NVDA Earnings Preview: Follow the CapEx

Hi {{first_name|Investor}} -

NVIDIA ($NVDA) doesn't just report its own earnings. It reports the health of the entire AI spending cycle.

NVIDIA's valuation is a derivative bet on how much the largest technology companies in the world spend on AI infrastructure. And here's what at stake going into this print: NVIDIA's stock has been roughly flat over the past eight months — despite hyperscaler CapEx guidance moving up roughly 50% during that same period. The market already capitalized the spending wave.

So the question isn't "will NVIDIA beat?" They almost certainly will. The question is whether their guidance and commentary bridge the $600B+ CapEx narrative into multi-year NVIDIA revenue — or leave a gap the market hasn't priced for.

Let's break down what to watch and why it matters.

Before you evaluate anything NVIDIA says, you need to understand the money flowing into the system.

The Magnificent 7 — Apple, Microsoft, Google, Amazon, Meta, NVIDIA, and Tesla — have collectively ramped capital expenditure by over 60% year-over-year. The core driver is AI infrastructure: data centers, networking, and the GPUs that power all of it.

Current estimates point to roughly $600–650 billion in annual CapEx from this group during the current AI build-out period, heavily skewed toward cloud and AI infrastructure. Hyperscalers like Amazon, Google, Microsoft, and Meta have each laid out multi-year AI infrastructure budgets that rise again in 2026 and beyond.

Here's why this matters for NVIDIA specifically: the vast majority of that AI infrastructure spend flows through GPU purchases. When Microsoft builds out Azure AI capacity, they're buying NVIDIA chips. When Meta scales its training clusters, they're buying NVIDIA chips. NVIDIA's revenue is downstream of these capital budgets.

When you own the pick-and-shovel company, your earnings call is really everyone else's spending call.

This is the core of what to listen for. NVIDIA's forward guidance needs to connect the macro CapEx story to NVIDIA-specific revenue. If it does, the "AI super-cycle" narrative holds. If it doesn't, the market will start asking whether the upside is already behind us.

There are three things to evaluate:

Does NVIDIA's next-quarter (and ideally full-year) guidance scale with the 50%+ CapEx increase that hyperscalers have already telegraphed?

Context: NVIDIA just printed quarterly revenue of roughly $68 billion — up about 73% year-over-year and 20% quarter-over-quarter. Full-year FY26 revenue came in around $216 billion, up 65% year-over-year, driven almost entirely by data center AI GPUs.

The bar is very high. If the guide implies growth is decelerating meaningfully from these levels — even if it still "beats" consensus — the narrative can shift from "AI super-cycle" to "GPU cycle is peaking." Watch the slope, not just the number.

This is what separates a secular shift from a cyclical wave. If management offers explicit commentary about their order book extending into 2027 or 2028, or references multi-year supply agreements, that supports the case that this is a durable platform shift — not a 2–3 year hardware upgrade that's already halfway done.

If they only talk about next quarter with confidence, investors will fill the silence with doubt.

Is hyperscaler AI CapEx being spent at a steady pace, or are companies pulling spend forward?

This is the subtlest and most important signal. If NVIDIA hints — even indirectly — that some hyperscaler spending is front-loaded and could moderate after 2026, the market will immediately start discounting the out-years that the stock is currently priced on.

Listen for the difference between "we see sustained, multi-year demand growth" and "our customers are building capacity ahead of anticipated need." Both sound positive. Only one supports a multi-year thesis.

Bottom line: If the guide doesn't bridge the $600B+ CapEx narrative into multi-year NVIDIA revenue, most of that upside story is already priced in — while downside to expectations is not.

Earnings calls aren't just about numbers. The specific language management uses signals where their confidence sits — and the market parses it in real time.

Here's a keyword decoder to what you're listening for:

One thing to watch specifically: the beat size relative to recent history. NVIDIA has been beating EPS estimates by mid-to-high single digits — around 8% last quarter. That pattern is now baked into expectations. A small beat or in-line result, after that history, can get treated as a psychological miss even if the absolute number is fine. In a stock priced for perfection, tone matters as much as numbers.

Here is my checklist for this call. More checks in the left column means the multi-year thesis is intact. Mixed signals — especially on duration, margins, or tone — means the "priced for perfection" math starts working against you.

For any company whose revenue depends on someone else's capital budget, the earnings call isn't about the backward-looking print. It's about whether the forward spending story still holds.

When you watch NVIDIA's call, you're not just watching one company report. You're watching whether the largest AI infrastructure buildout in history is still accelerating — or starting to find its ceiling. That's the signal. Everything else is noise.

One more thing before I go.

This email is an example of the kind of analysis I'll be moving to a paid tier. Earnings previews, deep dives, and research into high-growth themes — the work that takes real hours to put together — will live under a premium subscription going forward.

The weekly tips and frameworks emails you've been getting on Thursdays? Those stay free. That's not changing. The educational content is the foundation of what we're building here, and I want it accessible to everyone. Free subscribers will also get access to one premium post per month — so you'll still see the kind of analysis you're reading right now, just not every week.

Premium subscribers also get full access to my archive of previous issues — every deep dive and earnings analysis I've published to date.

There are two paid tiers:

Founder Subscription — $12/month or $120/year. Weekly deep dives, earnings analysis, research into high-growth themes, and full access to the archive. This is where the bulk of the work lives.

Founding VIP — $1,200/year. Everything in the Founder tier, plus access to my personal small and mid cap watchlist with periodic updates, early access to research before it goes out to Founder subscribers, and the ability to suggest a research topic or theme once per quarter within the technology sectors I cover.

The Founder launch rate goes up to $17/month or $170/year after the first 100 subscribers or on May 25 — whichever comes first.

Subscribe at the launch rate → HERE

You'll still get a few more deep dives in your inbox before the paywall goes up — I'm not pulling the rug. But the launch pricing won't last, so if you know you want in, don't wait on it.

Stay disciplined, Koh

Disclaimer: Nothing in this newsletter constitutes investment advice or a recommendation to buy or sell any security. Numbers and observations are as of publication. I may hold positions in companies discussed above. Always do your own research and consult a licensed financial advisor before making investment decisions.