Roald Coase probably has something to say on it
Something like it can work though. Many years ago I worked for Philip Semiconductor at the Mullard factory in Southampton in a department that designed and built production and test equipment. We had a formal project process where if the customer department wanted something they had to ask us first and allow us to perform a feasibility study or produce a prototype while spending up to ten percent of the rough estimate of the project cost. At the end of that we would present a detailed plan and cost. The customer department could then decide whether to offer the job to us or ask an external player such as British Aerospace to do it.
You can swap your MCRBs for real dollars and go out and purchase your own hardware, for example, or do your own service monitoring, or you can "pay" the team that manages hardware to maintain it for you. And the service monitoring team can integrate your project into the company-wide monitoring system. If you do the first option, it might be cheaper, but when things inevitably break, you're the one who takes the blame and has to fix it, instead of having in house experts who fix it.
There's central planning, so you don't have monopoly issues. The hardware provisioning team can't just jack up the price to infinity, because the company can see what the actual cost is and force the team to provide the stuff at cost.
Then you do a little magic like loosely tying bonuses to the amount of MCRBs you receive if your team is an infrastructural one, and perhaps allow management to redeem MCRBs for things like additional capacity to hire employees, and potentially even use MCRB allocation by end users as a way of figuring out what pieces of infrastructure are the highest impact/most in need of additional staffing/etc.
You end up with a system where infra teams have decent incentives to offer good tooling and to fix bugs/implement features (because there's always the looming threat of a client team leaving you and developing their own solution). User facing teams have an incentive to use existing tools and rely on the infrastructure that exists, unless there are compelling reasons not to, but when there are such reasons, they have some level of autonomy to develop unique solutions to fit their needs, and you have indirect, but still useful, signals of where upper management needs to invest additional resources, because something is being used a lot.
There's no competition, so there's no valuations, no local control (I have no company bucks I can use as an employee to, for example, get security to create a service account), just incompetent people using exorbitant pricing to provide utterly shoddy service that I can't go to any competitors over.
Want a group mailbox in Outlook? $450 dollars each.
There is a fair amount of shaking out to do but look at saas companies - to a large extent they are just being the HR department or whatever for their clients. Fake internal markets have generally failed utterly.
I think we are going to see much smaller much more productive individual companies.
maybe
This is basically what Amazon did with AWS, to phenomenal success. You can create a single-supplier market within the company for any product as long as you sell to the wider public too, and if the wider public buys from your single-supplier then you know your internal customer is getting a good product.
> We had a formal project process...
Been there at companies with similar models, done that, the entire model is fundamentally broken. You may pay the lowest amount of cash on paper but the process takes so long that the additional amount of money you spend on management and architects over the project planning process usually outweighs any supposed competitive savings. You also pay in agility / opportunity costs.