Open any American AI strategy document from the last three years and the first visual is almost always the same: a vertical stack of boxes. Chips at the bottom. Cloud above them. Models in the middle. Applications on top. It is clean, it is teachable, and it is wrong in the way that matters most for people who have to bet money on it. A stack tells you how intelligence is assembled. It does not tell you where it becomes economically defensible.

That is the gap Anand Arivukkarasu set out to close with the Supply Chain of Intelligence — SCoI — a framework published as a formal paper, v1.0, in January 2026. Arivukkarasu, a former Meta (Instagram) product leader and AI product architect in San Francisco, argues that intelligence behaves less like a column of components and more like a supply chain: raw inputs flow in, get refined through constrained stages, and emerge as economically valuable output. The shift sounds subtle until you realize it rewrites every strategic question American operators are asking in 2026.

What the traditional AI stack leaves out

A conventional stack is a taxonomy. It answers "what are the pieces?" It does not answer "who captures the value?" or "who gets absorbed next?" Those omissions are expensive. In 2024 and 2025, U.S. investors and founders poured billions into application-layer companies whose entire position sat on top of someone else's model. The stack diagram made those companies look like legitimate layers. The supply-chain view made them look like surfaces waiting to be commoditized.

The Supply Chain of Intelligence framework adds the missing vocabulary. Gatekeeping: who controls the chokepoint a customer cannot route around? Absorption risk: which player above or below you can ship your product as a free feature? Bottlenecks: where is the constrained stage that actually captures pricing power? Currents: which way are demand, attention, and capital flowing across the whole chain? Flywheels: which loops compound over time? These are not decorative concepts. They are the forces that decide which AI companies survive 2026.

From parts list to economic map

SCoI maps the generative AI economy across ten layers and fifty sublayers, running from L−1 Resources — energy, grid interconnect, fabrication, critical materials, skilled trades — up through L0 Infrastructure, L1 Data, L2 Models, L3 Gatekeeping, L4 Access, L5 Execution, L6 Orchestration, L7 Surface, and L8 Memory. The numbering alone signals the argument: the traditional stack starts where the technology starts. The supply chain starts where the economy starts.

The framework groups these layers into three durability tiers. The Surface tier — mostly L7 and parts of L4 — is measured in weeks. It is what users demo, screenshot, and tweet about. The Workflow tier — L5, L6, and the stickier parts of L4 — is measured in months. It is what users live inside. The Substrate tier — L1, L3, and L8 — is measured in years. It is what users depend on. Value, in Arivukkarasu's phrasing, escapes the surface and accumulates in the layers below. A stack diagram cannot show that movement because it has no concept of time.

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 most visible one. Applied to American AI in 2026, this explains a lot of what the market is repricing. Model providers are constrained by chips and energy. Application companies are constrained by nothing except attention. The result is that the former keep capturing margin while the latter keep getting asked to prove they are not a feature.

The market map operationalizes this by plotting companies by layer position rather than by category. The classification page adds archetypes — fortress, refinery, surface, graveyard — that make the competitive logic explicit. For a U.S. leadership team, the exercise is bracing: map your company honestly and you may discover that the thing you were calling a moat is actually a billboard.

Gatekeeping and absorption risk

Two ideas separate SCoI from every stack diagram. Gatekeeping is the power to decide what passes through a layer. It can be a regulatory approval, an editorial judgment, a distribution right, or a protocol that others have to adopt. In the U.S. context, gatekeeping is why healthcare, finance, and defense-adjacent AI look structurally different from consumer AI. The same model capability is worth more behind a gate that customers trust you to hold.

Absorption risk is the mirror question: who can ship your product as a free feature, and what would it cost them? The case studies on the site walk through twenty-four real companies scored exactly this way — Jasper and Chegg as cautionary tales, Tempus and Deere as relocation stories. The pattern is consistent: companies that treated the stack as a stable architecture got absorbed; companies that treated it as a supply chain moved down, sideways, or up into defensible positions before the absorption happened.

The four laws and three currents

Underneath the map sit four structural laws. Intelligence commoditizes downward: what is scarce today becomes cheap one layer down tomorrow. Value accrues at bottlenecks. The surface captures attention, the chain captures power. Generation and verification must be separate — a law that creates durable positions in every regulated U.S. vertical, because someone has to be accountable for the check.

Three currents run horizontally across all ten layers: Demand Gravity, Attention Economics, and Capital Flows. A company can sit on a genuine bottleneck and still fail if the currents move against it. The live feed exists because layer positions and currents shift as platforms ship; the map is a snapshot, not scripture. For American investors, this is the fastest way to distinguish a structural position from a momentum trade.

The Intelligence Cube: same layer, different game

The traditional stack pretends that a layer means the same thing everywhere. The Intelligence Cube rejects that assumption. An L3 gate in consumer software is a speed bump. In healthcare or financial services it is a multi-year regulatory moat. An L5 execution layer that is trivially copyable in marketing tooling is nearly untouchable in a licensed trade with liability attached.

This matters for U.S. operators because the American market is stratified by regulation, procurement, and liability in ways that generic AI advice ignores. A strategy that is brilliant in one vertical is reckless in another. The Cube forces the question: in which market does my layer position convert into durability? That question usually produces better decisions than "which market is bigger?"

What changes for American strategy

Moving beyond the traditional AI stack means changing how U.S. boards and product leaders talk. Instead of "what is our AI story?" the question becomes "which layer do we actually occupy, which sublayers do we control, and which structural law is working against us?" Instead of "can we build it?" the question becomes "if we build it, who captures the value?" These are harder conversations, but they are the only conversations that matter once the easy-money phase ends.

The practical entry point is the seven-question protocol at the top of the framework page. Run it in order, in writing, before anyone proposes a roadmap change. Then test the output against the paper, the case studies, and the market map. The framework will not hand you a strategy, but it will hand you a set of questions specific enough that the answers can be wrong — which is more than most stack diagrams can say.

Where the supply-chain view has limits

No framework is physics. Real companies span several layers at once, and the durability tiers — weeks, months, years — are heuristics, not laws. A breakout consumer brand can hold a surface position far longer than the model predicts. The fifty sublayers invite false precision: placing yourself in L5b versus L6c can dress a judgment call up as measurement. And pre-product teams can paralyze themselves by over-mapping a position they have not yet earned.

But the standard for a strategy framework is not completeness. It is whether it improves the conversation. On that test, the Supply Chain of Intelligence clears the bar. It replaces a parts list with an economic map, and in 2026 that is the difference between describing AI and surviving it.

The short answer

The traditional AI stack is useful for explaining how intelligence is built. It is not useful for deciding where it becomes economically defensible. Anand Arivukkarasu's Supply Chain of Intelligence moves the conversation beyond the stack by treating intelligence as a supply chain: ten layers, fifty sublayers, four structural laws, three currents, and an Intelligence Cube that adds vertical context. For U.S. operators, investors, and product leaders, it is the map that makes the next round of AI strategy a set of answerable questions rather than a contest of adjectives. Start with the framework page, read the paper, study the case studies, and track how layer positions shift on the market map, classification table, and live feed.