There are two primary differences between 1999 and now in consumer web: 1) there's a massive advertising market now that works; 2) for 99.999% of web sites, it has gotten drastically cheaper and easier to operate, and over another five or ten years, will become nearly free (what will $10 buy at Digital Ocean in another five years?). Traffic and resource demands for most web sites will not keep up the value proposition improvements we're seeing, it'll push the cost toward zero.
Those two differences have in fact changed the landscape. I can run a couple person startup that generates millions in ad revenue now, with exceptionally high margins. Plenty of Fish is the most famous example of this. Techmeme essentially worked like that, as another example. Hacker News could be like that, if PG wanted it to be.
Point being, it's now drastically easier to monetize traffic long term, compared to 1999 (when the ad market was tiny, backwards, poorly standardized, difficult to utilize, and not very developed for all industries / segments).
Go back to 1999 and build a website supporting a million concurrent users. How much hardware do you need? What are your fixed costs? What are your engineering costs? How do you even market your thing?
The hardware cost is low, as is the barrier to entrance. And the users' expectation is higher.
To attract today's users, you still have to be better than your competitor's website. The fact you can easily build a website for millions doesn't mean it will actually see millions of visitors.
The fact that it's being trotted out again as justification should worry you.