GPU Rental Prices
SearchCtrl K

GPU rental for students: free credits and the cheapest hours

A student usually needs a GPU for hours, not months: a course project, a model for a thesis chapter, a paper's experiments. The cheapest on-demand GPU on our index is $0.10/hr on the RTX A4000, so a weekend of work on a small card costs little. Free programmes help most when they fit the work, so check three things before signing up.

Three questions before you sign up

  • Does it give GPUs at all? A general cloud credit does not always reach GPU machines. Google Cloud's free trial cannot add GPUs until you activate a paid billing account, and Azure for Students says nothing about GPU machines on its page, so check the quota in the console before you plan around it.
  • Does it need a credit card? Azure for Students does not; most welcome credits and pay-as-you-go accounts do. A card on file also means the bill continues when the credit ends.
  • When does it end? Credits expire: some after 30 or 90 days, some renew each year while you are a student. Spend them on the runs that matter, not on a machine left idle.

Make a small budget go far

  • Size the card to the model. Most course projects fit a 16 or 24 GB card; check yours in the LLM VRAM calculator before renting an 80 GB one.
  • Stop the machine when the run ends. Some providers bill a stopped machine at its full rate; the fees page shows what each charges while stopped.
  • Use interruptible capacity for runs that can restart. Spot and community rates cost less for the same card (spot, on-demand or serverless); save checkpoints so a lost machine costs minutes, not the run.
  • Ask your department first. Universities often run their own GPU clusters or hold cloud agreements for courses; that time costs you nothing.

The programmes below are read from each one's own page, with the day they were checked.

Programmes a student or researcher can use

Azure for Students: $100 credit (12 months), renewable each year while a student

Full-time university students signing up with a school email; no credit card. One account per person. Its page says nothing about GPU machines, so check the quota before counting on one.

Checked 31 October 2026 · source

Modal for Academics: Up to $10,000 compute credits

For researchers; Modal takes sign-ups for its grant programmes on the page, with no published criteria or expiry. "Up to" is a ceiling, not a promise.

Checked 31 October 2026 · source

Modal: $30/month recurring free compute (Starter), $100/month (Team)

Renews monthly on the plan; usage and concurrency limits differ by tier. The only evergreen recurring credit on this list.

Checked 31 October 2026 · source

Google Cloud: $300 welcome credit (90 days)

New accounts. GPUs cannot be added during the free trial: activate a paid billing account first; unused credit still ends 90 days after sign-up.

Checked 31 October 2026 · source

AWS: Up to $200 credits for new customers (6 months)

$100 at sign-up plus up to $100 through exploration activities. The free plan is limited to select services and the account closes after 6 months or when the credit runs out, whichever comes first.

Checked 31 October 2026 · source

Microsoft Azure: $200 credit (30 days)

Eligible new users; sign-up needs a phone number and a credit or debit card. Credit does not carry over past 30 days.

Checked 31 October 2026 · source

Oracle Cloud: $300 credit (30 days) + Always Free tier

One free trial per person. Credits expire at 30 days or when consumed.

Checked 31 October 2026 · source

Every offer we track, with who can apply: free GPU credits.

The cheapest cards with 16 GB or more, by the hour

On-demand rates checked 5 November 2026.

  • RTX A4000 (16 GB): from $0.10/hr at TensorDock; ten hours cost $1.00.
  • Tesla V100 SXM2 (16 GB): from $0.17/hr at TensorDock; ten hours cost $1.70.
  • RTX 3090 (24 GB): from $0.20/hr at TensorDock; ten hours cost $2.00.
  • RTX 2000 Ada (16 GB): from $0.24/hr at RunPod; ten hours cost $2.40.

Compare your offers

Saved prices keep their original dates. Confirm the configuration, terms and availability at the source.

Select up to three offers to compare their rates and terms.

Type a GPU, a provider, a model or a question.