Intel Tiber AI Cloud (free Notebook tier)
Intel
Free JupyterLab and training-oriented notebooks, runnable on Intel Gaudi accelerators and Intel Data Center GPU Max, no credit card required to register
This is a standing free tier rather than a one-time credit. Sessions run as batch/shared queued jobs, not always-available standing instances, and don't support always-on GPU instances — that requires payment or redeeming a coupon to unlock. The tech stack is based on Intel oneAPI plus the PyTorch XPU backend, not CUDA, so running existing CUDA-specific code needs extra adaptation. Community feedback shows some confusion: an official Intel community post mentions that 'after migrating to the new Intel Developer Cloud Console, instances themselves no longer offer a free version,' while also confirming that JupyterHub batch notebooks remain entirely free for Standard/Premium accounts — the two statements need to be read together: 'free GPU instances' have been tightened, while 'free notebooks' remain available.
One of the few 'hardware vendor's own free cloud' examples outside NVIDIA, running the non-CUDA Gaudi/oneAPI ecosystem — suited to developers wanting to try an accelerator alternative to CUDA. Community accounts of whether the free tier has been tightened are inconsistent, though — after registering, check the console for what's actually currently available rather than assuming it matches past free-quota levels.