One package. One price. One accountable entity. Every layer of the stack — from bare-metal hardware through software, engineering team, monitoring, and executive reporting.
Assigned exclusively to your workload. NVIDIA Blackwell-generation GPU pods — scale and configuration confirmed in writing at engagement. No resource contention, no shared infrastructure, ever.
NVIDIA AI Enterprise full license included — CUDA, cuDNN, TensorRT, and the complete stack. MLOps tooling configured for your workload profile. Full-term LTS version freeze for deployment stability.
A dedicated engineering team assigned exclusively to your project — sized to your deployment configuration and confirmed at engagement. NVIDIA L3 minimum for field engineers. L4-certified lead on every deployment. No shared NOC, no split attention.
Weekly operational reports. Monthly SLA compliance documents. Quarterly executive briefings. Every incident logged, root-caused, and formally reviewed. Governance that matches institutional procurement requirements from day one.
Every item below is part of the engagement. No add-ons, no optional modules, no per-seat charges. One package, one price, one contract.
All packages below are built on infrastructure already in place under the primary engagement. No new procurement or separate vendor agreements required. Each is separately priced and added to the base MSA as an exhibit.
Optional packages are scoped and priced at engagement. Final availability confirmed in the applicable Master Services Agreement exhibit. All packages are delivered using infrastructure already contracted at Financial Close — no incremental procurement or vendor approval required.
Every SLA threshold is written into the Master Services Agreement. No ambiguity in what constitutes a severity, no discretion on response times.
| Capability | Verde Compute | Cloud GPU | Colo Bare Metal | Shared HPC |
|---|---|---|---|---|
| Hardware Generation | Blackwell architecture — Confirmed | Varies by region | Varies by provider | Mixed / unknown |
| Resource Isolation | Full pod dedication | Shared hypervisor | Depends on config | Queue-based sharing |
| Engineering Team | Dedicated · Exclusively assigned | Shared support tiers | Remote NOC only | No dedicated team |
| Software Stack | NVAI Enterprise, LTS frozen | Self-managed | Self-managed | Partial / outdated |
| Pricing Model | Fixed · Zero variable | Variable / per-hour | Fixed + extras | Variable / metered |
| Legal Structure | Delaware SPV per client | Standard EULA | Standard MSA | Academic / gov. |
| Executive Reporting | Weekly + QBR included | Dashboard only | Manual request | None |
Comparison reflects typical market offerings. Specific configurations may vary. Verde Compute terms are defined in the Master Services Agreement executed per engagement.
The base Verde package is complete — these modules are activated per client request at engagement or during the term. Each is a separately scoped and priced line item within the MSA Exhibit B. None are required for go-live.
Burst compute capacity via pre-negotiated AWS, Azure, or Google Cloud reservations — activated only during a declared DR event. Verde NOC triggers automatically upon P1 incident declaration. Client workloads fail over as quickly as infrastructure allows. Includes VPN tunnel and credential handoff procedures, tested in bi-annual DR drill.
A private 10Gbps (or 100Gbps) cross-connect from Verde's cage to the client's own network — bypassing the public internet entirely. Required for defense and government clients with classified or sensitive workloads. Verde manages installation coordination with the DC meet-me room and client's network carrier. Provisioned within 60 days of order.
Dedicated hyperscaler interconnect — AWS Direct Connect, Azure ExpressRoute, or Google Cloud Interconnect — enabling hybrid workloads that span Verde's dedicated cluster and a client's existing cloud environment. Useful for clients needing to move data from cloud storage to on-premises GPU compute. Verde activates and manages the physical cross-connect.
Verde's ML Infrastructure Architect integrates with the client's existing CI/CD pipeline — GitHub Actions, GitLab CI, Jenkins, or custom — to automate model versioning, deployment triggers, A/B testing, and rollback procedures on Verde's cluster. Includes integration testing with existing monitoring stack (Weights & Biases, MLflow, or equivalent). Verde provides a named ML Infrastructure Architect for this engagement.
Sector-specific compliance hardening for regulated clients. HIPAA: BAA signed, PHI handling procedures, encryption at rest confirmation. FedRAMP Moderate baseline alignment: NIST SP 800-53 control mapping, documentation package. ITAR-adjacent: export control review, personnel screening, data routing restrictions. Includes one compliance review per quarter by Verde's designated Compliance Officer.
An independent third-party hardware asset valuation and lifecycle report — suitable for lender requirements, insurance renewal, or client board reporting. Commissioned by Verde, prepared by a Verde-approved independent appraiser (e.g., CBRE, JLL Technologies, or equivalent). Covers fleet valuation, depreciation schedule, remaining useful life, replacement cost estimate. Delivered annually from Year 2.
A Verde Compute engagement begins with a confidential qualification process. Investment-grade organizations with multi-year commitment capacity may begin here.
All engagements subject to NDA, client qualification review, and executed MSA. US export controls and OFAC compliance apply. Submitting an inquiry is non-binding.
Every Verde engagement includes a minimum of 15 named engineers assigned exclusively to your project — listed by full name in the MSA Named Personnel Schedule (Exhibit A). Named engineers cannot be reassigned to any other Verde engagement without your written consent. Recruited and onboarded in parallel with SPV incorporation at contract signing.
Every assigned engineer is listed by full name, role, and seniority in the Master Service Agreement. This is a contractual commitment — not a commercial promise. If Verde wishes to reassign a named engineer to another engagement, Verde must obtain your written consent first. No exceptions.
All Verde Inc. employees sign a 24-month non-solicitation agreement as a condition of employment. Verde cannot poach your own AI team. Your client-facing engineers are protected from being recruited away by Verde during and for 24 months after your engagement.
Verde Compute engagements require an investment-grade institutional commitment — a single fixed monthly fee covering hardware, software, engineering, and operations. No per-GPU metering. No variable costs. No billing surprises. Exact pricing is confirmed in writing at engagement, prior to any financial obligation.