Research and Development
Operations
March 26, 2026
Applied R&D across five engineering domains — energy-compute integration, thermal management, GPU architecture, modular facility design, and remote NOC automation
5
R&D Domains
4
Products in Dev
4
Services in Dev
Q4 2028
Commercial Launch
01
R&D Scope & Focus Areas
Energy-Compute Integration
BTM hydro co-location, power delivery architecture, load-matching to minimum guaranteed output, zero-grid operation
Thermal Management & DLC
Direct liquid cooling with mountain river water (4–12°C), passive free-cooling, PUE < 1.11 target
GPU Infrastructure Architecture
NVIDIA Rubin Ultra GR200 NVL72 deployment, InfiniBand NDR800, NVMe storage, inference optimization
Modular Facility Design
Prefab seismic-rated containers, standardized site assessment, 3–6 month build cycle
Operational Automation (Remote NOC)
Unmanned facility operation — monitoring, alerting, automated failover, remote diagnostics without on-site staff
02
Product Pipeline
Active Dev
GPU Compute Platform
288 GPUs / facility, 3 revenue tracks (PPU, Reserved, Private AI Factory)
→ Launch Q4 2028
Architecture
API Layer
RESTful + SDK — provisioning, monitoring, billing
→ Beta Q3 2028
Requirements
Orchestration Console
Web dashboard — cluster monitoring, SLA tracking
→ MVP at Beta
Design
Facility Replication Kit
Prefab specs, site checklist, commissioning playbook
→ Validated Site 1
03
Service Pipeline
SLA Design
Dedicated Cluster Ops
Full-stack managed GPU infra, 1–2 clients/facility, SLA-backed
→ Primary Service
Concept
Inference Optimization
Workload tuning advisory on GR200 NVL72 architecture
→ CTO-delivered
Specification
Zero-Carbon Reporting
Automated carbon certificates, BTM verification, EU CSRD support
→ ESG Compliance
Process
Onboarding & Migration
Compatibility assessment, benchmarking, cutover planning
→ Anchor Clients
04
R&D Ownership — Distributed, No Dedicated Dept
CTO
GPU cluster architecture • API & platform dev • DLC cooling spec • InfiniBand + NVMe selection • Performance benchmarks & commissioning criteria
Head of Energy
BTM power delivery design • Load-matching methodology • DLC-river integration • Site assessments • Electrical safety & UPS specs
CEO
R&D priority setting • Facility standardization • Cross-team coordination • Replication methodology approval
Advisory Board
Architecture review • GPU roadmap intelligence (NVIDIA cadence) • BIS export controls & EU regulatory advisory
05
R&D Time Allocation
Pre-Launch (Current → Q2 2028)
CTO
Head of Energy
CEO
Post-Launch (Q4 2028+)
CTO
Head of Energy
CEO
06
Target Conferences & Events
NVIDIA GTC
GPU roadmap, CUDA, early access programs
Mandatory
GITEX Global
Dubai — sovereign AI pipeline (UAE, KSA)
Priority
OCP Summit
DC design, cooling, power efficiency, DLC standards
Target
Datacloud Congress
European DC industry, client intel
Target
SuperComputing
HPC, AI infra, networking benchmarks
Target
World Energy Congress
Renewables, hydro-compute positioning
Target
07
Additional R&D Activities
NVIDIA Partner Program
Early access to Rubin Ultra, pre-release testing, deployment best practices. Critical for Day 1 readiness.
Regulatory Monitoring
BIS export controls, EU AI Act, Georgian Virtual Zone regime. Ongoing compliance research.
Hydro Site Pipeline
Systematic assessment of Sites 2–5 in mountain Georgia. Criteria: output, river temp, connectivity, seismic.
Open-Source Engagement
vLLM, TensorRT-LLM, Triton Inference Server — benchmarking and potential contributions.
