/blog / review / vultr
Vultr Cloud GPU review: A100s on tap, but read the fine print
Vultr's GPU offering is finally GA. We rented an A100 for two weeks and ran the suite.
- gpu
- review
- vultr
- a100
Vultr’s GPU instances are a real product now — A100 40GB and A100 80GB SKUs across most regions. We rented an 80GB instance for two weeks.
Performance
Roughly on par with the equivalent on Lambda or Runpod. Llama-3 70B inference at 128 tok/s, SDXL at 33 imgs/min. Disk was the slowest of the four providers we’ve tested.
The fine print
Egress is metered. There’s a 2TB-included tier and then $0.01/GB after. For an inference API, that ramps fast. We hit $30 in egress over the two-week test running synthetic load — not catastrophic, but worth pricing in.
The console is fine. The API is fine. Spin-up was ~3 minutes, slower than competitors. Not a disqualifier; not a draw either.
Verdict
7.0. Solid second choice if your stack is already on Vultr. Not a reason to migrate.
review · runpod
Runpod review: bare-metal H100s without the enterprise tax
Six weeks on Runpod across Community Cloud, Secure Cloud, and Serverless. The benchmarks, the bills, and where it falls short of AWS and Lambda.
14 min
review · lambda
Lambda Labs review: pretty UI, predictable bill, painful queue
Lambda is the boring-good option for ML teams that hate surprises. Here's what we'd warn you about.
10 min
review · paperspace
Paperspace Gradient review: Jupyter-first GPU rentals in 2026
DigitalOcean's GPU subsidiary, three years after acquisition. Still a good notebook tool, still expensive.
8 min