NVIDIA Engagement — One Page

Initial thoughts from Bliss + Dustin on how a working engagement could be shaped, per Greg and Lonny’s ask in the 2026-06-17 call. Problem-first, scope-light, designed to get to a real working session next week.


Problem

  1. Premise. AI silicon is the backbone of the next decade of the information economy. The buildout has outrun the service envelope around it — the layer that handles replacement, repair, escalation, and customer-state. NVIDIA is the first OEM to feel that mismatch at scale; it will not be the last.

  2. Problem. NVIDIA doesn’t have a coherent view of its commitments, what customers have installed in the field, or how either side is performing against the other. Greg: “We don’t even know the score of the game.” Lonny: “We come to the conversation, and we’re not even armed.” The data exists. It lives across SFDC, SAP, the NVIDIA data lake, customer reply-all chains, and 3PL spreadsheets, with no single source of truth and no shared view between NVIDIA and the customer.

  3. Why an arm’s-length operator is the right shape. A correct solution needs data NVIDIA cannot collect directly — customer-side performance, CM repair and dispositioning logs, 3PL shipment events, distributor inventory. Customers won’t share that with NVIDIA because they treat NVIDIA as a partial competitor. They will share with a neutral party. We build and operate that party. NVIDIA-sourced data remains perpetually NVIDIA’s.

  4. Target outcomes, twelve months. 95%+ SLA attainment on standard RMAs, measured for the first time. Hyperscaler RMA cycle time reduced 30%+ against current baseline. Customer-initiated escalations per open RMA cut by half. Stranded-shipment incidents (the recent Phoenix five-truck event) driven to zero.

Proposed timeline — six-week dual-track sprint from kickoff

Discovery and build run in parallel from week one, with effort shifting between them over time. Two named checkpoints anchor decisions: problem framing and initial prototype scope locked in week 3; prototype demos to NVIDIA stakeholders and customers in week 6.

Customer DiscoveryDesign / PrototypeDeliverable / Checkpoint
Week 1NDAs executed. Systems access provisioned. NVIDIA interviews (CSO Ops, IT Architect, Field Service, Demand Planning, Quality / FA). Initial data pulls.Information-architecture sketches based on week-1 conversations.Kickoff meeting with NVIDIA executive stakeholders.
Week 2Customer interviews (5–10 hyperscalers, neoclouds, etc.).Low-fi wireframes exploring 2–3 design directions.Provisional problem-synthesis document. Wireframes shown to Greg + Lonny.
Week 3Partner-firm interviews: CMs (5) and 3PLs (5). Cross-source synthesis.Design direction narrowed into concrete mockups. Engineering scaffolding bootstrapped.Checkpoint: problem framing and initial prototype scope locked with NVIDIA executives. Design direction approved.
Week 4Targeted gap-fill interviews as the prototype build surfaces new unknowns.Prototype built against real NVIDIA data.Informal “deskside” demos to NVIDIA point of contact for ad hoc feedback.
Week 5Validation follow-ups with customers already interviewed on implementation ambiguities.Expand prototype features based on week-4 ad hoc feedback.Informal “deskside” demos to NVIDIA point of contact for ad hoc feedback.
Week 6Final validation of prototype against 1–2 customers.Polish and debug. Proposed expansions for future development sprints identified.Checkpoint: prototype demos to NVIDIA stakeholders and select customers.

What we need from NVIDIA

Access to people

  • NVIDIA: Customer Service Ops lead (Lonny’s org), Field Service / on-site engineer, Demand Planning specialist, IT Solutions Architect (customer-service systems), Quality / Failure Analysis engineer.
  • Customers: warm intros to 2–3 hyperscaler hardware-operations or SRE leads (anonymized engagements OK).
  • Contract manufacturer: warm intros to 1–2 repair-line leads + the NVIDIA account manager at the CM.
  • 3PL: warm intros to 1–2 reverse-logistics operations leads + the NVIDIA account director at the 3PL.

Systems access (read-only, scoped to the RMA / customer-service domain)

  • SFDC case data.
  • Relevant SAP modules (RMA, inventory, fulfillment).
  • The NVIDIA data lake slice covering RMA, SLA, and customer state.
  • Current customer-facing web portal.
  • WMS/TMS rollout sandbox, once available.

Operational data

  • Customer breakdown by units deployed and by revenue.
  • Failure and RMA breakdown by customer and SKU.
  • Current SLA attainment baseline on standard RMAs.
  • 90-day escalation log.
  • Representative customer email and spreadsheet threads (anonymized) — the actual artifacts of today’s communication failure mode.

Sponsorship and people on NVIDIA’s side

  • Greg as executive sponsor, biweekly review cadence.
  • Lonny as day-to-day point of contact.
  • IT-designated point person for systems access and provisioning.
  • Legal point person to move the NDAs by Wednesday.

Anchor docs: latent-problem codification; fuller engagement structure. Source meetings: Greg + Lonny @ NVIDIA, 2026-06-17; Joe Malchow advice, 2026-06-17.