From the NVIDIA edition of November 6, 2026
Accruex: GPU buyers should demand facility-level PUE/WUE, clean-energy evidence, and comparable power benchmarks
Accruex · Independent procurement guidance (AI compute sourcing) · Global guidance for GPU/compute buyers
- Publisher
- Accruex Insights
- Original publication
- 15 September 2026
- Source retrieved
- 2026-11-07
What happened
Accruex's Sept 15, 2026 buyer guide argues sustainability claims for AI compute must be facility-specific: ask for PUE and WUE of the site that will run the workload, standards and coverage behind clean-energy claims, comparability to MLPerf Power or AI Energy Score, emissions accounting including material Scope 3, and whether total energy for the workload class is rising or falling—not fleet-wide averages or chip-only efficiency stories.
Why it matters
Pairs with NVIDIA AI-factory efficiency marketing (including throughput-per-megawatt claims). U.S. SLED procurement can put these five questions into RFPs for on-prem or colocation GPU capacity. Limits: guidance essay, not an audit of NVIDIA or a specific Utah/municipal factory.
Evidence and measured results
Lists five due-diligence questions covering facility PUE/WUE, clean-energy standard/share, independent efficiency benchmarks, emissions boundary/Scope 3, and year-over-year total energy for the workload class. Emphasizes separating GPU efficiency from facility efficiency.
Limitations and uncertainty
Commercial advisory content; not a standards body. Does not validate any named provider's numbers. MLPerf Power / AI Energy Score coverage may not match a buyer's exact model.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-07; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- SLED buyers hear AI-factory efficiency claims without facility-level evidence.
- Stakeholders
- Procurement
- Sustainability office
- Facilities/energy manager
- CIO
- Discovery questions
- Can the bidder name PUE/WUE for the exact facility?
- What share of load is covered by the clean-energy claim?
- Will they report Scope 3 and year-over-year total energy?
- Value hypothesis
- Facility-specific diligence may prevent locking into efficient GPUs in inefficient halls—and creates comparable bid scoring.
- Potential engagement
- RFP language pack for sustainable AI compute diligence.
- Unsupported claims
- That any vendor already meets all five answers
- Numeric PUE targets in this essay
- NVIDIA-specific audit findings
Pre-sales engineering
Role takeaway
- Fit
- Applies to on-prem, colocation, and GPU-cloud buys for SLED AI factories.
- Architecture and integration
- Tie cooling/power design docs to the same facility ID used in the contract.
- Prerequisites
- Facility identity in the contract
- Access to PUE/WUE methodology notes
- Optional MLPerf Power comparable run
- Constraints
- Advisory essay not a standard
- Benchmarks may not match custom models
- Corporate averages still common in bids
- Security
- Facility packets often include physical access controls—review together.
- Proof of value
- Bid responses with facility-named metrics
- Contract clause for annual re-disclosure
Delivery
Role takeaway
- Work
- Insert five-question schedule into RFP and scorecard; verify evidence in BAFO.
- Dependencies
- Sustainability office review
- Facilities engineering validation
- Legal for disclosure clauses
- Ownership
- Procurement owns scorecard; facilities validates metrics; platform owns workload energy tracking.
- Skills and adoption
- Short briefing for evaluators on PUE vs chip efficiency.
- Governance checkpoints
- Approve RFP diligence schedule
- Approve annual reporting obligation
- Acceptance criteria
- Proposal: every eligible bid names facility PUE/WUE or is marked deficient
- Proposal: contract requires annual update
- Risks
- Vendors game averages
- Over-weighting green claims vs security/residency
- Essay treated as regulation
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Architecture choices (liquid cooling, density, location) show up in facility PUE/WUE; buyers should not score only GPU TDP or vendor tokens/watt slides.
Governance
Who approves, reviews and stays accountable for outcomes?
Make facility evidence contractual disclosure obligations, not brochure claims.
Security and privacy
What data, permissions and controls need testing?
Indirect—facility due diligence often travels with physical security and access reviews in the same vendor packet.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Procurement and sustainability officers need shared literacy with facilities engineers to interpret PUE/WUE answers.
Procurement
What should contracts, pricing and exit terms secure?
Drop the five questions into AI-factory and GPU-cloud RFPs; reject responses that only cite corporate fleet averages.
Operating model
Which teams own the service once it runs?
Ongoing reporting of facility metrics should be an operations KPI for shared state AI factories, not a one-time bid artifact.
Publication history
- 2026-11-06NVIDIA · Issue 626 resources
Stable resource ID: accruex-sustainable-ai-compute-pue-due-diligence-20260915