Growth is accelerating at all three. The businesses themselves don't seem to be the problem.
In a single week, roughly $1.3 trillion in market value vanished from the stocks of Nvidia, AMD, and Broadcom — not because their businesses faltered, but because markets, prone to sudden crises of confidence, began questioning whether extraordinary growth had been priced too generously. The companies themselves reported revenue expansions of 38% to 143%, yet fear of what valuations demand of the future proved more powerful than the evidence of the present. This is an old tension in markets: the difference between the worth of a thing and the price the world assigns it. For patient observers, such moments have historically been less endings than invitations.
- A single Friday erased $1.3 trillion in AI chip market value, with Broadcom losing a fifth of its worth in seven days after its own earnings report triggered a sector-wide panic.
- The paradox at the heart of the sell-off is stark: Nvidia's data center revenue grew 92%, AMD's 57%, and Broadcom's AI chip revenue surged 143% — yet all three stocks fell sharply.
- Investor anxiety is not about broken businesses but about the mathematics of valuation — whether multiples above 30x or even 100x earnings leave any room for the inevitable stumble.
- Broadcom's margin compression, as high-growth AI chips displace its more profitable legacy products, has become a proxy for a broader fear: that AI's ascent may quietly erode the profitability it appears to generate.
- The sector now sits at an unresolved crossroads — fundamentals pointing upward, sentiment pointing down — with no clear signal that the selling has exhausted itself.
The semiconductor sector endured one of its most turbulent weeks in recent memory, with Nvidia, AMD, and Broadcom collectively shedding $1.3 trillion in market value. Nvidia fell 6% in a single session and is now down 18% from its peak. AMD dropped nearly 11% and has retreated from $542 to around $452. Broadcom, whose earnings report ignited the panic, lost roughly a fifth of its value across seven days — a striking outcome for a company that had just beaten its own guidance.
What makes the sell-off philosophically interesting is what it was not about. All three companies reported growth figures that would have seemed implausible just a few years ago. Nvidia projected 95% data center revenue growth. AMD's data center segment expanded 57% and is accelerating. Broadcom's AI chip revenue jumped 143% year over year, and management reiterated a target of more than $100 billion in AI chip revenue by fiscal 2027. The businesses are not weakening — they are compounding.
The market's concern is instead a question of price versus value. Nvidia trades at roughly 31 times earnings, a multiple that demands its extraordinary trajectory continue uninterrupted. AMD trades at more than 100 times trailing earnings, leaving almost no tolerance for execution missteps. Broadcom faces a subtler anxiety: as AI semiconductors grow to dominate its revenue mix, they are mathematically diluting the company's overall gross margins, since AI chips carry lower margins than its legacy software and non-AI products. CEO Hock Tan emphasized that semiconductor margins remain stable, but the optics of compression unsettled investors already primed for disappointment.
For those with long time horizons, the central question is whether this represents panic or prophecy. The risks are real — AI infrastructure spending could slow, competition could intensify, and valuations leave little cushion. But these are different risks from business deterioration, and distinguishing between them matters enormously. Investors who believe AI infrastructure buildout will continue accelerating — and that these three companies will remain its primary architects — now find themselves facing something that was unavailable just weeks ago: the sector's leaders at meaningfully lower prices.
The semiconductor market just experienced one of its most violent weeks in recent memory. On a single Friday, roughly $1.3 trillion in value evaporated from AI chip stocks as investors suddenly lost their nerve. Nvidia fell 6% that day. AMD dropped nearly 11%. Broadcom, which had triggered the panic with its earnings report earlier in the week, shed about a fifth of its market value in just seven days. The selling has continued into the following week, piling fresh losses onto an already wounded sector.
Yet beneath the wreckage lies something worth examining. All three companies—Nvidia, Advanced Micro Devices, and Broadcom—reported earnings that by any historical standard would be considered exceptional. The businesses themselves are not deteriorating. They are accelerating. Nvidia's data center revenue grew 92% year over year in its most recent quarter, with management projecting 95% growth ahead. AMD's data center business, which now drives the company's expansion, grew 57% in the first quarter and is expected to grow 46% in the second. Broadcom's AI chip revenue jumped 143% year over year, crushing the company's own forecast. These are not the numbers of struggling enterprises.
Nvidia remains the sector's heavyweight, valued at roughly $4.9 trillion even after its 18% decline from its 52-week peak. The stock now trades at about 31 times earnings—a multiple that assumes the company's extraordinary growth trajectory will persist. CEO Jensen Huang framed the opportunity plainly during the earnings call, describing the buildout of AI infrastructure as "the largest infrastructure expansion in human history" and noting it is "accelerating at extraordinary speed." If that assessment holds, Nvidia's valuation may prove conservative rather than excessive.
AMD presents a different profile. The stock had reached $542.52 just before Broadcom's report spooked the market, but has since fallen to around $452—still more than double its value at the start of 2026. The company's data center segment, which CEO Lisa Su identified as the primary engine of growth, is expanding at a 57% clip. AMD plans to ship Helios, its first complete rack-scale AI system, later this year, with OpenAI and Meta already committed as customers. The risk here is valuation: the stock trades at more than 100 times trailing earnings, leaving virtually no margin for error in execution.
Broadcom's situation is perhaps the most puzzling. The company delivered a quarter that exceeded its own guidance—$22.2 billion in total revenue, up 48% year over year, with AI chip revenue of $10.8 billion, up 143%. Management guided to $16 billion in AI chip revenue for the next quarter and reiterated a target of more than $100 billion in AI chip revenue for fiscal 2027. Yet the stock fell hardest of the three. Part of the market's concern appears rooted in margin pressure. As AI semiconductors grow from a smaller portion of the business to a dominant one, the overall gross margin is being diluted—AI chips carry lower margins than some of Broadcom's legacy software and non-AI semiconductor products. CEO Hock Tan addressed this directly, emphasizing that semiconductor margins themselves remain "very stable and very solid," but the product mix shift is mathematically compressing the company's overall profitability metrics. Some investors also seemed disappointed that management merely reiterated its $100 billion target rather than raising it.
For long-term investors, the question becomes whether this sell-off represents genuine deterioration or temporary panic. The underlying businesses show no signs of weakness. Growth is accelerating across all three companies. The concern instead centers on whether current valuations can be justified if AI infrastructure spending slows, and whether competition might erode these companies' market positions. These are legitimate risks. But they are different from the risk that the businesses themselves are breaking down.
A measured approach may make sense. Rather than deploying capital all at once into stocks that have fallen sharply but may fall further, investors with a long time horizon could consider building positions gradually. The market has shifted sentiment with remarkable speed this month, and there is no guarantee that the selling has finished. But for those convinced that AI infrastructure spending will continue accelerating—and that these three companies will remain central to that buildout—the prices now available represent something that was unavailable just weeks ago: genuine entry points into the sector's leaders.
Citações Notáveis
The buildout of AI factories—the largest infrastructure expansion in human history—is accelerating at extraordinary speed.— Jensen Huang, Nvidia CEO
Semiconductor margins remain very stable and very solid. It's the mix, particularly the mix between software and non-AI to the very, very rapidly growing AI semiconductor that is just diluting gross margin.— Hock Tan, Broadcom CEO
A Conversa do Hearth Outra perspectiva sobre a história
Why did the market suddenly turn on these companies when their earnings were so strong?
The earnings themselves were strong, but they raised a different question: can these valuations hold if growth doesn't stay at these extraordinary rates? Investors started doing the math and realized there's almost no room for disappointment.
So it's not that the businesses are failing—it's that the prices got ahead of reality?
Exactly. Nvidia's trading at 31 times earnings. AMD at over 100 times. Those multiples only work if the growth story stays intact. When Broadcom's report came out, some investors started wondering if maybe the growth was peaking.
But Broadcom's AI revenue jumped 143%. That doesn't sound like peaking.
It doesn't. But the company only reiterated its targets rather than raising them. And there's a margin issue—as AI chips become a bigger part of the business, they're dragging down overall profitability because they carry lower margins than the legacy products.
So the fear is that success in AI is actually making the business less profitable?
Not less profitable in absolute terms. The AI business is growing fast and making money. But it's replacing higher-margin products, so the company's overall margin percentage is shrinking. That's a real dynamic, but it's different from the business breaking down.
If you believed in these companies before the sell-off, why would you believe in them now?
Because nothing fundamental changed about the businesses. The data center buildout is still accelerating. These companies are still winning. The only thing that changed is the price. For a long-term investor, that's actually when things get interesting.
What's the risk that this keeps falling?
Plenty. AI spending could slow. Competition could intensify. And valuations this high leave no room for mistakes. But the risk of further decline is different from the risk that you're buying a broken business.