Obviously it's not a hard truth that no capacity improvement is worth it, but the idea of induced demand is describing something that is quantifiably true in many cases. And it's all about cost and physical space - that's why nobody worries much about induced demand in, say, bike lanes or train lines, because the route capacity of those modes is so much higher than car lanes.
The issue at the end of the day is that, assuming a certain population density and a growing population, roads tend to become extremely expensive (because of the required property buybacks, etc.) while at the same time cars fundamentally take up too much space in urban/suburban areas for the number of people they tend to carry. (Obviously this is almost never an issue in rural areas, given the low population density).
So it's not so much the issue that you couldn't theoretically build enough capacity, it's for most brownfield areas, you won't add enough capacity to fix traffic because of the huge cost - not to mention environmental destruction, or knocking down people's houses to replace with more and more roads, etc.. This is of course very different to, your example of internet bandwidth, where you can trivially increase the capacity by many multiples without taking up any more space physically. The closest analog in the transport world would be rail, which uses very little real estate compared to huge potential capacity.
Given that fundamental issue (the cost and space that roads take up, and the fact that cars are one of the least volumetrically efficient modes of transport), what becomes observable is the ironic fact that the cities that tend to be best to drive in are the ones that have provided many good alternatives of good public transit, bikeways, etc. as opposed to the cities that were designed around the car.