Thanks for your Question @Kenneth .
If my understanding is correct, your question is more of around the maximum number profiles that can be stored / created in the Speaker Recognition (Speech Service). In case your question doesn't pertain to this. Request you ignore the below answer and request you to provide additional information to the question.
One subscription can create/contain 10,000 speaker profiles at most. Reference
You are likely to encounter the below error when you exceed the below limit :
{
"error": {
"code": "Forbidden",
"message": "Maximum allowed number of profiles is 10,000."
}
}
Coming back to your requirement of 70000+ profiles :
Unfortunately, currently there is no means to increase this 10000 profiles limit.
Having said that, the other possible solution/workaround I could think of is to have multiple subscriptions (in your case 7+ for 70000+ profiles) and map the profiles into a global pool (kind of Subscription-ProfileId Map) that the custom code would need to establish the tracking of where the profile was located in what subscription and resolve it accordingly.