Have you ever watched a stock you love get hammered on a headline and felt that little twist in your stomach? Yeah, me too. This week it was Nvidia’s turn again. Shares slid another 4% on Tuesday after a report suggested one of the biggest buyers of its chips might start leaning on something else entirely. The whispers are getting louder: is the reign of the GPU king finally facing a serious challenge?
Let’s be real for a second—Nvidia has been the undisputed heavyweight champion of the AI boom. But markets hate complacency, and the second someone mentions “alternatives,” the algo traders hit sell faster than you can say “margin call.” So I dug into what actually happened, what the smartest people on the Street are saying, and whether this is a blip or something investors genuinely need to worry about.
What Sparked the Latest Sell-Off?
The trigger was pretty straightforward. Word leaked that a major hyperscaler—think one of the handful of companies dropping tens of billions on AI infrastructure—was seriously exploring the idea of using custom-designed tensor processing units instead of piling even more Nvidia GPUs into their data centers. For the non-technical crowd, that’s basically choosing a purpose-built race car over a very fast, very flexible Swiss Army knife on wheels.
These custom chips, often called ASICs, promise better power efficiency for specific workloads. And when you’re burning enough electricity to power small countries, every watt matters. The fear trade kicked in immediately: if the biggest spenders start building their own solutions or licensing someone else’s, does that cap Nvidia’s growth story?
Short answer from almost everyone who actually covers the space professionally? Not yet.
The “Compute Scarcity” Reality Check
Here’s the part a lot of headline readers miss. Right now the AI world isn’t swimming in excess chips—it’s starving for compute. Demand is so insane that companies are literally fighting over whatever silicon they can get their hands on, regardless of whether it says Nvidia, AMD, or was cooked up in a custom lab.
“GPUs are clearly not going anywhere. Right now the overarching theme is of compute scarcity, and if anything this feels like an effort to secure more.”
— Senior semiconductor analyst, Bernstein
Think of it like the California Gold Rush. Yes, someone just invented a slightly better pan, but there’s still way more gold in the river than anyone can dig up. Having more tools doesn’t mean the original shovel suddenly becomes worthless.
ASICs Aren’t New—And They’ve Never Killed GPUs Before
People forget that custom silicon has been around for years. Bitcoin miners went all-in on ASICs a decade ago and completely abandoned GPUs for mining. Did that destroy Nvidia? Hardly—they pivoted to gaming, professional visualization, and now AI. The company has a habit of finding the next multi-billion-dollar workload when one matures.
In machine learning, Google has been using its own TPUs internally since 2016. The cloud version has been available to outsiders since 2018. Yet somehow Nvidia’s data center revenue went from $4 billion in 2019 to over $100 billion expected this year. Clearly the existence of alternatives hasn’t exactly crippled the GPU juggernaut so far.
The Real Winner Might Be… Broadcom?
Here’s where things get interesting. While Nvidia catches the headlines (and the selling pressure), another company quietly surged this week—Broadcom. Up over 13% in a handful of days while Nvidia bled.
Why? Because Broadcom is one of the few companies on Earth that actually helps design and manufacture these fancy custom AI chips for the hyperscalers. More ASICs in the world can very easily mean more business for them, not less.
- They already build Google’s TPUs
- They’re deeply involved with multiple other custom programs
- They have the manufacturing relationships and IP that make these projects possible
Several analysts pointed out the irony: the same trend scaring Nvidia holders could be rocket fuel for Broadcom shareholders. In fact, some now argue AVGO deserves an even higher valuation premium than Nvidia going forward.
Valuation Reality: Nvidia Trading Like a “Normal” Company
Perhaps the most surprising takeaway from the research notes flying around this week? Nvidia, the company that basically prints money in AI, is suddenly trading at roughly 25 times forward earnings. That’s not nosebleed territory—that’s roughly in line with the broader market.
“NVDA is trading at ~25x market multiple, essentially valuing the company as another run-of-the-mill franchise, which we disagree with.”
— Bank of America securities analyst
When the perceived “AI trade” darling starts trading like Coca-Cola or Procter & Gamble, you have to wonder if the fear has gone a bit too far.
The Multi-Trillion Dollar Pie Everyone Keeps Forgetting
Step back for a minute and look at the actual numbers being thrown around. Some of the more aggressive forecasts suggest the total addressable market for AI accelerators and networking could grow five-fold or more by the end of the decade—from roughly $240 billion today to something north of $1.2 trillion.
Even if Nvidia’s share of that pie shrinks from 80-90% to “only” 60-70%, we’re still talking about a company that could reasonably triple or quadruple its revenue from here. The idea that a few custom chip projects suddenly derail that math feels… premature, to put it kindly.
So Should You Panic-Sell Nvidia?
Look, I’m not here to give financial advice—everyone’s risk tolerance and time horizon is different. But from everything I’ve read this week, the consensus among analysts who’ve spent decades covering semiconductors is remarkably calm.
- The overall AI buildout is still in early innings
- Compute demand continues to outstrip supply by a wide margin
- Nvidia’s next-generation products are reportedly sold out for the next 12 months
- Diversification into alternatives is natural and healthy for the ecosystem
In my experience, when a growth story this powerful starts trading at normal-market multiples amid a minor competitive scare, it’s usually worth paying attention—often on the long side.
The bottom line? Competition is coming. It was always coming. But the AI infrastructure buildout is so enormous that there appears to be room for multiple winners—for now. Whether Nvidia keeps 90% market share or “only” 70%, the absolute dollars could still be staggering.
For long-term investors, this week’s dip might end up looking like noise. For traders? Sure, ride the volatility. But the idea that the Nvidia story is suddenly broken because hyperscalers want more tools in their toolbox doesn’t seem to hold up under scrutiny.
As one analyst put it best: right now it’s not about GPU versus ASIC. It’s about whether the AI opportunity is still massive or whether we’re already late in the game.
All the evidence I’ve seen suggests we’re still very, very early.
