From the cited report:

This post explains why a correction should be expected even if current valuations are rational...

And from the linked article:

When investors price in excessively steep growth paths, even strong actual AI performance can disappoint. Markets aren’t betting on good — they’re betting on perfect.

So unless straining all NAND production out to late 2027 for chips that haven't been made for data centers that haven't been built, to train on data that doesn't exist, is perfect in it's returns, there will be a much more drastic "correction" than dotcom. All while as the linked article points out:

Unlike 2001, central banks don’t have much runway to cut rates. Fiscal space across the euro area is already constrained by high public debt and existing spending commitments. The usual cushions are thin. That’s the detail that elevates this beyond routine caution — when the correction arrives, policymakers will have fewer options to slow the fall than they did during the dot-com bust.