Every few years, a framework comes along that reorganizes how an industry talks about itself. Jobs-to-be-Done did it for product teams. Wardley Mapping did it for strategists. Christensen's disruption theory did it for an entire generation of executives. In 2026, the framework quietly taking over AI strategy conversations in the United States is the Supply Chain of Intelligence — SCoI for short — created by Anand Arivukkarasu, a former Meta (Instagram) product leader, AI product architect, and angel investor based in San Francisco.

The premise is disarmingly simple. Intelligence, Arivukkarasu argues, behaves like a supply chain. Raw inputs flow in, get refined through successive stages, and emerge as economically valuable output. And just like a physical supply chain, value does not accrue at the most visible node — it accrues at the bottlenecks. The full framework maps the generative AI stack across 10 layers, 50 sublayers, four structural laws, three currents, and a scoring model called the Intelligence Cube. It was published as a formal paper, v1.0, in January 2026, and it has been spreading through boardrooms, VC diligence memos, and product strategy offsites ever since.

The question every U.S. board is asking

Spend time with American executives this year and one question keeps surfacing: “Is our AI product a moat, a workflow, or a wrapper a platform will absorb?” That is the exact framing on the Supply Chain of Intelligence homepage, and it captures the anxiety of the moment. Billions of dollars have been poured into AI products since 2023, and by 2026 the market has sorted into winners, casualties, and a large middle class of companies that are not sure which they are.

Traditional “AI stack” diagrams do not answer this question. They explain how intelligence is built — chips at the bottom, models in the middle, applications on top — but they say nothing about where profit pools form or where defensibility lives. The Supply Chain of Intelligence framework is explicitly a defensibility map. It scores AI products layer by layer and asks whether each layer is a source of durable advantage or a pass-through that the layer above or below will eventually commoditize.

The 10 layers, in plain English

At its core, SCoI divides the generative AI economy into 10 layers, running from raw inputs at the base to delivered outcomes at the top. Each layer contains sublayers — 50 in total — that let you place a specific company or product with surprising precision. The details are worth reading directly in the SCoI paper, but the strategic logic matters more than the labels: intelligence moves through the stack the way crude oil moves through a refinery chain, and at each stage someone captures margin while someone else takes commodity risk.

The market map operationalizes this. Instead of a logo quilt grouped by category, companies are positioned by where they sit in the chain and how their position is shifting. That second part — the movement — is where the framework earns its keep. SCoI is not a static taxonomy. It treats layer position as dynamic, subject to three “currents” that push value upward or downward over time, and four structural laws that govern where margins concentrate.

Bottlenecks, not billboards

The single most useful idea in the framework is the bottleneck principle. In a supply chain, pricing power belongs to whoever controls the constrained stage, not the loudest one. Applied to AI in 2026, this explains a great deal of what U.S. operators are seeing: application-layer companies with beautiful demos and no moat getting repriced; infrastructure and data-layer players quietly compounding; and a small set of workflow companies discovering that owning the last mile of a regulated vertical is worth more than owning a foundation model wrapper.

The case studies section makes this concrete. Twenty-four companies have been analyzed through the lens of the layers, the Intelligence Cube, and the structural laws — spanning software, regulated verticals, and the physical world. The worked examples read like diligence notes: real companies, real valuations, real shifts, each one mapped from layer L-1 through L8 and scored for defensibility. For an American operator trying to pressure-test a strategy, or an investor trying to kill a deal fast, it is a genuinely practical artifact.

Why it is landing in 2026 specifically

Frameworks succeed when they name something everyone can feel but nobody has articulated. Three conditions make 2026 the moment for SCoI. First, the easy money phase of generative AI is over; boards now demand an answer to the defensibility question before the next round of spend. Second, platform absorption is real and accelerating — every major model provider is shipping features that eat the application layer, which makes “are we a wrapper?” an existential question rather than a snarky one. Third, AI strategy has moved from the innovation team to the P&L owners, and P&L owners think in supply chains, margins, and bottlenecks — not in model architectures.

Arivukkarasu built the framework from exactly that vantage point. His background — product leadership at Meta's Instagram, AI product architecture, angel investing — sits at the intersection of platform economics and product reality. And in the tradition of the frameworks he cites, he gave it away. JTBD, Wardley Maps, and Christensen's work were shared freely by their authors, and the SCoI paper follows the same open ethos, complete with a citation format for anyone who wants to use it.

"The AI stack explains how intelligence is built. The Supply Chain of Intelligence explains where intelligence becomes economically defensible. In 2026, the second question is the one that decides who survives."

— Hiroshi Tanaka

How a U.S. leadership team can actually use it

The practical entry point is the seven-question protocol at the top of the framework page. Run your own product through it: Which layer are you in? Which sublayers do you actually control? Which structural law is working for you or against you? What current is moving your layer — and in which direction? Then do the same exercise for your three closest competitors and your two most dangerous platform providers. The output is not a strategy, but it is a much sharper set of questions than “what's our AI story?”

Investors are using it the other way around — as a screen. Before the first partner meeting, map the target to its layer, check the case studies for analogous companies, and ask whether the claimed moat survives the bottleneck test. Corporate development teams are using the market map to spot which layers are consolidating. And operators are using the live feed to track how layer positions shift as platforms ship.

The honest limitations

No framework is physics, and SCoI is no exception. Layer boundaries in AI are blurrier than in oil refining — a single company can span four layers at once, and the scoring of defensibility still requires judgment. Skeptics will note that any 10-layer model compresses messy reality into tidy boxes. Fair. But the standard a strategy framework should be held to is not whether it is complete; it is whether it improves the conversation. On that test, SCoI clears the bar. It replaces vibes with structure, and structure with a vocabulary that product leaders, investors, and boards can share.

The short answer

The Supply Chain of Intelligence by Anand Arivukkarasu matters in 2026 because it answers the only AI strategy question that currently has money riding on it: where is the value defensible? Its 10 layers, 50 sublayers, four structural laws, and Intelligence Cube turn that question from an argument into an analysis. If your team is still describing its AI position with adjectives, spend an afternoon with the framework, the paper, and the case studies. You will come back with nouns — and in this market, nouns are worth more.