{"resourceId":"semianalysis-rubin-vs-gb200-tco-unverified-20260723","versions":[{"version":"external-5645d0871c9dce3d28dc57914eaadae9d9316521e3b8b49b8f220bba807695a8","resource":{"id":"semianalysis-rubin-vs-gb200-tco-unverified-20260723","title":"SemiAnalysis InferenceX: CoreWeave early Rubin metrics not independently verified","organization":"SemiAnalysis / InferenceX","sector":"Independent semiconductor and AI-infra analysis","geography":"Global (analyst publication)","publishedAt":"July 23, 2026","publicationDate":"2026-07-23","eventDate":"2026-07-23","sourceName":"SemiAnalysis InferenceX","sourceLabel":"Independent research","sourceUrl":"https://inferencex.semianalysis.com/blog/vera-rubin-nvl72-vs-gb200-nvl72-inference","evidenceClass":"independent-research","outcomeClass":"cautionary","topics":["infrastructure","governance-procurement","developers-agents"],"finding":"SemiAnalysis’ July 23, 2026 InferenceX deep dive on Vera Rubin NVL72 vs GB200 NVL72 states that early VR NVL72 metrics come from CoreWeave and that SemiAnalysis has not independently verified them, while noting NVIDIA committed to submitting verifiable numbers to InferenceX by Q3 CY2026. The article analyzes architecture (including Rubin LUT-based tensor-core features), performance-per-megawatt and performance-per-TCO framings, and software-stack maturity, cautioning that leads versus Blackwell depend on baseline and operating point. Newly relevant amid Sep 2026 multi-rack/AgentX marketing that still rests on early partner measurements.","sledRelevance":"Supplies the independent caveat SLED buyers need when neocloud or vendor decks cite early Rubin tokens/MW: treat CoreWeave-origin figures as unverified until InferenceX/MLPerf-class submissions land.","evidence":"Originally published July 23, 2026; explicit “We have not independently verified them”; NVIDIA verifiable-numbers commitment by Q3 CY2026; architecture and TCO/perf-per-MW analysis; software runway discussion.","architectureImplications":"Separate silicon/architecture upside from production software maturity; pin evaluation to disclosed engines, interacting TPS targets, and power measurement methods.","governanceImplications":"Disallow unverified partner metrics as sole basis for capital or interconnect filings; require third-party or public benchmark commitments in contracts.","securityPrivacyImplications":"None primary; TCO analysis does not address confidential computing.","caveats":"Analyst publication with subscription/context; Q3 verification commitment may slip; not a lab report reproducing CoreWeave runs.","streamIds":["nvidia"],"roles":{"sales":"When a buyer cites early Rubin 10×-class tokens/MW, introduce SemiAnalysis’s non-verification statement before debating architecture. Bound the conversation to what will be independently submitted by Q3 and what on-site PoC metrics the buyer will accept—do not resell unverified partner charts as fact.","engineering":"Fit as a diligence filter on Rubin vs Blackwell TCO claims. Require disclosed workload, interactivity target, power metering, and engine versions. Proof of value is either InferenceX/MLPerf-verifiable numbers or a buyer-controlled replay harness—not slideware.","delivery":"Delivery implication: put verification gates in the PoC plan. Ownership: architecture lead. Risks: baselining against outdated Blackwell software; accepting partner-only telemetry."},"roleDetails":{"sales":{"customerProblem":"Buyers see large Rubin efficiency claims without knowing verification status.","stakeholders":["CIO/CTO","procurement","architecture"],"discoveryQuestions":["Which metric source is cited?","Is InferenceX/MLPerf verification in the timeline?","What PoC harness will you trust?"],"valueHypothesis":"Independent non-verification warnings prevent over-commitment while preserving architecture upside.","potentialEngagement":"Claim-diligence workshop mapping vendor charts to verification status.","unsupportedClaims":["SemiAnalysis validated CoreWeave 10×","Q3 numbers already public","TCO lead guaranteed at all TPS points"]},"engineering":{"fit":"Essential diligence for Rubin procurement evaluations.","architectureIntegration":"Compare like-for-like engines, power domains, and interactivity targets.","prerequisites":["disclosed configs","power measurement method","workload pins"],"constraints":["Unverified early metrics","Software maturity moving","Analyst framing"],"security":"N/A primary; keep evaluation data rooms controlled.","proofOfValue":["Verifiable submission or buyer PoC","Config pin documented","Sensitivity across TPS points"]},"delivery":{"work":"Build verification gates into PoC and acceptance.","dependencies":["benchmark owners","vendor SE","facilities metering"],"ownership":"Architecture/evaluation lead.","skillsAdoption":"Teams need benchmark literacy for agentic/TCO metrics.","governanceCheckpoints":["No CapEx on unverified charts alone","Q3 verification tracked","PoC ethics/data approved"],"acceptanceCriteria":["Verification path written","PoC metrics reproducible","Caveats in decision memo"],"risks":["Hidden baseline softwares","Metric shopping","Timeline slip on verifiable numbers"]}},"retrievedAt":"2026-09-22T03:12:35Z","enrichedAt":"2026-09-22T03:21:46Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"None stated.","procurementImplications":"Contract for verifiable benchmark deliverables; treat early partner metrics as informational.","operatingModelImplications":"Evaluation governance must track verification status as software stacks mature.","sourceVerification":{"openedUrl":"https://inferencex.semianalysis.com/blog/vera-rubin-nvl72-vs-gb200-nvl72-inference","referenceExcerpt":"The early metrics gathered on VR NVL72 come from CoreWeave. We have not independently verified them.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}