Why is it hard to access OpenAI and Anthropic models from AWS, GCP and Azure?

From the last week I'm trying to access Anthropic and OpenAI models specifically Claude Sonet 5, Opus 4.8 and GPT 5.5 U from these cloud platforms but they block my account from even being able to request access.

I did kyc, deposited funds and even upgraded support from free to paid. Still no luck.

I contacted AWS support, it's been 5 days, no response and support ticket is still unassigned. I think they're on holiday. WTF.

AND GCP support basically told me to contact the sales team through their contact form and the sales team repsonse was to open a support ticket.

At this point I'm about to give up.

Is anyone out there able to access the models from these platforms? Please let me know what you did.

Cloud marketplaces add a procurement layer on top of the model provider, so the failure mode is different from using a normal API provider directly.

I would split the problem into three paths:

1. official cloud access through AWS/GCP/Azure - best for enterprise billing/compliance, but slow approval and regional availability can be painful
2. direct model-provider API - usually simpler if your account/country/payment profile is accepted
3. OpenAI-compatible third-party access - faster to test, but you need to verify provider legitimacy, base URL contract, model list, concurrency cap, setup timing, and refund/support policy

For your case I would not wait forever on an unassigned cloud ticket if the goal is just development/testing. Use the cloud route when procurement/compliance matters. Use a direct or OpenAI-compatible API route when you need to build this week.

Before paying any third party, ask for the boring proof points: /v1 base URL, API key delivery flow, supported model IDs, whether images/embeddings work, stated concurrency/rate limits, and whether setup is instant or manual.