Who Pays for the Internal Platform in a Portfolio
Who pays for the internal platform when it serves twenty products? The first product carries it, and every product after that pays almost nothing. Here is the math.
The first product pays for the internal platform. Every product after it pays almost nothing. That is the whole economic argument for building a shared foundation, and it is also the reason most people never get there: the bill lands entirely on product one, before you have any proof the platform will pay off. If you cannot stomach that front-loaded cost, you will build twenty products twenty times and wonder why the twentieth still takes as long as the first.
Who should the cost of a shared foundation land on?
It lands on whoever builds the second product. Product one does not need a platform. It needs to ship. You build auth, billing, and a deploy pipeline the direct way, tied to that one product, and you get it live. There is no shared anything yet, and there should not be. Abstracting for a portfolio of one is premature abstraction with a business plan attached.
The cost of the platform is really the cost of extracting product one's plumbing into something product two can reuse. That work is not free and it does not help product one at all. It helps product two, three, and the twentieth. So in accounting terms, the platform is an investment charged to the portfolio, funded by the surplus product one throws off, paid back by every product that skips the rebuild.
If you try to spread the cost evenly across all future products, you never start, because none of them exist yet. Someone has to eat the setup cost up front. In my portfolio that someone is me, and the payoff shows up as each next venture being cheaper to ship than the last.
The marginal cost math nobody runs
Here is the calculation that justifies the whole thing. Say a product from scratch costs you X in effort. The first three products cost roughly X each: X to build product one, X-plus to extract the platform on product two, and product three still shaking out the rough edges.
Then the curve bends. Product four inherits auth, billing, deploy, database access, and monitoring for free. Its cost is X minus everything the platform already solved, which in my experience is well under half of X. By the tenth product the platform work is close to zero and the only real cost is the thing that makes that product different from the others.
That is the point of the marginal cost of the next product falling. The platform converts a fixed per-product cost into a one-time cost the whole portfolio amortizes. The more products you run, the cheaper the average product gets. A single product can never justify this. A portfolio pays it back many times over.
The ongoing tax, and who carries it
The platform is not a one-time bill. It has a running cost: maintenance, upgrades, keeping twenty products compatible as it evolves. That tax is real and it is easy to underprice.
The mistake is letting the platform tax hide inside product work, where it never gets measured and slowly starves every product of attention. I keep it explicit. Platform maintenance is its own line of work, funded by the fact that it makes every product cheaper. When I version the foundation carefully, that tax stays small because upgrades are additive and products pin their version. When I get sloppy, the tax balloons into a portfolio-wide migration and eats a week.
The way to keep the tax affordable is to treat the platform like a product with an owner, not a side effect. That is why I say your internal platform is a product. Products get maintained on purpose. Side effects rot.
Why the front-loaded cost is worth it
The uncomfortable truth is that you pay before you profit. Product two costs more than it would have if you had just cloned product one, because you spent extra to extract a reusable base. The return does not arrive until product four, five, and beyond.
I fund that gap with tooling that makes extraction cheap in the first place. A build platform that turns "make this reusable" into a fast operation lowers the price of product two, which lowers the bar you have to clear before the payoff starts. That is the practical reason I run my portfolio on Bootspring and host it on my own infrastructure: both push the front-loaded cost down far enough that the platform pays for itself by the fourth product instead of the tenth. Get the cost curve to bend early and the answer to "who pays for the platform" stops mattering, because by then it is paying you.