingly common case of a company cheaping out on office chairs, computer hardware, etc. when the total cost of creating a comfortable working environment for an employee is less than the company pays them for a month of work!

Or, related, and perhaps even more common in tech: skimping on infra (be it physical or virtual), despite again wasting more money in lost productivity than they save. Like e.g. at one place, I used to regularly complain about the amount of build bots we had for the team, and the resources allocated to them. The response was always that it's as much as we can get from the company cloud, to which I'd reply, why not buy more compute? Never got a straight answer to that. Sure, more and beefier machines cost some money, but when the CI is running at capacity at all times, and you have several well-paid devs sitting there and getting distracted while waiting for their jobs to finish... Penny-wise, pound-foolish was the saying?

In fact, I'd happily spend a fraction of my salary to buy a beefy PC to use as a compilation server. And another one to lend to the team's CI infra. But I can't do that, because procedures and company IP protection and whatnot. In the end, I was left with the feeling that the company just wants to make it hard for me to do the very job they're paying me for.

You'd think that management and finance education would let the company do the obvious math, but apparently, it doesn't. But then, the whole problem seems eerily similar to what's apparently a Business Wisdom today (and to me is total insanity): always being focused on acquiring new customers, and not paying attention to retaining existing ones. Maybe the reasons are similar, whatever they are?