Executive Summary: Dell AI Factory with NVIDIA

Dell has one of the most credible on-prem AI Factory infrastructure stacks in the market. Its credibility comes from physical infrastructure (compute at 0A, fabric at 0B and storage at 0C), data lifecycle integration (Layers 1A/1B/1C, where the Dataloop acquisition gives Dell its first proprietary software asset in the data lifecycle), and an AI ecosystem it validates rather than sells (Layer 3 reads gap). From the buyer's seat, the authority profile is just as legible: the enterprise retains commodity compute, cedes its storage and pipeline opinions to Dell, and inherits NVIDIA's judgment for everything GPU-aware above the rack.

Dell's capture is decoupled and therefore mostly invisible. The data and table formats are genuinely open: Iceberg, Trino, S3-interface object storage. Schemas lift and catalogs swap. That openness is true and reassuring, and it's the layer Dell's own teams point to when they push back on authority scores. The captive layers sit above the data: the MetadataIQ catalog structure, Data Orchestration Engine pipeline logic, PowerRack integration, and Spectrum-X fabric opinions. The buyer feels portable because the bytes are portable. The opinions aren't.

Dell's owned capability above the rack is thin by design: validated ISV blueprints and human services at Layer 2B, fleet management at Layer 2A. GPU scheduling judgment is NVIDIA's regardless of whose paper the license rides on. The deskside agentic runtime is pre-GA and watch-listed, not scored.

The reasoning-plane gap is no longer just this instrument's finding; it's Dell's stated strategy. In a July 2026 briefing, Dell confirmed both halves of the boundary: the telemetry exists from Layer 0 up, and the decision layer is deliberately left to the ecosystem. 'I can give you the telemetry, but the ultimate decision comes down to somebody making the decision.' Security is not governance. Security constrains who can access the platform. Governance constrains what the platform does. The Dell AI Factory has security. It does not yet have governance at the infrastructure level.

That does not make the AI Factory weak. It exposes where the next control-plane battle will be fought.

Layer-by-layer status: Layer 0A (PowerEdge HGX Servers and PowerRack the Enterprise Owns), Layer 0B (Spectrum-X and PowerSwitch, With Dell SONiC on the AI Back End), Layer 0C (PowerScale and Lightning FS, With PowerStore and ObjectScale), Layer 1A (MetadataIQ, a Metadata-Only Index; Classification Is Watch-Listed), Layer 1B (Elastic-Powered Retrieval), Layer 1C (Dell + Dataloop), Layer 2A (Fleet Management), Layer 2B (Blueprints + Services), Layer 2C (Enterprise Responsibility), Layer 3 (+1) (No AI Applications on Dell Paper; the Ecosystem Program Is Validation).

Assessment framework: 4+1 Layer AI Infrastructure Model. Scoring model: Decision Authority Placement Model (DAPM) — Retained, Delegated, or Ceded. Published by The CTO Advisor LLC (DBA The Advisor Bench). Author: Keith Townsend. Date assessed: October 27, 2026. Version: v2.43 - 4+1 v3: Layer 0 Sub-Layers.

Dell AI Factory with NVIDIA

Mapped to the 4+1 Layer AI Infrastructure Model

v2.43 - 4+1 v3: Layer 0 Sub-LayersAssessed October 27, 2026Sources & revision history
ACTIVE ASSESSMENT

Summary Finding

Dell has one of the most credible on-prem AI Factory infrastructure stacks in the market. Its credibility comes from physical infrastructure (compute at 0A, fabric at 0B and storage at 0C), data lifecycle integration (Layers 1A/1B/1C, where the Dataloop acquisition gives Dell its first proprietary software asset in the data lifecycle), and an AI ecosystem it validates rather than sells (Layer 3 reads gap). From the buyer's seat, the authority profile is just as legible: the enterprise retains commodity compute, cedes its storage and pipeline opinions to Dell, and inherits NVIDIA's judgment for everything GPU-aware above the rack.

Dell's capture is decoupled and therefore mostly invisible. The data and table formats are genuinely open: Iceberg, Trino, S3-interface object storage. Schemas lift and catalogs swap. That openness is true and reassuring, and it's the layer Dell's own teams point to when they push back on authority scores. The captive layers sit above the data: the MetadataIQ catalog structure, Data Orchestration Engine pipeline logic, PowerRack integration, and Spectrum-X fabric opinions. The buyer feels portable because the bytes are portable. The opinions aren't.

Dell's owned capability above the rack is thin by design: validated ISV blueprints and human services at Layer 2B, fleet management at Layer 2A. GPU scheduling judgment is NVIDIA's regardless of whose paper the license rides on. The deskside agentic runtime is pre-GA and watch-listed, not scored.

The reasoning-plane gap is no longer just this instrument's finding; it's Dell's stated strategy. In a July 2026 briefing, Dell confirmed both halves of the boundary: the telemetry exists from Layer 0 up, and the decision layer is deliberately left to the ecosystem. 'I can give you the telemetry, but the ultimate decision comes down to somebody making the decision.' Security is not governance. Security constrains who can access the platform. Governance constrains what the platform does. The Dell AI Factory has security. It does not yet have governance at the infrastructure level.

That does not make the AI Factory weak. It exposes where the next control-plane battle will be fought.

Strength
Moderate
Gap
Partner
Layer 0A · ComputeComputePowerEdge HGX Servers and PowerRack the Enterprise Ownsdecides: enterprise · Retained▼

CPUs, accelerators, servers, integrated systems, OS and hypervisor, physical plant

Vendor-Provided

PowerRackCeded

Factory-integrated, pre-validated rack-scale systems, documented in Dell's PowerRack data sheet (August 2026): PowerRack 5000 for 19-inch deployments and PowerRack 7000 for dense liquid-cooled 21-inch ORv3 racks, spanning rack-scale compute, networking and Exascale Storage with integrated PowerCool cooling. Integrated-system rule: the buyer's rack-integration opinions cannot be lifted to another vendor as deployed. The Integrated Rack Controller (power, thermal, firmware, device inventory) is the rack's management half, not a component of its own (Keith, October 16, 2026): its baselines are captive to PowerRack, which is part of why the rack can't be lifted as deployed.

Dell Pro Precision With GB300 (Deskside; Formerly Pro Max GB300)Retained

Deskside system on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, model FCT6263: a 72-core Grace CPU with 496 GB LPDDR5X and a ConnectX-8 SuperNIC, with Dell's MaxCool liquid cooling. Commodity deskside substrate, substitutable across OEMs.

PowerCool Rack-mount CDU (RCDU)Retained

Dell's rack-mount coolant distribution unit, which "provides up to 160 kW of cooling capacity," enough for GB300 NVL72 density, per Dell's power and cooling page. Physical plant accumulates no portable opinions, no lock-in surface.

PowerEdge XE9780 / XE9785 With NVIDIA HGX (8-Way HGX B300 or B200; iDRAC10)Retained

Standalone 10U air-cooled accelerated servers, with liquid-cooled L variants. The XE9780 takes eight NVIDIA HGX B300 or B200 GPUs, and the XE9785 takes "8 NVIDIA HGX B300 NVL8 270 GB 1100W SXM6 GPUs, fully interconnected with NVIDIA NVLink technology" on two AMD EPYC 9005 CPUs, per Dell's product pages. iDRAC10 is the embedded controller and reads with the server. A standalone server: a CUDA or PyTorch workload moves to another OEM's server without a rebuild, Retained.

PowerEdge XE9712 (NVIDIA GB300 NVL72)Ceded

Dell's GB300 NVL72 system: 1 RU sleds with "Two NVIDIA Grace with 72 cores per processor" and "4 Blackwell Ultra GPU's," in the IR9048 48 RU rack, listed as a current PowerRack compute option on Dell's integrated rack page. Ceded on the integrated-system rule, the Lenovo NVL-class reading: a rack-scale system built around a captive NVLink fabric can't be lifted as deployed.

PowerEdge XE9680 With NVIDIA HGX (8-Way HGX H200 or H100)Retained

Dell's earlier eight-way accelerator server, still on the XE spec sheet, which offers "your choice of GPU modules": here "8 NVIDIA HGX H200 141GB 700W SXM5 GPUs, fully interconnected with NVIDIA NVLink technology," or H100. A standalone server: a CUDA or PyTorch workload moves to another OEM's server without a rebuild, Retained.

PowerEdge XE9785 With AMD Instinct MI355X (8-Way OAM; iDRAC10)Retained

The XE9785 with "8 AMD Instinct™ MI355X 288GB 1400W OAM with AMD Infinity Fabric connectivity," per Dell's product page, on two AMD EPYC 9005 CPUs, with a liquid-cooled XE9785L. A standalone server on AMD's open ROCm stack: the workload moves to another OEM's MI355X server without a rebuild, Retained.

PowerEdge XE9680 With AMD Instinct MI300X (8-Way OAM)Retained

The XE9680 configured with "8 AMD Instinct MI300X 192GB 750W OAM accelerator with AMD Infinity Fabric connectivity," per Dell's spec sheet. A standalone server on ROCm: the workload moves to another OEM's MI300X server without a rebuild, Retained.

PowerEdge XE9680 With Intel Gaudi 3 (8-Way OAM)Retained

The XE9680 configured with "8 Intel Gaudi 3 128GB 900W OAM accelerator with embedded RoCE ports for ethernet connectivity," per Dell's spec sheet, a build that changes the server around it (eight PCIe slots, six 800 Gb ports). Read as the Intel row reads Gaudi 3, Retained.

NVIDIA-Provided

GPU/Accelerator Silicon

Blackwell and Blackwell Ultra, the compute engines Dell builds around.

NVLink / NVSwitch

The XE9712's GB300 NVL72 scale-up domain, captive inside the rack.

◆ Gap Analysis

Dell sells the compute the enterprise racks and runs. PowerEdge XE9780 and XE9785 are eight-way HGX servers with B300 or AMD MI355X, the XE9680 carries Intel Gaudi 3 and AMD MI300X, and the XE9712 is Dell's GB300 NVL72 system. PowerRack is a factory-integrated rack platform spanning compute, networking and Exascale Storage, and Pro Max GB300 sits at the desk. Dell's own engineering shows in thermal design and rack integration, with PowerCool CDUs and liquid cooling built for NVL72 density. The buyer gets current-generation accelerated compute on hardware it owns. The architect's concern is the integrated system. The eight-way servers lift to another OEM; the XE9712 rack and PowerRack can't be lifted as deployed. The XE9812 isn't scored until Dell publishes product documentation for it. Calibration: HPE, Lenovo and Supermicro read strong on the same OEM shape; AWS and GCP strong on compute the enterprise rents.

◆ Borrowed Judgment

Retained where the grade rests. A server the enterprise owns makes no runtime call on Dell's behalf; the enterprise's own hypervisor, scheduler and firmware baselines decide. Portability: the eight-way servers, Pro Max and the CDU read Retained; the XE9712 and PowerRack read Ceded on the integrated-system rule, and the rack controller is part of why.

◆ Working Notes

Watch-list (ISC, June 22, 2026, not scored): PowerEdge XE8812, Vera Rubin NVL4, up to 144 GPUs per rack with fanless direct liquid cooling, 'globally available early next year' (early 2027). Named: the 'Dell AI Platform with AMD' line (MI350P PCIe, ROCm and vLLM) is a separate set of SKUs. Named, not scored (October 22, 2026): PowerEdge XE9812 (Vera Rubin NVL72). Dell's March 2026 press release says it "will be globally available 2H 2026," and Dell's June 1, 2026 blog says PowerRack systems with XE9812 servers "were delivered to CoreWeave," but no product page or spec sheet documents it; Dell's XE-series spec sheet (October 9, 2026) doesn't list it. Any product documentation lifts it. PowerRack was moved here the same day and restored on Dell's PowerRack data sheet (last modified August 12, 2026). Named, not scored (October 23, 2026): PowerCool CDU C7000, announced in Dell's May 2026 press release for Q3 2026. Dell's power and cooling page lists only the rack-mount CDU, and the XE spec sheet doesn't list the C7000. Any product documentation lifts it.

Layer 0B · FabricNetwork FabricSpectrum-X and PowerSwitch, With Dell SONiC on the AI Back Enddecides: vendor · Ceded▼

Interconnects, switching and routing, network virtualization

Vendor-Provided

PowerSwitch SN6000-seriesCeded

NVIDIA Spectrum-6 Ethernet switches for the Spectrum-X AI back end, sold as Dell PowerSwitch. Through NVIDIA's network OS the adaptive routing and congestion opinions are Spectrum-X's and don't lift to Juniper or Aruba switching without rebuilding: Ceded. Dell's datasheet also offers Dell Enterprise SONiC on these switches, which reads on its own chip, Delegated.

PowerRack (fabric facet)Ceded

PowerRack's integrated networking, captive to the rack as deployed. Integrated-system rule at this facet.

PowerSwitch with Enterprise SONiC Distribution by Dell TechnologiesDelegated

Dell's licensed commercial SONiC distribution with enterprise feature additions and Dell support, running on PowerSwitch hardware and validated on select third-party switches. Licenses, binaries and docs ship through Dell Digital Locker. It carries Dell's AI back ends as well as general switching: Dell's AMD design runs the PowerSwitch Z9864F back end on it, RoCEv2 with PFC, adaptive routing and congestion control, and Dell's NVIDIA design comes in a version where NVIDIA Spectrum-4 SN5610 switches run Dell SONiC instead of NVIDIA Cumulus.

PowerEdge XE9712 GB300 NVL72 (NVLink facet)Ceded

The captive NVLink/NVSwitch scale-up fabric inside the XE9712's GB300 NVL72 rack. Integrated-system rule at this facet.

NVIDIA-Provided

Spectrum Ethernet Silicon

Dell brands and rack-integrates NVIDIA's switching silicon.

NVLink / NVSwitch

The XE9712's GB300 NVL72 scale-up fabric.

◆ Gap Analysis

Dell sells the AI fabric on its own paper, and its design guides lay it out as three networks. The GPU back end runs RoCEv2 with PFC and adaptive routing, on PowerSwitch Z9864F with Broadcom Thor 2 adapters in the AMD design and on NVIDIA Spectrum-4 switches with ConnectX-8 SuperNICs in the NVIDIA designs. A front end on PowerSwitch Z9432F or Spectrum-4 carries management, applications and PowerScale storage. Out-of-band management runs on its own switch. PowerSwitch SN6000 brings Spectrum-6 to the back end. The buyer gets an Ethernet AI fabric from the same vendor as the servers. The architect's concern is the roadmap behind it. The switching silicon and the Spectrum-X opinions are NVIDIA's, and they don't move to Juniper or Arista without rebuilding. SONiC is the exit, and it now reaches the AI back end: Dell documents its NVIDIA design with Spectrum-4 switches running Dell SONiC, and its AMD design runs entirely on Dell SONiC. Calibration: Lenovo and Supermicro read strong on Spectrum-X on their own paper; HPE strong on Slingshot it owns.

◆ Borrowed Judgment

Ceded where the grade rests. Spectrum-X makes the adaptive routing and congestion calls at runtime, with no per-decision override: vendor decides, visible, not overridable, Ceded. Portability: the SN6000 through NVIDIA's network OS and the PowerRack fabric facet read Ceded; Enterprise SONiC reads Delegated on an open network OS, on general switching and on the AI back end.

◆ Working Notes

Named, not scored (October 22, 2026): the XE9812's NVLink facet, with the 0A chip. Evidence: Dell design guides H20060.8 (August 2026), H04724 (July 2026) and H04688.1 (July 2026). The Spectrum-4 SN5610 switches in Dell's NVIDIA designs are NVIDIA's product, and Dell part numbers for them aren't confirmed, so they read on NVIDIA's row. Dell's own path on them is Dell SONiC, which reads in the Enterprise SONiC chip.

Layer 0C · StorageStorage SubstratePowerScale and Lightning FS, With PowerStore and ObjectScaledecides: vendor · Delegated▼

Arrays, storage servers and software, front-end protocols: where the data sits

Vendor-Provided

PowerScale (File Engine)Ceded

MetadataIQ integration. NeMo Retriever connector. pNFS 25% throughput improvement. PowerScale storage opinions (configs, tiering policies, performance tuning) are captive to the Dell platform: the enterprise cannot lift them to Alletra or VAST without rebuilding.

ObjectScale (Object Storage)Ceded

S3-compatible. S3 over RDMA. NVIDIA Omniverse integration. Palantir Ontology deploys here. The S3 API is its own chip. ObjectScale behind it, its buckets' lifecycle and policy engine and MetadataIQ, is Dell's and doesn't lift to another S3 store: Ceded.

PowerRack (storage facet)Ceded

PowerRack's integrated storage, captive to the rack as deployed. Integrated-system rule at this facet. Not scored within this chip under the GA evidence table: Exascale Storage block engine ("Block availability in 1H CY2027.").

Cyber Resilience (Built-in) (protection half: encryption, immutable snapshots, air-gapped backup)Ceded

The data-protection half of Cyber Resilience: encryption, immutable snapshots and air-gapped backup on Dell storage. Ceded, as the chip reads.

PowerStore (Unified Block and File, All-NVMe)Ceded

Dell's active/active NVMe array: block over FC, iSCSI, NVMe/FC and NVMe/TCP, file over NFS and SMB, vVols. Protection policies carry snapshot rules, a replication rule (async, metro sync) and a remote backup rule, assigned per volume, volume group, file system or NAS server.

PowerStore Front-End Protocols (FC, iSCSI, NVMe/FC, NVMe/TCP, NFS, SMB)Delegated

Standards the enterprise's hosts and switches speak; the array behind them is Dell's. Delegated, the Everpure, IBM FlashSystem and NetApp storage-networking reading. Protection policies, replication rules and the active/active engine stay in the PowerStore chip.

Lightning File System (Software-Defined Parallel File System)Ceded

Dell's parallel file system for extreme-scale AI, sold as software on PowerEdge R660 or R6715 nodes with E3.S NVMe and ConnectX-7 400GbE. Direct NVMe access, redirect-on-write and native indexing. Dell positions it above PowerScale and ObjectScale for clusters past roughly 16K GPUs or 4 TB/s.

S3 Front End on ObjectScale (S3, S3 Over RDMA)Delegated

The S3 API, which independent object stores implement. Workloads confined to the supported common S3 operations move by changing endpoints, credentials and provider-specific configuration; the data, policies and unsupported service features migrate or adapt. Delegated. S3 over RDMA rides the same API. The object store is the ObjectScale chip, Ceded.

Front-End Protocols on PowerScale (NFSv3, NFSv4, NFS Over RDMA, SMB, HDFS, S3)Delegated

NFSv3, NFSv4, NFS over RDMA, SMB, HDFS and S3, the protocols the PowerScale spec sheet lists, which independent implementations share: hosts and applications re-point to another server or target speaking the same protocols, and the data and provider-specific policies migrate with it. Delegated. OneFS and its policies stay in the PowerScale chip, Ceded.

NVIDIA-Provided

CX-8/CX-9 SuperNICs

Storage-side RDMA for GPU-direct access.

◆ Gap Analysis

Dell's storage substrate covers file, parallel file, block and object. PowerScale is scale-out file with pNFS, Lightning File System a parallel file system (generally available March 16, 2026), PowerStore block and file, and ObjectScale object behind S3. Immutable snapshots, encryption and air-gapped backup are built in. The buyer gets AI-throughput file storage from its server vendor. The architect's concern is the storage OS. OneFS decides layout, protection and rebuild, and those opinions stay on Dell's platform. Calibration: HPE reads strong on K3000 and E2000; VAST, WEKA and NetApp strong on storage alone.

◆ Borrowed Judgment

Delegated where the grade rests. OneFS places and protects the bytes, and SmartPools file pool policies send a path's files to a named node pool or tier, a per-path handle the engine must honor: vendor decides, visible, overridable, Delegated. Portability: PowerScale, Lightning FS and PowerStore read Ceded; ObjectScale's S3 interface and PowerStore's front-end protocols read Delegated.

◆ Working Notes

Watch-list (rolled July 12, 2026, not scored): Dell Exascale Storage (PowerScale, ObjectScale and Lightning FS on one platform), still 'available 2H 2026' in Dell materials and expanding to 4-in-1 with PowerFlex in 1H 2027. Out of the PowerRack chip under the GA evidence table (October 24, 2026): Exascale Storage block engine ("Block availability in 1H CY2027.").

Layer 1A · DataData Foundation & GovernanceMetadataIQ, a Metadata-Only Index; Classification Is Watch-Listeddecides: code · Retained▼

The data itself, organized and governed: the Governance Catalog that Layer 2C queries

Vendor-Provided

MetadataIQCeded

Exports OneFS ChangeListCreate output (path, file type, pool, cluster, timestamps, change type) to an off-cluster Elasticsearch index with Kibana, across clusters. Metadata only: Dell's documentation shows no content classification, sensitivity, or lineage. Metadata indexing opinions and index structure are captive to Dell.

Cyber Resilience (Built-in) (access half: Zero Trust, RBAC, data masking, XDR)Ceded

Zero Trust, encryption, RBAC, immutable snapshots, XDR, data masking, air-gapped backup. Security implementation captive to Dell storage platforms: policies and configurations do not lift to another vendor's storage.

NVIDIA-Provided

cuVS (Vector Search)

GPU-accelerated vector indexing (12x claimed). The GPU-accelerated path is not yet GA on Dell paper (watch-listed at 1B): dependency finding, not shipping capability.

CX-8/CX-9 SuperNICs

Storage-side RDMA for GPU-direct access.

NeMo Retriever Connector

PowerScale integration for GPU-accelerated retrieval.

◆ Gap Analysis

1A is graded on a catalog that classifies the data, so the grade turns on whether a reasoning plane can ask Dell's index what the data is, and today it can't. MetadataIQ exports OneFS changelist metadata (path, file type, pool, cluster, timestamps, change type) to an off-cluster Elasticsearch index across clusters, the same shape as WEKA's Data Catalog. Dell's documentation, through OneFS 9.15, describes file system metadata, not content classification, sensitivity or lineage. Access control and data masking from Dell's cyber resilience suite are 1A functions and read here; its snapshots and encryption protect the bytes and read at 0C. Trust3 AI does classify sensitive data on Dell storage, but Dell doesn't sell it. Moderate, at the top of the band.

◆ Borrowed Judgment

MetadataIQ indexes the paths the administrator configures, executing without inference: code decides, visible, overridable, Retained. Portability: the catalog structure and indexing opinions stay on Dell's platform, Ceded.

◆ Working Notes

Watch-list (announced October 6, 2026, not scored): the Dell AI Data Platform Unified Semantic Layer and Enterprise Knowledge Graph, 'set for release in the first half of 2027'. Named, not scored: Trust3 AI governance on Dell storage, a partnership, not Dell paper.

Layer 1B · RetrievalContext Management & RetrievalElastic-Powered Retrievaldecides: vendor · Delegated▼

Low-latency retrieval for RAG — vector/hybrid search, context windows

Vendor-Provided

Dell Data Search Engine (Elastic)Ceded

Elasticsearch 9.4. Hybrid keyword+vector search. MetadataIQ integration. LangChain support. Available now on Dell paper; the GPU-accelerated path is watch-listed pending Dell doc confirmation. The Elasticsearch API surface is its own chip. The service behind it, its operations, management and configuration, is Dell's and doesn't lift: Ceded.

Analytics Engine Agentic Layer + MCP ServerDelegated

Unifies vector stores across Iceberg, Data Search Engine, PostgreSQL+PGVector. Agent-queryable through the open MCP interface.

Elasticsearch API on Dell Data Search EngineDelegated

The Elasticsearch API is Elastic's interface, and the enterprise can also run it itself from the AGPLv3 source: code and definitions written to it move to a self-run deployment, the data moves by snapshot or export, and the service's control-plane configuration doesn't lift. Delegated. The service behind it is its own chip, Ceded.

NVIDIA-Provided

NVIDIA cuVS

GPU-accelerated hybrid search (12x indexing claimed). Upstream Elasticsearch 9.4 took cuVS to GA April 2026; not yet GA on Dell paper, watch-listed, dependency finding.

NVIDIA STX Architecture

BlueField-4 + ConnectX-9 + Spectrum-X + DOCA. Storage-side acceleration: available to ALL storage vendors.

NeMo Retriever

PowerScale connector for GPU-accelerated retrieval.

◆ Gap Analysis

Three-party dependency: Dell (storage + metadata), Elastic (search intelligence), NVIDIA (acceleration). STX is non-differentiating: every storage vendor has it. Dell's differentiation is MetadataIQ integration and the Elastic partnership, not NVIDIA acceleration. Frontier check (July 2026, strong pegged to the best shipping implementation): the 1B frontier is now integrated AI-retrieval platforms delivered on vendor paper: VAST InsightEngine natively and through the Cisco channel (strong). Dell's DSE is a general-purpose search engine (the general-purpose test favors its generality: Elasticsearch runs retrieval workloads Dell never anticipated) that the customer assembles into RAG, with the GPU-accelerated path still unconfirmed on Dell paper. Real, deployable, general, but below the frontier and gated: moderate holds because the GPU path is gated and strong is pegged to the best shipping implementation.

◆ Borrowed Judgment

Moderate, distributed across two partners. Search intelligence is Elastic's. Acceleration is NVIDIA's. Dell's durable value is the data substrate: if you swap search engines, PowerScale data doesn't move. From the buyer's seat the retrieval opinions themselves (Elasticsearch indices, queries, pipelines) lift to any Elastic deployment through the Elasticsearch API, a Delegated door; Dell's engine and its management read Ceded.

◆ Working Notes

No retrieval quality observability (recall@k, latency percentiles) that a Layer 2C could use for placement decisions. The July 2026 briefing's telemetry discussion was token/usage metrics, not retrieval quality. Watch-list (rolled July 12, 2026, flag sharpened, not scored): GPU-accelerated vector indexing for the Data Search Engine. Dell's own materials dated it 'available 1H 2026'; 1H 2026 has ended, and the Q2 2026 AI Data Platform administrators guide (April 13, 2026) does not confirm it. Upstream Elasticsearch cuVS went GA April 2026, but upstream GA never moves an OEM cell. This is now a slipped date, not a pending one: the top fact question for the next Dell briefing; scores when Dell docs confirm. Briefing (July 2026): knowledge graphs, graph databases, and unified streaming ingest named as AIDP capabilities: none documented GA; fact questions pending written feedback. Placement note: the instrument scores retrieval (DSE) at 1B and movement (DOE/DAE) at 1C; Dell's AIDP framing bundles them. Vendor-feedback posture: portability claims on pre-GA features are untestable and carry no DAPM weight. Written feedback should arrive with GA documentation. Rolled September 4, 2026 (Elastic row): upstream cuVS GPU indexing is GA in Elasticsearch 9.4 for self-managed Enterprise clusters only (unavailable on Elastic Cloud Serverless; CUDA 12.9 or 13, cuVS 2025.12 per Elastic's support matrix). Dell Data Search Engine is a self-managed packaging, so a Dell-side GA is possible on that path, but it still has to appear in Dell docs. Unconfirmed; no change. GA evidence table (October 24, 2026): Incremental Indexing goes to notes; only Elastic's May 2026 blog describes it, and Dell's 1.9.0.0 admin guide doesn't. The Data Search Engine is scored on that guide.

Layer 1C · PipelinesData Movement & PipelinesDell + Dataloopdecides: code · Retained▼

Move/transform data — ETL/ELT, lineage, cost-aware movement, KV cache tiering

Vendor-Provided

Data Orchestration Engine (Dataloop)Ceded

No-code/low-code AI data lifecycle. Dell's most meaningful software acquisition (~$120M, Dec 2025). GA confirmed: deployment documentation on Dell InfoHub and an orderable product page (verified July 2026). Proprietary pipeline orchestration: pipeline definitions built in the engine are captive to Dell.

Data Analytics Engine (Starburst)Ceded

GPU-accelerated SQL. Agentic Layer + MCP Server for agent access. Trino/Iceberg substrate. The Trino SQL surface is its own chip. Starburst owns the proprietary engine features and Dell Data Engines supplies deployment and lifecycle management; neither lifts: Ceded.

Trino SQL on the Dell Data Analytics Engine (Starburst)Delegated

Trino SQL has independent implementations: code and definitions written to it move to another compatible implementation within the features both support, the data moves by export, replication or rewrite, and the provider's control-plane configuration and operational state don't lift. Delegated. The service behind it is its own chip, Ceded.

NVIDIA-Provided

NVIDIA CMX

BlueField-4 powered context memory tier (G3.5). Pre-GA, watch-listed here and on the NVIDIA row; dependency finding, not shipping capability.

NVIDIA STX Reference Architecture

Storage-side infrastructure reference. Non-differentiating for Dell.

Blueprints, NIMs, AI-Q Blueprint

Pre-built pipeline components through the Marketplace.

◆ Gap Analysis

Dell's most significant strategic software move. Dataloop gives Dell proprietary orchestration logic: its first owned software asset in the data lifecycle. From the buyer's seat that ownership reads as decoupled capture: the data and table formats are open (Iceberg, Trino), so schemas lift and engines swap, while the pipeline opinions built in the no-code tooling are captive to Dell. The July 2026 briefing pushback made the instrument's case: open formats prove data portability, and data portability is the decoy: it says nothing about where orchestration authority accumulates. Decision authority: Dataloop carries the grade, and its pipelines run the nodes the enterprise built. Autoscaling, retry attempts, and timeouts are configuration, and cycles can be rerun with resume or restart, so the engine makes no runtime choice of its own. Code decides, visible, overridable, Retained, the judgment test and its defaults clause (October 4 and 5, 2026). Portability stays with the chips: the enterprise inherits Dell's orchestration opinions.

◆ Borrowed Judgment

From the buyer's seat: the enterprise inherits Dell's orchestration opinions outright at the engine (Ceded, first-party: Dell borrows little from partners there). Starburst's engine reads Ceded; the Trino SQL in front of it is a Delegated door. KV-cache judgment moved to the watch-list with the capability.

◆ Working Notes

NAND Research flagged maturity concern: young acquisition as enterprise orchestration engine vs. established Databricks/Snowflake. HyperFRAME: only 14% of orgs have AI-ready data architecture. Watch-list (announced GTC, March 2026; checked July 7, 2026, not scored): KV cache offload to PowerScale/ObjectScale/Lightning FS via NVIDIA CMX: 19x TTFT and 5.3x QPS are demo claims; CMX is pre-GA (watch-listed on the NVIDIA row) and Dell's language is staged availability 'throughout the year.' Scores when Dell docs confirm; the 'Context Moves to Storage' framing travels with it. Briefing (July 2026): unified streaming ingest named as an AIDP capability, not documented GA; fact question pending. The Orchestration Engine Marketplace is a catalogue of models, NIMs and blueprints; it doesn't move or transform data, so it isn't scored at 1C.

Layer 2A · OrchestrationInfrastructure OrchestrationFleet Managementdecides: code · Retained▼

GPU scheduling, quotas, RBAC, fair-share scheduling, utilization optimization

Vendor-Provided

OpenManage EnterpriseCeded

Infrastructure lifecycle management. Manages the chassis, not GPU workloads. Fleet templates, policies, and baselines do not lift to another OEM's management plane. A fleet management plane run on its own across many servers, so it's a 2A component (Keith, October 16, 2026). The embedded iDRAC in each server, and the Integrated Rack Controller in PowerRack, are the management half of that hardware and read with it at Layer 0.

Dell Container Storage Modules (CSM) OperatorDelegated

Dell's Kubernetes operator for installing and managing its CSI drivers and Container Storage Modules, the successor Dell names for the CSI Operator ("The Dell CSI Operator is no longer actively maintained or supported."). Dell's one K8s operator: storage provisioning, not compute orchestration. The consumed interface is CSI/Kubernetes, a genuine multi-vendor standard: provisioning opinions live in Kubernetes-standard objects (StorageClasses, PVCs) and survive a swap to another vendor's CSI driver without rebuilding the layer.

NVIDIA-Provided

GPU Operator + Network Operator + NIM Operator

Three of four K8s operators in the reference architecture are NVIDIA's.

NVIDIA Run:ai

GPU scheduling, quotas, fair-share. THIS IS the Layer 2A function. NVIDIA-acquired.

NVIDIA AI Enterprise

Commercial platform wrapping the full GPU orchestration and management stack.

◆ Gap Analysis

Dell manages the server fleet. NVIDIA manages the GPU-aware substrate. That distinction matters because AI Factory differentiation depends less on whether the rack can be deployed and more on how scarce accelerated capacity is scheduled, partitioned, licensed, and governed at runtime. ClearML provides floating NVAIE license management. Three authorities for one optimization function.

◆ Borrowed Judgment

High. GPU-aware orchestration primitives are NVIDIA-controlled. Dell's authority is limited to physical chassis management (Layer 0), storage provisioning (Layer 1A), and deployment automation (Day 0/1). No alternative GPU scheduler exists within the Dell AI Factory. The July 2026 briefing confirmed the posture: Dell positions everything above the rack as ecosystem, not owned capability. Decision authority: OpenManage Enterprise carries the grade (Keith, October 5, 2026) and runs the fleet templates, baselines, and policies the enterprise defines. Code decides, visible, overridable, Retained. The rack controller's firmware thermal and power logic now reads with PowerRack at Layer 0 (Keith, October 16, 2026): a controller embedded in one system is that system's management half.

◆ Working Notes

ClearML is the most interesting independent Layer 2A play. If Dell wanted proprietary 2A capability, acquiring or deep-partnering ClearML would be the most direct path. Pending (July 2026, not scored): Dell Automation Platform / private cloud briefing: the lifecycle-automation story (including the Nutanix-on-PowerStore private cloud landing July 2026) has not been briefed in over a year and may bear on this cell. Fact questions for that briefing: does the automation platform do anything GPU-aware itself (scheduling, partitioning, quota) or stop at deployment automation; who operates it day-2; can a customer run it without Dell services. The Dell CSI Operator chip is renamed to its successor, the Dell Container Storage Modules (CSM) Operator (October 24, 2026).

Layer 2B · RuntimeApplication Runtime & ExecutionBlueprints + Servicesdecides: model · Delegated▼

Model serving, agent execution, inference APIs, distributed inference

Vendor-Provided

Cohere North Blueprint on the Dell Agentic AI PlatformDelegated

Cohere North deployed on Dell infrastructure through Dell’s Agentic AI Platform. Dell holds the menu, not the capture: a production North estate’s orchestration opinions are Ceded to Cohere through any paper, so the channel reads Delegated (owner cedes, channel delegates).

Accelerator Services for Agentic AIDelegated

Dell's human-delivered value: strategy, deployment, optimization. Delegated, not Retained: the Delegated definition covers a genuinely substitutable delivery partner, and a different systems integrator could deliver the same outcome. The enterprise keeps the right to change providers without rebuilding the layer, which is Delegated by definition; Retained is reserved for a commodity substrate the enterprise operates or a layer no vendor has claimed.

DataRobot Blueprint on the Dell Agentic AI PlatformDelegated

DataRobot’s agentic AI platform deployed on Dell infrastructure through Dell’s Agentic AI Platform. DataRobot’s opinions are Ceded to DataRobot through any paper; the channel reads Delegated.

Mistral Blueprint on the Dell Agentic AI Platform (Mistral’s Full Stack Above the Dell Hardware)Delegated

Mistral’s full stack, everything above the Dell hardware (models, serving and Mistral’s agent platform), deployed through Dell’s Agentic AI Platform. Mistral’s opinions are Mistral’s through any paper; the channel reads Delegated.

NVIDIA-Provided

NemoClaw (OpenClaw Stack)

Open-source agent runtime (Apache 2.0). Alpha, not GA, watch-listed. Jensen: 'the operating system for personal AI.'

OpenShell

Sandboxed agent runtime with security/privacy controls. Spans deskside to data center. Alpha, not GA, watch-listed.

NeMo Guardrails

Runtime safety boundaries: what agents are NOT allowed to do. Constraint enforcement, not placement.

Dynamo

Distributed inference framework. KV-aware routing to cache-warm nodes. Closest thing to a placement decision in the stack, but single-variable optimization.

NIMs + AI Enterprise

Containerized model serving + commercial platform.

◆ Gap Analysis

What ships on Dell paper at 2B is exactly what a customer can buy and keep or swap today: validated ISV agent platforms delivered as Dell blueprints, and Dell's human expertise to deploy them. Dell owns no runtime software: model serving, agent execution, guardrails, and distributed inference are NVIDIA or ISV capabilities, carried in the dependency column. The NVIDIA deskside agent runtime (NemoClaw/OpenShell) is alpha and watch-listed, consistent with its treatment on the NVIDIA row. Dynamo's KV-aware routing remains the closest thing to placement reasoning in the stack: single-variable cache-locality optimization, not multi-variable policy.

◆ Borrowed Judgment

Total for the runtime the enterprise actually consumes: the NVIDIA path arrives with NVIDIA's judgment end to end: closed NIM serving decisions the enterprise can't inspect, Dynamo routing behavior, guardrail defaults. That judgment is inherited through the dependency column, not through a scored Dell component. Dell's owned contribution is curation and services; the blueprint menu keeps pre-purchase choice open (deployed platforms capture per-choice, to the ISV rather than to Dell), and Accelerator Services expertise transfers to the customer. Decision authority: the runtime is NVIDIA's, and it reads as it does on NVIDIA's row. The model makes the runtime call inside it, and the enterprise's own tool code, executing client-side and able to refuse, gates the effect: model decides, visible, overridable, Delegated. Portability is a separate reading, carried by the components.

◆ Working Notes

Jensen's 'OS for personal AI' is a Layer 2B claim. An OS manages execution. A control plane manages placement and policy. Watch-list (checked July 7, 2026, not scored): Deskside Agentic AI (Pro Max workstations + NVIDIA NemoClaw/OpenShell + Dell Services): the runtime is alpha (Apache 2.0), watch-listed on the NVIDIA row; scores when GA on Dell paper. Likely lands Delegated at GA (open substrate, self-hostable), noted, not pre-scored. The deskside hardware is scored at Layer 0 (Pro Max GB300, Retained); the services are scored here (Retained). Dell's written feedback on the 2B ratings is pending (July 2026 briefing). Out of the Agentic AI Platform chip under the GA evidence table (October 24, 2026): ClearML (Dell's Agentic AI Platform partner list names Cohere, DataRobot and Mistral blueprints; ClearML appears nowhere on it and no other Dell page was found (web search budget ran out).).

Layer 2C · ReasoningAgentic Infrastructure — The Reasoning PlaneEnterprise Responsibilitydecides: absent · Absent▼

Policy-driven placement and resource coordination — the Autonomy Layer

Vendor-Provided

NVIDIA-Provided

AI-Q 2.0 Reference Architecture

Multi-agent workflow scaffolding. Does NOT make placement decisions.

OpenShell Governance

Runtime security sandboxing. Layer 2B constraint enforcement, not 2C placement reasoning. Alpha, not GA.

Dynamo KV-Aware Routing

Performance-aware routing (single variable). Not multi-variable policy optimization.

◆ Gap Analysis

The 4+1 model defines Layer 2C as a required function of AI infrastructure: policy-driven decisions about where compute runs relative to data, which model serves which request, and how cost/compliance/latency are arbitrated in real time. Dell does not provide this function. The enterprise retains responsibility for it. Applying the 'Routing Is Not Reasoning' test: AI-Q 2.0 = workflow scaffolding. OpenShell/NeMo Guardrails = constraint enforcement. Dynamo = performance routing. None provides multi-variable policy-driven placement. The July 2026 briefing confirmed the gap in Dell's own words, twice. First, the telemetry-versus-decision boundary: Dell has telemetry from Layer 0 'all the way to 2B or even 2C,' token metrics per query, tools and guardrails, 'but the ultimate decision of what is economical or not comes down to somebody making the decision. I can give you the telemetry.' Dell provides the inputs to a reasoning plane and explicitly disclaims the decision layer. Second, the strategy: 2C and above is a deliberate ecosystem play: 'we designed the ecosystem program for 2C and 3 to give the flexibility and not try to be the winner of all, because we don't know who's going to win those layers.' The gap is a decision, not an omission. For the buyer that matters: bringing your own 2C (Kamiwaza, ClearML, potentially Palantir Ontology) collides with no Dell roadmap. The enterprise must build custom 2C logic (6-12 months), bring a partner, or operate without it. Most will choose option 3. The gap isn't visible until production agentic workloads expose it. Dell + Intel 'actively addressing' an AI-factory control plane (SiliconANGLE, May 2026) remains security/governance framing with no product announced, still not scoreable.

◆ Borrowed Judgment

There is no judgment to borrow: the enterprise retains full responsibility for this function. That responsibility is implicit: most enterprises operating Dell AI Factory infrastructure do not yet recognize Layer 2C as a distinct function they need to provide.

◆ Working Notes

Dave Vellante (theCUBE): 'The AI factory requires a new control plane, one that governs data, models and agents in real time.' That control plane is Layer 2C. Briefing (July 2026), vendor position recorded: 2C can't be productized because the intelligence is subjective: 'I can give you the tools, but I can't create the intelligence for you.' That argues the gap is universal, not that Dell fills it; it moves no cell. Scope clarification from the same exchange: the instrument rates infrastructure 2C (policy-driven placement: which model, which region, which cost/compliance tier), not domain reasoning; the GDPR-placement example is deterministic given telemetry Dell confirms it has. The capability spans 1B/1C (data boundaries) and 2C (agent/model placement), confirmed as the instrument's split. You can't prompt your way to deterministic output: the controls here answer whether an action is legal, never whether the outcome was right. That deterministic outcome-validator is missing instrument-wide, and like live placement it is noted as universal rather than charged to one vendor.

Layer 3 (+1) · ApplicationsAI Application Layer — The Value PlaneNo AI Applications on Dell Paper; the Ecosystem Program Is Validationdecides: absent · Absent▼

AI-powered business capabilities — business logic, workflow automation

Vendor-Provided

NVIDIA-Provided

NemoClaw / OpenClaw Runtime

Execution surface for Layer 3 applications. NVIDIA provides substrate; ISVs provide business logic. Alpha, not GA, watch-listed at 2B.

◆ Gap Analysis

Dell sells no AI business applications, and its AI ISV programs are validation. The Dell AI Ecosystem Program "provides AI software and platform providers access to Dell labs, validation tools and reference architectures" (Dell, May 18, 2026), and the named partners (Google, Hugging Face, OpenAI, Palantir, Reflection, ServiceNow, SpaceXAI) appear in collaboration language, not on Dell's paper: Palantir's Foundry and AIP is "coming on-premises to the Dell AI Factory" with no SKU or GA date. Dell Enterprise Hub is a curated catalog of open-weight models, which is model choice and read at 2B. The one independent software vendor (ISV) product Dell resells is CrowdStrike Falcon, with CrowdStrike providing support for purchases since February 2023, and Dell runs a managed detection and response service on Falcon; that's a single endpoint-security product, not an application program, and it's named, not scored. Layer 3's grade rests on business applications the enterprise can buy or get supported on the vendor's paper (as ruled October 19, 2026: business applications bought or supported on the vendor's paper). Certification isn't procurement. Calibration: Cisco, Arista and Nokia read moderate on first-party security applications; HPE reads partner on ISV applications customers buy from HPE on the GreenLake Marketplace; Nutanix, Red Hat, VMware, VAST and Supermicro read gap on the same certification-only shape.

◆ Borrowed Judgment

None on Dell's paper. The enterprise buys its applications from the ISVs; each brings its own governance domain (Palantir's Ontology, ServiceNow's workflows), and nothing on Dell's paper governs across them.

◆ Working Notes

Named, not scored: Dell AI Ecosystem Program (validation, May 18, 2026), Palantir Foundry and AIP on the Dell AI Factory (announced, coming), Dell Enterprise Hub on Hugging Face (2B), CrowdStrike Falcon resale and Dell MDR on Falcon (single security product; Dell KB 000125178, service description March 25, 2026), Fortanix and F5 (no Dell SKU found). 5,000+ AI Factory customers (Dell, May 18, 2026 release).

Sources & revision history · v2.43 - 4+1 v3: Layer 0 Sub-Layers

DTW 2026, GTC 2026, Dell press releases (March 16, May 18, June 22, 2026), Dell InfoHub documentation, Dell AI infrastructure product/marketing briefing (July 2026), published 4+1 model. v2.5 (briefing + DAPM axis reconciliation): DAPM re-reasoned from the customer's seat throughout: the rating measures the enterprise-to-vendor authority relationship; vendor-to-vendor framing (Dell-cedes-to-NVIDIA) retired to the nvidia[] dependency column and borrowedJudgment narrative. 2A Dell CSI Operator Ceded→Delegated (consumed interface is the multi-vendor CSI/K8s standard). GA-gate descores to dated watch-lists: 1A Exascale (targeted early 2H CY26, unconfirmed), 1C KV cache offload (NVIDIA CMX pre-GA), 2B Deskside Agentic AI (NemoClaw/OpenShell alpha). 1B GPU-accelerated DSE watch-listed: upstream Elasticsearch 9.4 took cuVS to GA April 2026, but upstream GA never moves an OEM cell. Dell-side GA unconfirmed in Dell docs. statusLabel convention ratified: capability vocabulary only; authority lives in the per-component DAPM chips. Summary rewritten: capture mechanism named (decoupled/invisible); 2C gap upgraded from instrument finding to vendor-confirmed deliberate ecosystem strategy. No layer status changed. v2.6 (July 12, 2026, codified-criteria audit): the canonical row re-read under the methodology's new capability grading criteria and channel-substitution rule. Trust3 Delegated→Ceded (proprietary governance platform; policies captive to the partner through any paper). 2B blueprint component re-reasoned to menu-altitude Delegated (deployed platforms capture per-choice, to the ISV). 1B held at moderate with the frontier check written in (DSE is general but below the integrated-retrieval frontier, GPU path unconfirmed). Watch-lists rolled: Exascale still 2H 2026 (now expanding to 4-in-1 with PowerFlex, 1H 2027); GPU-accelerated DSE flagged as slipped: Dell's '1H 2026' passed without doc confirmation. No layer status changed. /reconcile (September 1, 2026): NVL-class rack-scale systems moved Retained to Ceded under the integrated-system rule; multi-OEM availability is channel substitution, not authority. /reconcile (September 1, 2026): human-delivered services moved Retained to Delegated; a substitutable delivery partner is the Delegated definition. /reconcile (October 5, 2026): 2B authority re-read from {decides:vendor,visible:true,overridable:true,boundary:vendor,direction:Delegated} to model / visible / overridable / Delegated, matching NVIDIA's row for the same NVIDIA AI Enterprise runtime (the open-source seam and tool-call ruling, September 2, 2026). Components and grade unchanged. /reconcile (October 5, 2026): 1C authority column (rulings of September 15 and October 4 and 5, 2026). 1C authority vendor / Delegated to code / Retained: Dataloop pipelines run nodes the enterprise built; autoscaling, retries, and timeouts are configuration (docs.dataloop.ai service configuration), and cycles rerun with resume or restart. Grade unchanged. /reconcile (October 5, 2026): Layer 0 authority vendor / Ceded to enterprise / Retained: hardware the enterprise owns and operates makes no runtime tradeoff on the vendor's behalf (Keith, October 5, 2026). Grade and components unchanged. /reconcile (October 5, 2026): authority re-read against product docs under the judgment test, override rule, and Layer 3 ruling: layer1a to code / Retained. Grades and components unchanged. /reconcile (October 5, 2026): layer2a authority to code / Retained on the carrier Keith named. Grade and components unchanged. v2.16 (October 15, 2026, catalog-classification storage re-read): layer1a strong to moderate. MetadataIQ indexes OneFS changelist metadata only (Dell docs through OneFS 9.15), the WEKA Data Catalog reading. Trust3 AI moved from components to notes: Dell doesn't sell it, so it's a partnership, not Dell paper (Keith, October 15, 2026: whose paper asks whether Dell sells the thing). Unified Semantic Layer and Enterprise Knowledge Graph (first half of 2027) watch-listed. Authority unchanged. v2.17 (October 16, 2026, hardware-management scope ruling): a controller embedded in one system is that system's management half and reads with it; a fleet management plane run on its own across many systems is a 2A component (Keith, October 16, 2026, after AWS, Dell, and HPE pressure tests). Integrated Rack Controller moved from layer2a into the PowerRack chip at Layer 0. OpenManage Enterprise stays at layer2a as the fleet plane and keeps carrying the grade; statusLabel Rack-Level Management to Fleet Management. Grades, authority, and the other chips unchanged. v2.18 (Layer 0 sub-layers, October 16, 2026): Layer 0 split into 0A Compute, 0B Network Fabric and 0C Storage Substrate by runtime decision (Keith, October 16, 2026); storage left 1A for 0C, and 1A, now Data Foundation & Governance, is graded on the catalog that classifies the data. 0A strong, 0B strong, 0C strong; summary edits (summary worksheet, Keith, October 16, 2026). v2.19 (October 19, 2026, /reconcile partner ruling): Layer 3 partner to gap, authority Absent. The AI Ecosystem Program is validation, Palantir is announced, Enterprise Hub is a model catalog (2B), and the CrowdStrike resale is one security product, named, not scored (Keith). (Same pass) Plain-language pass: methodology rule numbers in the prose and audit notes replaced with the reasoning they stood for; no grade, chip or authority reading changed. v2.20 (October 19, 2026, /reconcile voice pass): em-dashes in the prose replaced with periods, colons, commas or parentheses; wording, grades, chips and authority readings unchanged. v2.21 (October 21, 2026, /reconcile): 0C door chip added for PowerStore's front-end protocols (FC, iSCSI, NVMe/FC, NVMe/TCP, NFS, SMB), Delegated, from reconcile's peer calibration: Everpure, IBM, NetApp and Nokia carry the same door on their arrays. Grade and cell authority unchanged; the cell rests on OneFS. v2.22 (October 22, 2026, /reconcile): the PowerEdge XE9812 (Vera Rubin NVL72) and PowerRack move to named, unscored notes at 0A, 0B and 0C. Dell's March 2026 press release ("globally available 2H 2026") and its June 1, 2026 blog (racks delivered to CoreWeave) are the only availability statements, and three Dell design guides from 2026 (H20060.8, H04724, H04688.1) don't document either product. 0B rebuilt on those guides: three networks (GPU back end, a front end carrying PowerScale storage, out-of-band); the SN6000 detail corrected ("800+ Tb/sec" was PowerRack's total, and Dell's datasheet offers Enterprise SONiC); Enterprise SONiC now documented on the AI back end (Z9864F in the AMD design, Spectrum-4 SN5610 in the NVIDIA design). Grades and authority unchanged. v2.23 (October 22, 2026, /reconcile): correction to v2.22. PowerRack is restored at 0A, 0B and 0C on Dell's PowerRack data sheet (last modified August 12, 2026), which documents PowerRack 5000 and 7000 as factory-integrated rack-scale systems spanning compute, networking and Exascale Storage; any product documentation lifts a product, and v2.22 missed it. Ceded on the integrated-system rule, as before. The XE9812 stays unscored. Found by the 0A compute research. Grades and authority unchanged. v2.24 (October 23, 2026, /reconcile): 0A adds the PowerEdge XE9712 with GB300 NVL72 (Ceded, on Dell's integrated rack page and XE spec sheet) with its 0B NVLink facet, and the XE9680 with H200, MI300X or Gaudi 3 (Retained, on the XE spec sheet). The PowerCool CDU C7000 moves to a named, unscored note (press release only), and the documented rack-mount CDU takes its chip. The Pro Max chip drops an undocumented parameter range and a break-even claim. The NVIDIA columns point at the XE9712 instead of the unscored XE9812. Found by the 0A compute research. Grade and authority unchanged. v2.25 (October 23, 2026, /reconcile): 0A: The XE9785 takes NVIDIA or AMD; the XE9780 is NVIDIA only. 0A: One model, three documented GPU-module configurations on three platforms. Found by the Layer 0 data pass after the sniff read. Grade and cell authority unchanged. v2.26 (October 23, 2026, /reconcile): Layer 1 research pass. 1C: "Orchestration Engine Marketplace" removed (a catalogue of models, NIMs and blueprints doesn't move or transform data). Found by the Layer 1 slot cut. Grade and cell authority unchanged. v2.27 (October 23, 2026, storage door pass): PowerScale gets a front-end door and ObjectScale splits into an S3 door and the Ceded store; Lightning File System documents no client protocol. Grade unchanged. v2.28 (October 23, 2026, Layer 1 managed-services pass): Dell Data Search Engine splits; Data Analytics Engine splits. Grade unchanged. v2.29 (October 24, 2026, /reconcile): 1B calibration updated: VAST reads moderate at 1B, InsightEngine unscored without a product guide. Found in the 1B dependency audit follow-ups. Grade unchanged. v2.30 (October 24, 2026, /reconcile): 1B calibration restored: VAST reads strong at 1B again, InsightEngine GA on VAST's own announcement. Grade unchanged. v2.31 (October 24, 2026, /reconcile): GA evidence recorded on 3 of 4 Layer 1B chips under the GA evidence table. Grade unchanged. v2.32 (October 24, 2026, /reconcile): Layer 1B under the GA evidence table version 2: GA evidence on every chip, preview items and unaffirmed chips to notes. Grade unchanged. v2.33 (October 24, 2026, /reconcile): GA evidence recorded on 2 of 2 Layer 1A chips under the GA evidence table version 2. Grade unchanged. v2.34 (October 24, 2026, /reconcile): GA evidence recorded on 3 of 3 Layer 1C chips under the GA evidence table version 2. Grade unchanged. v2.35 (October 24, 2026, /reconcile): GA evidence recorded on 4 of 4 Layer 0B chips under the GA evidence table version 2. Grade unchanged. v2.36 (October 24, 2026, /reconcile): GA evidence recorded on 9 of 9 Layer 0A chips under the GA evidence table version 2. Grade unchanged. v2.37 (October 24, 2026, /reconcile): GA evidence recorded on 8 of 9 Layer 0C chips under the GA evidence table version 2. Grade unchanged. v2.38 (October 24, 2026, /reconcile): GA evidence recorded on 1 of 2 Layer 2A chips under the GA evidence table version 2. Grade unchanged. v2.39 (October 24, 2026, /reconcile): GA evidence recorded on 1 of 2 Layer 2B chips under the GA evidence table version 2. Grade unchanged. v2.40 (October 24, 2026, /reconcile): GA evidence table version 2.1 applied: 3 chips recorded or narrowed. Grade unchanged. v2.41 (October 24, 2026, /reconcile): Stale text corrected from the GA backfill: the deskside GB300 renamed to Dell's current store name. Grade unchanged. v2.42 (October 24, 2026, /reconcile): Layer 2 bundle split: The Agentic AI Platform blueprints split by platform. Grade unchanged. v2.43 (October 27, 2026): each component now carries a durable offering id; nothing else changed.

4+1 Layer AI Infrastructure Model · Vendor Assessment Series · The CTO Advisor LLC (DBA The Advisor Bench) · thectoadvisor.com