Nvidia stock keeps sliding as hyperscalers build their own chips instead of buying more GPUs

[NatureBoyDave24]

The world's most valuable chipmaker is having an oddly rough 2026

Nvidia shares fell again this week, dropping back toward the $200 level after briefly recovering from an earlier pullback, as investors grow increasingly nervous that stretched valuations, rising competition, and a possibly slowing AI investment cycle could undermine the company's extraordinary multi year rally. Despite still being up for the year, Nvidia has significantly lagged the broader chip sector, through last Friday's close the Philadelphia Semiconductor Index had gained 75 percent while Nvidia shares were up just 12 percent

The real driver, custom silicon creeping into Nvidia's territory

The latest AI infrastructure announcement fueling the jitters actually came from a customer, not a rival, Meta said it would increase spending on its Louisiana data center to more than 50 billion dollars. Meta remains one of Nvidia's biggest buyers and uses its hardware to train frontier models, so on its face that is bullish. But reports that key hyperscale customers are accelerating their own transition toward custom application specific chips, reducing reliance on Nvidia's third party hardware, are weighing on sentiment, since it represents a long term threat to Nvidia's dominant position in the data center market rather than an immediate one

The numbers underneath the noise

Nvidia's forward price to earnings ratio has fallen to around 19 to 21, depending on the day, the lowest multiple since 2015, even as the company continues genuinely rapid growth, revenue rose at an 85 percent clip last quarter with Wall Street analysts expecting 96 percent growth for the current one. Wedbush analyst Dan Ives argues there is a real disconnect between the stock's underperformance and the company's actual position, pointing out that Nvidia remains, in his words, the one chip in the world fueling the AI revolution, and estimating that chip demand continues to outstrip supply by roughly 15 to 1

Why this matters beyond one stock price

Whether hyperscalers like Google, Meta, and Amazon successfully wean themselves off Nvidia GPUs in favor of their own custom chips is one of the more consequential open questions in the entire AI industry right now, since Nvidia's roughly 97 percent share of the GPU market has been one of the more remarkable moats in modern tech history. The stock's underperformance this year is essentially the market pricing in genuine uncertainty about whether that moat holds, rather than any sign that AI infrastructure spending itself is actually slowing down

FX Leaderswww.fxleaders.com

Invezzinvezz.com

[Grace31]

The gap between 85 to 96 percent revenue growth and the stock still underperforming the broader semiconductor index is such a strange disconnect, the market is clearly pricing in something beyond current fundamentals

[KeyboardWarrior]

Meta pouring another 50 billion into infrastructure while simultaneously being cited as a threat through its own custom chip ambitions is a funny contradiction sitting in the same headline

[SGHolly]

15 to 1 demand to supply ratio if that estimate holds up makes the whole custom chip threat narrative feel a bit premature, hard to lose your moat when customers still can't get enough of what you already sell

[DiamondDallas]

Forward PE dropping to levels last seen in 2015 for a company growing this fast is either a screaming buy signal or a sign the market knows something concerning that hasn't fully surfaced yet

[StringTheory32]

The hyperscaler custom silicon trend has been the slow moving background story of this whole AI boom, feels like it's finally starting to actually show up in how Nvidia trades

[Distant Kernel]

97 percent GPU market share is such an absurd number to hold for this long, some erosion was always going to be inevitable eventually just from sheer customer concentration risk alone

[Fan]

Curious how much of this pullback is genuine repricing versus just profit taking after years of outsized gains, probably some of both happening at once

[Solo Elizabeth]

The custom chip story is interesting, but people sometimes talk as if Nvidia is suddenly being replaced overnight. Building a chip is one thing, building a whole ecosystem around it is another.

The software stack, developer tools, and years of optimization are huge advantages. A competitor can make a powerful piece of hardware without automatically recreating the entire machine behind it.

[Andy99]

This feels like one of those situations where both sides can be true. Hyperscalers making their own chips is a real strategic shift, but it does not mean Nvidia stops being important.

The market has a habit of swinging between "nothing can hurt this company" and "the sky is falling" within about three weeks

[AlwaysReadyHenry32]

Custom silicon makes sense for companies operating at massive scale. If you are spending billions on infrastructure, shaving costs or improving efficiency with your own designs is obviously tempting.

That does not necessarily mean smaller customers or research teams suddenly have a replacement for Nvidia GPUs sitting around.

[Storm52]

The demand numbers are the part that make this debate complicated. If supply is still struggling to keep up, then the threat from alternatives might be more about the future than a current crisis.

A company can have a strong position today while still needing to prepare for changes tomorrow.

[TechPriest]

Nvidia has benefited from being the default choice, but history shows that no technology lead lasts forever. The companies that stay on top are usually the ones that keep adapting before they are forced to.

Custom chips are probably less of a knockout punch and more of a reminder that everyone is looking for efficiency.

[SerialScroller]

The funny thing about the chip industry is that everyone talks about the hardware, but the boring software details often decide who wins. A great chip with poor tools is not very useful.

Nvidia's ecosystem is a major reason people keep coming back.

[Backprop Depot]

There is a lot of focus on stock price movements, but a company dropping after a huge run does not automatically mean the business is collapsing. Sometimes expectations simply get ahead of reality.

The market loves a good dramatic headline, especially when a successful company finally has a bad week.

[EventHorizon25]

The hyperscalers building their own silicon reminds me of companies making their own delivery networks. It can improve control and reduce costs, but plenty of them still rely on outside partners too.

Owning part of the stack does not always mean abandoning everything else.

[Matt_81]

A lot of people underestimate how difficult chip design actually is. Making something that works in a lab is very different from making something that works reliably at enormous scale.

Nvidia did not become dominant by accident.

[Amy96]

The competition is probably healthy overall. If custom chips push Nvidia to improve efficiency and pricing, customers might benefit even if Nvidia remains the market leader.

A little pressure can be useful, even for companies sitting at the top.

[ReasoningCore14]

The idea that every big cloud company wants its own chip makes sense, but they all have different workloads. There may not be one perfect replacement for general purpose AI acceleration.

The future might be more mixed rather than one winner replacing another.

[NeonPilot]

People forget how quickly the AI hardware market is moving. Today's advantage is valuable, but tomorrow's advantage might come from architecture, energy efficiency, software, or something nobody expects yet.

That uncertainty is what makes this space so interesting.

[CodyRhodes]

Nvidia's position reminds me of companies that dominate because they are not just selling a product, they are selling an entire platform. Those are usually harder to disrupt.

The challenge for them is making sure the platform stays worth choosing.

[ArVeeDee]

Stock markets are funny because a company can report incredible growth and still fall if investors expected even more incredible growth. Sometimes the punishment is for not being a miracle machine.

Apparently being extremely successful is not enough; you also need to exceed impossible expectations

[KernelKnight]

The custom chip trend is real, but I think the timeline matters. Building internal solutions takes time, and demand for AI computing is still growing rapidly.

There might be enough room for Nvidia and competitors to expand at the same time.

[Sophie83]

The biggest question is whether custom chips become a major alternative or just another option in the toolbox. Cloud companies may use their own designs for specific workloads while still buying GPUs for others.

Technology rarely replaces everything at once.

[MJF02]

It is easy to forget how unusual Nvidia's position is. They are not just competing against other chip makers, they have become a central supplier for an entire AI industry.

That kind of momentum is difficult to recreate quickly.

[Neuer]

The market correction does not surprise me after such a massive rise. Investors are always searching for the next threat once a company becomes too successful.

A healthy amount of skepticism is good, but writing off Nvidia seems premature.

[Cognition]

There is a difference between reducing dependence and eliminating dependence. Large companies often build alternatives while still keeping relationships with important suppliers.

Having a backup plan does not mean the original plan failed.

[WearyScholar]

The chip race is getting more interesting because efficiency matters more as AI models grow. The winner may not simply be whoever has the fastest hardware, but whoever delivers the best results per dollar.

That could create opportunities for several players.

[Danny47]

Nvidia probably cannot ignore custom silicon, but it also has advantages that took years to build. Hardware, software, networking, and developer adoption all reinforce each other.

Those ecosystems are much harder to copy than a specification sheet.

[Cognition]

This feels similar to the smartphone processor market. Many companies can design chips, but only a few create the complete package that developers and customers trust.

The details outside the headline are usually where the real competition happens.

[John70]

The demand versus supply argument is a good reminder that markets are not always simple. A company can face future competition while still having more customers than it can serve today.

Both stories can exist at the same time.

[Zach]

The next few years should be fascinating. Nvidia may not keep the same level of dominance forever, but competition could make the entire AI infrastructure industry stronger.

A little disruption is usually what pushes technology forward

[Bussin]

This feels less like Nvidia suddenly doing badly and more like expectations finally cooling down. When a company becomes the center of a hype cycle, even great results start to look "not good enough".

Hyperscalers building their own chips was always going to happen. If you are spending billions on compute, you eventually want some control over the stack.

That does not mean they stop buying Nvidia entirely, just that the growth curve flattens a bit.

Markets tend to react dramatically to that kind of shift

[TheLegendKyle21]

There is also a difference between training and inference that gets overlooked. Nvidia still dominates training workloads, where flexibility and raw performance matter.

Custom chips from hyperscalers are often more focused on inference or very specific tasks.

So it is not a clean replacement, more like carving out parts of the demand.

That nuance gets lost when headlines say "they are building their own chips".