GPU Cloud

Cloud GPU offers virtual GPU instances (vGPU) that can be flexibly scaled — ideal for variable workloads. GPU Servers are dedicated physical servers with exclusively assigned GPUs — optimal for continuous training with maximum performance.

We offer four NVIDIA GPU classes: Tesla T4 (16 GB GDDR6) as a cost-efficient entry-level GPU, A10 (24 GB GDDR6) as an all-rounder, A100 (80 GB HBM2e) for demanding AI workloads, and H200 (141 GB HBM3e) for maximum AI performance. Each GPU can be divided into different profiles — from small slices for inference to the full GPU for training.

The Tesla T4 (Turing, 8.1 TFLOPS FP32) is ideal for cost-efficient inference, VDI, and light ML workloads. The A10 (Ampere, 31.2 TFLOPS FP32) is an all-rounder for ML training, 3D rendering, and virtual desktops. The A100 (Ampere, 80 GB HBM2e, 312 TFLOPS FP16 Tensor) offers MIG isolation for demanding AI workloads. The H200 (Hopper, 141 GB HBM3e, 989 TFLOPS FP16 Tensor) delivers maximum performance for LLM training and large foundation models.

vGPU (Virtual GPU) allows a physical GPU to be divided into multiple virtual instances. Each vGPU instance receives dedicated GPU resources and VRAM. This way, you can book exactly the GPU power you need — without having to rent an entire GPU.

Full support for CUDA, cuDNN, and TensorRT. Compatible with all major ML frameworks such as PyTorch, TensorFlow, JAX, and ONNX Runtime. We provide pre-configured container images.

Yes, via Direct Connect you can seamlessly connect cloud GPU instances with your bare metal servers and colocation hardware — ideal for hybrid AI pipelines.

All resources are billed monthly. You configure vGPU, vCores, RAM, and storage individually and pay only for what you use. No minimum contract term.

Egress costs are fees that cloud providers charge for outgoing data traffic — i.e., for data leaving the data center toward the internet or other networks. Every API response, every download, every video stream, and every backup replication generates egress traffic. Most hyperscalers charge these costs per GB, which can quickly become a significant and hard-to-calculate cost factor. Especially with data-intensive applications, these fees add up rapidly.

No. At INGATE, outgoing data traffic is already included in the price — no per-GB fees, no hidden surcharges. This makes your costs fully predictable and transparent. Especially with data-intensive applications like CDN, streaming, large APIs, or backup replication, this results in a massive cost advantage over the major hyperscalers.

AWS charges approximately $0.09/GB (first 10 TB), Azure approximately $0.087/GB, and Google Cloud approximately $0.12/GB. A company transferring 10 TB per month pays approximately $900-1,200/month for outgoing traffic alone. At 100 TB, it is already $8,000-9,000+/month. These costs are difficult to predict as they depend on user behavior, API call volumes, and traffic patterns — complicating TCO calculations and frequently leading to unexpectedly high bills ("bill shock"). At INGATE, traffic is included, making total costs fully calculable from day one.