Evoflex AI (Beta) - Hosted AI Workstation
AI workflows before scale - without cloud billing surprises
Hosted DGX Spark-class AI workstation in Malaysia for AI workflows, development, validation and staging
Where this fits in your AI workflow
Dedicated access to a DGX Spark‑class AI workstation, hosted physically in Malaysia.
— suitable for local LLM fine‑tuning and inference, development, and both short‑term and extended work periods prior to cloud scale or on‑prem deployment.
Decision & Constraints
Workflow Lifecycle
- Pre-purchase hardware evaluation
- Predictable cost before cloud scale
- In-country data residency requirements (Malaysia)
Workflow Lifecycle
- Development on large models beyond typical local hardware
- Workflow validation & prototyping
- Pre-deployment and staging environments
This is hosted infrastructure – not a GPU cloud, not a managed AI platform, and not a hardware purchase.
Evoflex AI provides an alternative to traditional GPU-as-a-service or cloud-based DGX Spark rentals, offering dedicated access on a hosted basis in Malaysia.
What you get
Dedicated & Isolated Access
Control & Connectivity
- Dedicated access — one customer at a time
- Malaysia-hosted — data remains physically in-country
- Private network access — systems are not exposed to the public internet
Control & Connectivity
- Full administrative control — install tools, libraries, framework, and models
- Clean reset between sessions — no shared or persistent environments
- Network connectivity — up to 1 Gbps download / 500 Mbps upload, no data transfer limits.
Hardware Specification
Software Environment
- GPU — NVIDIA GB10 (Blackwell architecture)
- Memory — 128GB unified LPDDR5x memory
- Storage — 4TB NVMe SSD
- Interconnect — NVLink-C2C CPU-GPU fabric
- Practical capacity — Supports large-parameter models (70B-class and above) for development, validation and staging workflows
Software Environment
Choose between a pre‑loaded AI environment for rapid setup, or a bare system for full control and customisation.
- Pre-loaded: CUDA-enabled AI Stack, container runtime, common frameworks
- Bare: Clean DGX OS with full administrative control
Connectivity Validation (Cyberjaya) · Maxis · Verified: 2026-05-07
Access & Pricing
Access model (Beta)
Pricing (Beta)
- Minimum booking — 3 Days
- Capacity during beta is limited and prioritised by use case
- Access is confirmed manually during the beta phase
- Access is provided on a dedicated, scheduled basis
Pricing (Beta)
- Beta access is offered at a reduced beta rate while the service matures
- Pricing reflects dedicated access to a full DGX Spark‑class system, not shared or usage‑based compute
- No usage‑based billing, metering, or variable cloud fees
- Terms may evolve as Evoflex AI moves beyond the beta phase
Engagements may be short‑term or extended (e.g. multi‑week or monthly), depending on availability and use case.
Unlike shared GPU cloud services, Evoflex AI provides dedicated, hosted DGX Spark-class access rather than on-demand GPU slices.
Designed for you
Evoflex AI is suited for those who need direct access to DGX Spark‑class hardware to develop and run large‑model workflows, with data hosted and retained entirely within Malaysia, and the flexibility to work for days or months without cloud constraints.
Talk to us. Check Availability
- Fill up the form and we will get in touch
- Email us: sales.ai@evoplus.com.my
- WhatsApp: Chat with us for quick questions
