If you have heard the phrase "supply chain of intelligence" in a board meeting, a diligence call, or a strategy offsite this year and nodded along without being entirely sure what it covers, this is the briefing. The Supply Chain of Intelligence — SCoI — is a framework published in January 2026 by Anand Arivukkarasu, a San Francisco-based product leader whose career runs through Meta's Instagram, AI product architecture, and angel investing. In the months since, it has become one of the more quietly influential documents in American AI strategy circles. Here is what you actually need to know, without the hype.

First, who is Anand Arivukkarasu?

Frameworks live or die by the credibility of their authors, so the biography matters. Arivukkarasu spent years as a product leader at Meta, working on Instagram at a scale where platform economics are not an abstraction — they are the daily weather. He has since worked as an AI product architect and angel investor, which puts him in the unusual position of evaluating AI companies from both sides of the table: as someone who has built products on top of platforms that could absorb him, and as someone who now writes checks into companies facing the same risk.

That vantage point shows in the work. SCoI is not an academic model and not a vendor whitepaper. It reads like what it is: a practitioner trying to answer a question he kept encountering in real meetings — where, exactly, does the money stay in AI?

What the framework actually says

The core claim is one sentence long: intelligence behaves like a supply chain. Raw inputs — compute, data, energy, talent — flow in at the base, get refined through successive stages, and emerge at the top as economically valuable outcomes. And as in any physical supply chain, pricing power does not live at the most visible stage. It lives at the bottlenecks.

The full framework maps the generative AI economy across 10 layers and 50 sublayers, from raw inputs at the bottom to delivered outcomes at the top. Layer position is not static: three "currents" push value up or down the chain over time, and four structural laws describe where margins concentrate. On top of the map sits a scoring model called the Intelligence Cube, which rates a company's defensibility rather than its demo.

The distinction that matters most is the one on the framework's homepage: a conventional "AI stack" diagram explains how intelligence is built. SCoI explains where intelligence becomes economically defensible. In a year when U.S. boards are asking whether their AI initiative is a moat, a workflow, or a wrapper, that is the question with money riding on it.

The five ideas worth remembering

You do not need to memorize 10 layers to use the framework. Five ideas carry most of the weight.

One: bottlenecks, not billboards. Value accrues to whoever controls the constrained stage, not the loudest one. Two: layer position is dynamic. The currents mean a comfortable position in 2025 can be a commodity position by 2027, and the framework forces you to ask which direction your layer is moving. Three: defensibility is scored, not asserted. The Intelligence Cube turns "we have a moat" from an adjective into an analysis. Four: platform absorption is a structural force, not bad luck — when the layer below you ships your product as a feature, the framework treats that as predictable, not tragic. Five: the map is meant to be used. The market map and the case studies — 24 companies analyzed through the lens so far — exist so you can test the ideas against real businesses, including your own.

Who is actually using it

Three audiences have picked up SCoI fastest in the United States. Investors use it as a screen: map the target to its layer, check the analysis archive for analogues, and ask whether the claimed moat survives the bottleneck test before the first partner meeting. Corporate strategy and development teams use the market map to watch which layers are consolidating and which are fragmenting. And product leaders use the seven-question protocol on the framework page as a structured way to pressure-test a roadmap — which layer are we in, which sublayers do we control, which current is moving us, and in which direction.

There is also a fourth, less visible audience: operators who simply track how positions shift as platforms ship. The site's live feed treats layer movement as news, which is a reasonable description of what platform feature launches have become.

Why it is spreading now

Frameworks catch on when they name something people can already feel. Three conditions made 2026 the moment for this one. The easy-money phase of generative AI is over, and boards want a defensibility answer before the next dollar goes out the door. Platform absorption has moved from a paranoid hypothetical to a quarterly event. And AI strategy has migrated from innovation teams to P&L owners — people who think natively in supply chains, margins, and bottlenecks rather than model architectures. SCoI speaks their language because it was built in it.

It also helped that Arivukkarasu gave the work away. In the tradition of Jobs-to-be-Done, Wardley Mapping, and Christensen's disruption theory — all of which he cites — the formal paper, v1.0, is published openly, with a citation format for anyone who wants to build on it.

The honest caveats

A fair briefing includes the limits. Layer boundaries in AI are blurrier than in oil refining; a single company can span four layers at once, and the Intelligence Cube's scoring still requires judgment. A 10-layer model compresses a messy reality into tidy boxes, and skeptics are right to say so. The standard for a strategy framework, though, is not completeness — it is whether it improves the conversation. On that test, SCoI replaces vibes with structure and gives product leaders, investors, and boards a shared vocabulary. That is a real contribution, even if the map is not the territory.

Where to start

If you have an afternoon: read the framework overview, then the paper, then two or three case studies from companies that resemble yours. Run your own product through the seven-question protocol, then do the same for your closest competitor and your most dangerous platform provider. You will not come back with a strategy — frameworks do not hand those out — but you will come back with sharper questions than "what is our AI story?" In the 2026 market, sharper questions are the asset.