Nvidia is not Cisco, but the AI supply chain can still crack first
The clean rebuttal to the 'Nvidia is the next Cisco' trade is not that valuation suddenly stopped mattering. It is that AI demand is still being funded by real hyperscaler capex, while the more fragile expectations now sit deeper in the supply chain where memory, networking, and adjacent semis are priced for a smoother runway than this cycle is likely to deliver.
The market is arguing about the wrong comparison. NVDA is not Cisco in the simple dot-com sense because today’s buyers are not speculative telecom balance sheets; they are giant cloud platforms still writing very real capex checks. But that does not mean the whole AI complex is safe. The better way to frame this selloff is that the top of the stack can remain durable even as weaker links crack first across second-order suppliers, where expectations have run ahead of how clean this buildout will actually be.
Start with the anchor: hyperscaler spending is still the fact that matters most. Public commentary this year points to the largest cloud platforms pushing toward roughly $750 billion of capex in 2026, and that is why the broad 'AI is fake' argument still misses the setup. Microsoft, Meta, and Alphabet are not behaving like late-stage tourists; they are funding AI from operating strength. That is exactly why the Cisco analogy is too blunt. The issue is not whether the demand base exists. The issue is where the market has become too confident about who captures that spend cleanly.
That distinction matters because NVDA still looks more like the platform toll collector than the weakest cyclical node. On the numbers available here, Nvidia trades at 34.73x earnings with 65.5% revenue growth and a 63.0% net margin. That is expensive in absolute terms, but it is not the profile of a hollow bubble stock detached from fundamentals. The market is paying for dominant economics and still-rare pricing power. If the next leg of AI spending gets more selective, Nvidia may feel it, but it is still closer to the center of the budget than many names investors have treated as automatic sympathy winners.
The more vulnerable part of the chain is where valuation and operating leverage are leaning on a best-case continuation. AVGO trades at 45.61x earnings and 23.38x sales, richer than Nvidia on both counts, even though its AI upside depends more on custom silicon and networking demand staying hot across a narrower set of customers. MU is cheaper on earnings at 18.82x, but the stock is up 169.1% year to date, which tells you how much future memory tightness and AI-driven pricing power are already embedded in expectations. Memory is exactly where cost inflation can become both a bottleneck and a margin squeeze elsewhere in the system. When the market starts questioning AI returns, that is the kind of second-order exposure that usually gets repriced before the core compute thesis does.
A quick comparison shows where the market has stretched the chain:
NVDA: 34.73x P/E, 65.5% revenue growth, 63.0% net margin
That list is the real tell. TSM at 33.20x earnings is not cheap for a manufacturer, even with 33.0% revenue growth and a 47.0% net margin. The market has moved beyond rewarding only the obvious winner and is now capitalizing much more of the ecosystem as if the entire chain deserves premium multiples at once. That is usually where cracks form first. Foundry, memory, packaging, networking, and custom silicon all benefit from AI capex, but they do not all have the same ability to defend margins if customers push back, product cycles slip, or component costs rise.
Yes, bulls have a fair point: this is not 2000 because the buyers are real and cash-rich, and upcoming results from MSFT, META, and Alphabet are more likely to confirm continued capex than a sudden collapse. We agree with that much. But the stronger that point becomes, the more investors should focus on the next question: if hyperscalers keep spending, who is most exposed to lower returns on that spend? Recent concerns around slower AI agent progress and rising memory costs matter because they challenge the timing and efficiency of monetization, not necessarily the existence of the budget. That is a much bigger problem for the suppliers priced for frictionless volume than for the platforms that can absorb a longer payoff period.
This is also why the recent weakness in AI-linked semis should not be read as a clean verdict on NVDA alone. MSFT trades at 24.32x earnings with just 14.9% revenue growth and is down 16.7% year to date, a sign that the market is already discounting capex pressure more heavily at the buyer level than the headline spend numbers imply. META, at 19.16x earnings and 7.63x sales, is not being priced like a euphoric AI bubble stock either. In other words, some of the skepticism is already showing up in the hyperscalers and platform names, while parts of the hardware chain still carry richer assumptions. That mismatch is where we would expect more pain if earnings over the next few weeks confirm that spending remains high but the economics get messier.
The cleanest way to separate signal from noise here is simple: stop asking whether NVDA is Cisco and start asking where the AI buildout is most vulnerable to disappointment. We think the answer is not the top-line capex commitment from hyperscalers, at least not yet. It is the broader supplier complex that has been priced as if every layer of the stack will enjoy Nvidia-like durability and pricing power.
What we are watching now is not whether Microsoft, Meta, and Alphabet keep spending, but whether their commentary implies rising friction in the form of memory costs, slower productization, or tougher return hurdles. If those pressures ease and second-order suppliers keep converting AI demand into durable margins, this view weakens. If capex holds up while adjacent semis and memory names stumble first, the market will have its answer: Nvidia was not Cisco, but plenty of the AI supply chain was still priced like the cycle could not crack.
Our take, not advice. This is opinion commentary — informational only, not personalized investment recommendations. Markets carry risk. Do your own research and consider your own situation before any trade.
▌The Daily Briefing · Free
A new stock idea, every evening.
One stock worth watching each weekday, plus the analysis behind it. Free, in your inbox.
