The AI Stack: Why Where a Stock Sits Matters More Than Whether It Says "AI"

Index post for an 11-part series. September 2026.

Two companies both describe themselves as AI plays. One has a $163 billion order backlog stretching years into the future. The other has no revenue and a story about what happens in 2030. Both go up when the AI narrative is hot. Only one of them is still standing when it isn't.

That difference has almost nothing to do with how much AI is in the press release, and almost everything to do with where the company sits in the chain that makes AI physically possible.

The layer a company operates in tells you more about its risk than anything it says about AI.

That chain is real and it is sequential. Before a model can answer a question, someone generated the electricity. Someone fabricated the chip. Someone built and cooled the building. Someone trained the model. Someone wrapped it in software a business will pay for. Each of those steps is a different business with a different risk profile — different customers, different contract structures, different things that break.

This series maps that chain into 11 layers, sorts every major US-listed name into the layer it actually operates in, and ranks each layer from strongest to weakest forward setup.

What "strongest" means here

Not "went up the most recently." Every ranking in this series is built on four things:

• Contracted demand — backlog, signed deals, capacity already sold. What's locked in beats what's hoped for.

• Analyst consensus and price targets — where the professional estimates sit, and how much they disagree with each other. Wide disagreement is itself a signal.

• Balance sheet — can this company fund its growth, or does it need to keep raising money and diluting you?

• Near-term catalysts — what specific event in the next two quarters could change the picture.

Where the data is thin — small caps with three analysts covering them, or price targets that differ wildly between providers — that gets flagged as unreliable rather than dressed up as a consensus. Roughly two dozen of the names in this series fall in that category. Knowing which ones is part of the point.

The 11 layers

The strongest setups across all 11 layers

If you read nothing else in this series, these are the names that came out best on the four criteria above — visible demand, workable balance sheets, and real analyst support rather than a story about the future. Full reasoning and the data behind each sits in its layer post.

• GE Vernova, Quanta, Eaton — the physical grid. Backlogs are contracted years out; GE Vernova pulled its $200 billion backlog target forward a full year.

• Micron, Western Digital — the memory cycle, with capacity reported as effectively sold out into 2027.

• Broadcom, Marvell, Astera Labs — custom AI silicon and the connections between chips. Broadcom's AI semiconductor revenue grew 221% year-over-year last quarter.

• Vertiv, Digital Realty, Equinix — the buildings and the cooling. All three posted record bookings or backlog recently.

• Alphabet, Meta — the two hyperscalers where analysts see the most room left after September's pullback.

• Caterpillar — the market has quietly re-rated it from a construction cyclical to an AI-infrastructure supplier.

Notably absent from that list: most of the names that get the loudest coverage. That is not an accident, and the layer posts explain why in each case.

If you only trade IDX

Most names here are US-listed, and not every reader has US market access. The framework still does work for you. The same question — where in the chain does this company actually sit, and is its demand contracted or merely hoped for? — applies to any market. Indonesian investors looking at energy, metals, and infrastructure names on IHSG face the same sorting problem: which companies have signed offtake and which have a press release. Layers 1 and 11 in particular map onto themes that exist on the local exchange.

Catalysts worth watching, next two quarters

• Earnings: Micron (Sept 30), Vertiv (Oct 20), Intel (Oct 21), Palantir / Arista / Super Micro (Nov 2), Qualcomm (Nov 3)

• Milestones: Rocket Lab's Neutron first launch, targeted for Q4 2026 — the company itself has described the window as narrowing

• M&A: Core Scientific's acquisition by CoreWeave closing — this is a merger situation tied to CoreWeave's share price, not a standalone bet

• Commodities: copper (FCX, SCCO), lithium (ALB, LAC), and rare-earth policy (MP, USAR, CRML, TMRC) can move these names regardless of how well the companies execute

Disclosure and method

The author holds positions in some of the companies discussed in this series, and those positions change over time. Individual holdings are not disclosed. No compensation has been received from any company mentioned, and nothing in this series is sponsored.

Rankings are based on publicly available analyst consensus, company filings and earnings releases, and reported guidance, gathered in September 2026. Analyst price targets are estimates and are frequently wrong. Market data moves; each layer post carries its own as-of date, and where a company's situation changes materially, that post will be revised rather than left standing.

This is research and general information, not personalized investment advice. Nothing here accounts for your financial situation, time horizon, or risk tolerance. Do your own work before acting on any of it.

Series index. Each layer post links back here, and this page updates with the link as each one goes live.