Verda (previously presented as Takomo and linked to the DataCrunch brand) sells cloud infrastructure shaped around AI workloads. The portfolio spans compute, storage, networking and, more recently, a managed inference layer. Available accelerators include the newest NVIDIA data-center parts such as GB300 NVL72, B300 SXM6, B200 SXM6, RTX PRO 6000, H200, H100 and A100, alongside dedicated CPU instances with high clock speeds and NVMe block storage.
For developers, the platform emphasises self-service access: on-demand GPU instances can be provisioned without sales negotiations, instant clusters scale across multiple racks with InfiniBand or RoCE interconnect, and serverless containers autoscale for inference and batch processing. A newer Inference offering exposes frontier open-weight models through an OpenAI-compatible endpoint billed per token, with cached input pricing, so existing code that targets the OpenAI API can switch base URL and run unchanged.
The company highlights European ownership and data-center control, renewable energy, and a compliance posture that includes SOC 2 Type II, ISO 27001 and GDPR alignment, with confidential computing options for hardware-attested inference and fine-tuning. An in-house AI Lab contributes to open-source projects such as SGLang and vLLM, and publishes GPU benchmarks for prospective customers.
