Ask HN: How do products get priced after the bubble bursts? | Hacker News Reader

Ask HN: How do products get priced after the bubble bursts?

2 pointsAbstractH244 comments

For those who remember dot com, any insight into how product pricing changes after a bubble starts to burst?

While part of me sees a world where companies rapidly raise prices to at least be break even, another side sees a world where things from data centers to careers get liquidated at a fire sale. Wiping out investors & careers, lowering cost of doing business for those that survive the burst and keep costs low for end users.

Trying to think through what it’ll cost to use those AI products that remain standing when this is all over? Do they 20x overnight or does that still take a decade, after the Google & AWS of the 2020s truly emerges.

View on HN

If VC capital dries up, prices can no longer be subsidized and skyrocket. Many participants in a given space go out of business, so there’s less competition and less of a race to the bottom. Those that survive are the ones that price highly enough and can sell enough at that price to at least break even. Yes, there are liquidations, and drops in the cost of shovels, but generally of things that are already in surplus and do not significantly produce value.

Any downward pressure on prices stems from competition for customers’ shrinking discretionary budgets.

No one will pay 20x today's prices. I could see free tiers basically going away, subscriptions staying the same price but with lower limits, and tokens staying the same price because that price already has margin built in. I agree that data centers will be recapitalized just like fiber backbones were in 2001.

> I agree that data centers will be recapitalized just like fiber backbones were in 2001.

It’ll be interesting to see what the equivalent breakthrough to multiplexing will be

One which makes all of those data centers and GPUs not just bargain-priced relative to their original cost, but also way more efficient than originally expected.

For AI I think we'll continue to see a series of incremental efficiency improvements like flash attention, MoEs, disaggregated prefill, etc.

Reply on news.ycombinator.com