Computing power is not a business model
Scale is frequently mistaken for strategy. Why the quantity of available computing capacity says little about whether a venture carries commercially.
In discussions of technological competitiveness, computing capacity is often treated as a result. More of it counts as progress, less as falling behind. That equation does not survive sober examination.
Capacity is a means, not an end
Computing power is a factor of production like energy or floor space. A company with a great deal of floor space is not thereby more successful than one with little; what matters is what happens on it. The same holds for computing capacity, but it is said less often, because capacity is easier to communicate than benefit.
That has an uncomfortable consequence: capacity becomes the metric because it is easy to measure, not because it is informative. Use it as a measure of progress and you are measuring the input while calling it the outcome.
A contest about size is a contest about capital
Where capacity is the yardstick, competition becomes a competition over investment volume. That favours participants with deep pockets and squeezes out those who cannot or will not match the stake. For those involved it means rising commitments against unchanged uncertainty about the return.
For companies outside that group, the obvious response is not to bid after all, but to examine whether their own use case demands that order of magnitude at all. Frequently it does not.
Specialisation beats generality where the task is narrow
A general-purpose system has to be adequate at everything. A specialised one has to be good at one thing. Where the task is clearly bounded, the second route is regularly cheaper, faster and easier to verify — not because specialisation is superior, but because less is being paid for than is needed.
The drawback is equally clear: specialised solutions age badly when the task changes. That is a trade-off, not a rule.
The more useful question
Instead of asking how much computing power a venture requires, one can ask which process inside the business becomes measurably better as a result. If that question cannot be answered, the capacity question is premature. If it can, the required scale usually follows by itself — and regularly turns out to be smaller than the discussion suggests.