What Is Layer 2C?

What Is Layer2C (the site)?

Layer2C is an independent AI infrastructure assessment resource from The CTO Advisor. It evaluates AI infrastructure vendors against the 4+1 Layer AI Infrastructure Model — a framework that maps where decision authority resides across the AI stack, from compute and networking through data, orchestration, runtime, reasoning, and applications.

The assessments are not paid rankings, certifications, or procurement recommendations. They are opinionated architectural analyses designed to help enterprise technology leaders understand where a vendor actually operates — and where the enterprise still owns governance, cost, evidence, and accountability.

Each assessment scores every vendor component using the Decision Authority Placement Model (DAPM): Retained, Delegated, Ceded, or Absent. The DAPM score answers a specific question: if the enterprise wants to change, audit, or override this capability, can it?

The 4+1 Layer AI Infrastructure Model

The 4+1 Layer AI Infrastructure Model is a framework developed by The CTO Advisor for analyzing AI infrastructure architectures. It divides the stack into eight layers grouped into four functional planes plus one application layer:

Layer 0

Compute & Network Fabric — GPUs, CPUs, networking silicon, and interconnects. The raw acceleration substrate.

Layer 1A

Data Storage & Governance — Durable storage, metadata, lineage, and the governance catalog that higher layers query.

Layer 1B

Context Management & Retrieval — Vector search, hybrid retrieval, RAG infrastructure, and context window management.

Layer 1C

Data Movement & Pipelines — ETL/ELT, KV cache tiering, data orchestration, and pipeline automation.

Layer 2A

Infrastructure Orchestration — Kubernetes, GPU scheduling, workload placement, and infrastructure control planes.

Layer 2B

Application Runtime & Execution — Model serving, inference engines, agent runtimes, and guardrails.

Layer 2C

Agentic Infrastructure — The Reasoning Plane — Policy-driven placement, cross-model routing, agent governance, and evidence capture.

Layer 3 (+1)

AI Application Layer — The Value Plane — AI applications, ISV ecosystems, and the interface between AI infrastructure and business outcomes.

The "+1" designation for Layer 3 reflects that the application layer is where business value is realized — it sits above the infrastructure stack rather than within it.

What Is Layer 2C Specifically?

Layer 2C — the Agentic Infrastructure layer — is where most enterprise AI governance failures happen. It is the reasoning plane: the infrastructure responsible for deciding how AI agents and models are orchestrated, constrained, monitored, and held accountable.

A vendor passes the Layer 2C test if its infrastructure can make policy-driven placement decisions — not just route requests, but reason about which model, agent, or resource should handle a task given current policy, cost constraints, compliance requirements, and available evidence.

Most vendors today have routing at Layer 2C but not reasoning. Routing is load balancing. Reasoning is governance. The distinction is whether the infrastructure applies enterprise policy or merely distributes traffic.

The Layer 2C gap is why enterprises find themselves with AI systems that are fast but ungoverned — capable of producing outputs but unable to explain, audit, or constrain them at the infrastructure level.

The Decision Authority Placement Model (DAPM)

DAPM classifies every assessed component on a single axis: where does operating authority reside?

Retained

The enterprise owns and controls this capability. It can audit it, modify it, replace it, or turn it off without vendor involvement.

Delegated

A substitutable partner provides this capability. The enterprise retains the authority to swap vendors. Authority is lent, not surrendered.

Ceded

The vendor controls this capability. The enterprise consumes it without governance authority. Changing it requires changing the vendor.

Absent

No capability exists at this layer. The enterprise must build it, buy it elsewhere, or operate without it.

DAPM is not a judgment about capability quality. A Ceded component can be excellent. The classification answers a different question: when your organization needs to change, audit, or override this capability, who has the authority to do it?

How Buyers Should Use These Assessments

Layer2C assessments are decision-support tools, not selection guides. They are most useful for:

• Understanding the architectural position a vendor occupies before a briefing or RFP
• Identifying which layers carry structural gaps or borrowed judgment — areas where the vendor's claim depends on a partner's roadmap
• Comparing authority profiles across vendors: which vendor maximizes Retained authority at which layers?
• Identifying the governance questions to ask in procurement conversations
• Stress-testing a vendor narrative against independent architectural analysis

Assessments should be read as a starting point, not a verdict. Architecture changes. Assessments reflect a point in time and a specific version of each vendor's platform.

How Vendors Should Use These Assessments

Vendors whose products are assessed are not clients of The CTO Advisor. Assessments are based on public materials, briefings, event coverage, analyst conversations, and independent judgment. Vendors may request factual corrections by contacting The CTO Advisor directly.

The assessment framework is consistent across all vendors. A vendor that wants to improve its Layer 2C profile should focus on the specific capabilities the framework evaluates: policy-driven placement decisions, cross-agent orchestration, evidence capture, and governance authority at the infrastructure layer.

Start Reading Assessments

The vendor assessment grid on the Layer2C homepage shows all assessed vendors with their DAPM profiles and layer-by-layer status at a glance. Each card links to the full assessment.

The cross-vendor comparison lets you see the layer status matrix and DAPM heat map across all vendors simultaneously.