It might sound weird, but yes. The economic perspective is the same.
You're making a huge upfront unprofitable investment into something, so that a lot of insanely profitable investments can be made 10 years in the future, that uses the result of todays investment as infrastructure.
Whether it's as important as roads, all of that is not relevant from an economic standpoint -- I'm just explaining the rationale behind the investments today.
> These companies
It's useful to mentally think of it as an implicit "team effort" among all relevant, rich companies. Openai (just an example) themselves are being "allowed" by everyone else in the "team" to not make money, because everyone knows that at the end there will be a gajillion new product companies they can all make their money back on. Openai/stargate is in that sense just a "front" for this massive infrastructure investment. Sama/Microsoft themselves will make money that way, either by building those products (just think of the ai integrations in MS enterprise in 10 years...) or by investing in others building those products. Defense folks have investments into this for the same reason.
That's why I said it's similar to roads, no one expects to make money from roads. But on those nice roads amazon will create a 1-day delivery product, charge you 10$ a month for it, and investors will make money from that [1].
[1] yes yes this specific example is bunk and historically wrong, but wanted to drive home the parallel with a small example