01What we do

We rent dedicated NVIDIA GPUs by the month, with root access, to people training and serving machine learning models. You get a private instance, a real GPU that is not time-sliced with anyone else, and full control of the operating system. Install PyTorch, TensorFlow, Jupyter, Docker, whatever your work needs.

Plans and specifications are on the pricing page. There are three: a single RTX card for solo work, an A100 for production training, and bare-metal configurations for heavier workloads.

02Why we exist

Renting a GPU from a hyperscaler is easy and expensive, and you rarely know what you actually got. Instances are frequently shared, throttled, or migrated. Egress is billed at rates that make moving your own data feel like a penalty. And when something breaks, you file a ticket into a queue.

Renting from a marketplace is cheap and unpredictable. Your job can be preempted, the host can vanish, and nobody is accountable for the machine.

There is room between those two: a dedicated card, at a fixed monthly price, from someone who will tell you exactly what hardware you are on and answer when you ask. That is the whole idea.

03How we are different

• We own the hardware. We are not reselling capacity from another cloud with a margin on top. The GPUs are ours, in a facility in Mumbai, and we know their thermal and power history.
• Dedicated means dedicated. When a plan says one A100, that card is allocated to your instance for the life of your subscription. It is not shared, partitioned, or oversubscribed.
• Your data stays in India. Instances run on our own hardware in Mumbai. For teams with data-residency requirements under the DPDP Act, that is a shorter conversation than it is with a US-headquartered provider.
• We do not train on your data. Your datasets, code and model weights are yours, and we do not use them to train our models or anyone else’s — Privacy Policy §2.3. We do not read them either, and we do not log into your instance unless you ask us to or there is a confirmed security incident — Privacy Policy §5.
• You reach a person. No tiered support, no ticket triage. The addresses on the contact page go to someone who works on the platform.
• Monthly, no lock-in. Cancel any time. If the first month is not what you expected, the 14-day guarantee refunds it in full, no explanation needed.

04What we will not do

Being small means being clear about limits rather than papering over them.

• We are not a hyperscaler. There is no global region list, no managed Kubernetes, no hundred-service catalogue. If you need those, use one.
• We do not hold compliance certifications we have not earned. No SOC 2 badge, no ISO logo in the footer. If you need an audited report for vendor review, we will tell you honestly what we do and do not have — see Privacy Policy §10.
• We will not host crypto mining. It is banned outright in Terms §5, along with attack infrastructure and the other things that get a network blocklisted. That restriction exists to protect the customers doing real work.
• We do not manage your operating system. Root access is yours, and so is patching it. We say so in Terms §7 rather than leaving it implied.

05Who runs it

VijayCloud is operated by Vijay, trading as VijayCloud, a sole proprietorship established in India and registered at 5th Floor, Technopolis Knowledge Park, Chakala, Andheri East, Mumbai, Maharashtra 400093.

It is a small operation, and we would rather say that than imply a company that does not exist. If you are evaluating us for something critical, ask us directly about capacity, redundancy and what happens when a card fails. Those are fair questions and we will answer them straight.

06Talk to us