The 5 layers of the AI trade: follow the money down the stack 

Why the layers matter now

The AI trade has pushed US shares back to the top. The S&P 500 has recently closed in the mid-7,700s, with the gains driven almost entirely by AI-related stocks. The index is up over 13% so far in 2026, while the Nasdaq Composite has climbed 17%.

But calling it “the AI trade” hides a lot. Every dollar a big tech company spends passes through several businesses before it turns into a chatbot answer. Each of those layers has its own drivers, its own profit margins and its own weak spots. Here’s the stack, from the bottom up.

Layer 1: silicon (the brains)

These are the graphics processing units (GPUs), custom AI chips and networking chips that do the heavy lifting. This is where the market looks first, and so far it hasn’t been disappointed. Nvidia hit a new intraday all-time high of $243.37 per share on Tuesday, and Marvell raised its ambitions sharply. Marvell now targets $70 billion to $90 billion in revenue in 2031, up from $8.2 billion in fiscal 2026.

What drives it: orders from big tech, new chip launches and chip export rules.

What breaks it: any sign that customers are working through the chips they already have rather than ordering more.

Layer 2: memory and plumbing (the nervous system)

Fast memory chips, data storage and the fibre-optic links that let thousands of chips talk to each other. This layer has quietly become one of the loudest. Micron is the clearest example. It is the only US-based maker of high-bandwidth memory (HBM), the fast memory that sits beside AI chips. On 30 September it reported fiscal fourth-quarter revenue of $54.2 billion, up 379% from a year earlier, and a non-GAAP gross margin of 87%, up from 45.7%. It guided to about $61.5 billion of revenue for the next quarter.

On Monday, Western Digital and Seagate rose 6% and 4% respectively, while Nvidia and Broadcom (which sit across multiple layers) each added 2%. The storage bounce followed a roughly 10% fall on Friday, after a report that Toshiba plans to double its data-centre hard-drive capacity. A day later, Ciena, Nebius and Astera Labs all posted big gains as enthusiasm returned to stocks tied to building AI infrastructure, even as Seagate and Western Digital fell 9% and 7%.

What drives it: pricing power. Memory costs are now big enough to shift big tech’s spending plans. In April, Meta raised its 2026 capital spending range to $125–145 billion from $115–135 billion, citing higher component prices, with memory singled out. Microsoft’s finance chief put $25 billion of its budget down to memory and chip costs.

What breaks it: memory has a history of boom and bust, and shortages tend to end in oversupply. The Toshiba plan shows how quickly new supply can be announced.

Layer 3: data centres and AI cloud providers (the body)

The landlords, builders, cooling specialists and companies that rent out AI chips. Think CoreWeave, Nebius, Vertiv and Equinix.

What drives it: signed orders waiting to be filled and how fast new capacity can be delivered.

What breaks it: financing costs. This layer leans heavily on debt, and the 10-year Treasury yield recently reached 5.365%, its highest level since 2002.

Layer 4: power (the heartbeat)

None of the above runs without electricity, and electricity is now the bottleneck. Higher power prices also feed back into hyperscaler or big tech considerations and could eventually affect the willingness of Layer 5 to spend. Find the full story in today’s companion piece, America’s Power Play.

Layer 5: platforms (the wallet)

Microsoft, Alphabet, Amazon and Meta are among the largest buyers. Together they plan to spend $720 billion to $745 billion in 2026. Alphabet alone expects to spend $195 billion to $205 billion, versus $91.4 billion in 2025, and it has raised that plan twice.

The catch is timing. Microsoft, Alphabet and Meta spent only 35% to 42% of their plans in the first half, so the second half has to be heavier. Cash flow is tight, too: Meta’s second-quarter spending of $31.1 billion came against $31.9 billion in cash generated by its day-to-day business.

The one thing that links all five

Every layer depends on Layer 5 continuing to spend. That makes the late-October results the key event. The Big Four tech giants report third-quarter results in late October, and any cuts in spending plans could ripple down the whole stack. The layers where a small change in sales can cause a big swing in profit would likely feel it most. Signs of weaker demand for AI services or margin pressure could also force a rebalancing of priorities.

The other risk to keep in mind is concentration. As Morningstar’s Lochlan Halloway put it, the market’s fortunes now rest heavily on a handful of companies and on the AI story continuing to deliver.

What to watch

• Big tech’s third-quarter spending plans (late October)
• Whether memory prices keep rising after Micron’s $61.5 billion guidance, and how fast new hard-drive capacity arrives
• Treasury yields: above 5%, layers that rely on debt feel it first

Stay tuned, this stack is just getting interesting!

Figures as of 8 October 2026. Past performance is not a reliable indicator of future performance. This content does not constitute investment advice.