Playbook
Credits for AI startups
For an AI startup the highest-value programs are the cloud AI tiers and NVIDIA Inception, because compute is your dominant cost. Apply to the cloud programs first: their AI tracks are several times larger than their general tiers, and they stack with model-provider credits.
AI startups burn money differently. Most early-stage companies worry about salaries first and infrastructure second; if you are training or serving models, that order is reversed and it happens before you have revenue.
The programs below are the ones that address that specific cost. They are worth applying to in a deliberate order, which is what the playbook underneath is for.
The order to do this in
- 1
Start with the cloud AI tiers, not the general ones
Google, AWS and Microsoft all run AI-specific tracks that are substantially larger than their standard startup tiers. Applying to the general tier first can leave you holding the smaller allocation.
- 2
Add NVIDIA Inception in parallel
It is free, it does not compete with cloud credits, and it brings GPU access and hardware discounts that the cloud programs do not.
- 3
Then the model providers
OpenAI and Anthropic credits are smaller but stack cleanly on top, and they cover the cost the clouds do not if you are calling hosted models rather than running your own.
- 4
Check the expiry before you optimise the amount
AI credits typically run 12 to 24 months. A larger credit you cannot consume before it lapses is worth less than a smaller one you fully use, and this is the mistake that costs AI startups the most.
Which of these do you clear?
Five questions against the real eligibility rules.