Considering the majority of existing HBM2 demand comes from the HPC sectors where price is far far less relevant than power efficiency, those customers will be wanting the cream of the silicon crop HBM stacks that come off the production lines at the foundry.
If all that demand falls at the far right of the bell curve for speed binned parts, there is current a large-ish inventory of rejected HBM stacks that are still too cost prohibitive for traditional mainstream consumer market segments.
If the customer demand at the top-end grows, then the foundry will struggle to keep up with the demand because the economics of having so many rejected stacks with no-one to sell to, means production volumes need to remain limited.
Say console manufacturers come along and ask for low speed binned HBM stacks, at over 10m a year.
All of a sudden, the foundry's economics look much better, and they can happily bring online new capacity allowing them to keep up with demand at the top end HPC segment, while providing enough low speed binned parts to the consoles.
it's similar I would say to how Sony negotiated double density GDDR5 chips for PS4 but took the hit on lower speed chips in order to get bumped to the front of the queue.
It's a real win win and the economies of scale bring down the costs of HBM for everyone.
TBH, low-spec. high volume console contracts are probably what the HBM-makers have been looking land since they commercialized the technology.
