If the profit of a robot is Xrobot and the cost Crobot, compared with a human with profit Xhuman and the cost (healthcare, training, and salary) Chuman. We can formulate a few scenarios.
It should also be noted that human’s lives can't be replaced and robots can be maintained indefinitely (while humans age and die), so even if all the variables are equal robots will have the edge.
I'll ignore your monetary figure as to leave this answer open to be used for other cases.
So, lets see:
- Crobot ≈ Chuman, and Xrobot ≈ Xhuman. Robots will slowly replace humans, as the effort turns from recruitment to manufacture. The proportion of humans to robots will change from 0% robots to 99.999% robots – the particular proportion at a given time depends on how much time has passed since the technology for robots became mainstream and how quickly the replace humans. How quickly they replace humans depends on migration costs… basically: the more humans there are to begin with, the slower will be the conversion rate, but the conversion rate will accelerate as issues are solved and production grows.
- Crobot ≈ Chuman, and Xrobot > Xhuman. If robots clearly perform better than humans by around the same cost, it will ease transition. The idea is the same as above, just faster.
- Crobot ≈ Chuman, and Xrobot < Xhuman. If robots underperform, then they will have niche uses... I imagine they would be used for very risky situations. There would be a hard cap on the rate of humans and robots. My guess based on reality is a rate of approx. 1 robot per each 10000 officers.
Crobot > Chuman, and Xrobot ≈ Xhuman. If robots are expensive but they perform nearly as equal to a human, there is not enough incentive to deploy robots. Those that exist are only test-types. I would expect less than 100 robots in operation, mainly in pilot programs (I mean, early field tests).
Crobot > Chuman, and Xrobot > Xhuman. Since robots are expensive but also outperform humans, the strategy depends on which factor is more relevant. If robots cost twice as much as humans but perform only one and a half times better, then having twice as humans is better… but the supply of humans is limited, so you get as many humans as you can and then robots. On the other hand if robots cost one and a half times what humans cost but perform twice as good, you want to start migration to reach a point when you have as many robots as you can and then a few humans to fill any niche positions.
- Crobot > Chuman, and Xrobot < Xhuman. Robots are trash; they are too expensive and perform badly. You may have a few prototypes in development, perhaps not more than 10, to develop the technology until they become better or cheaper.
- Crobot < Chuman, and Xrobot ≈ Xhuman. Robots are cheap and about as good as a human, you back up your migration strategy with a good human rights speech and you have your robot military forces. The bottle neck is production. The rate of robots to humans will rapidly move from 0% robots to 99.999% robots.
- Crobot < Chuman, and Xrobot > Xhuman. In this case robots are cheap and perform better than humans! Similar to the case above, you will migrate as fast a politics and production allows you.
- Crobot < Chuman, and Xrobot < Xhuman. If robots are cheap but underperform, they are cannon meat. You produce them mainly to have numbers. I would expect the rate to be at least 1:1 with humans, up to 3 or 4 robots per human.
Of course, it is expected that robots become cheaper as time passes, and also that they become better. So, we could imagine them going by a few of the stages above:
- Prototypes. About 10 robots in research and development.
- They start to perform better, and pass to a pilot problem with around 100 robots in operation.
- A new cheaper model is developed. This one performs worst but it is safe to use in the field. 1 robot for each 10000 officers.
- Robots continue to become cheaper. 1 robot for each officer.
- New models outperform humans. 1 human per 10000 robots.
Note: do not consider this to be linear time. While the development may be exponential, the production of new units is not. Since the production is the bottle neck, interpreting the list above to be logarithmic is the best approach.
How fast does it happens depends on the size of the military forces and the production of new units. Politic cycles may also delay the start of the migration project. Decide this things for your fictional work.