By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel consensus is bearish, with key risks including capex sustainability, custom ASIC acceleration, and delayed ROI feedback loops. The main opportunity lies in Nvidia's software ecosystem lock-in.
Risk: capex sustainability and custom ASIC acceleration
Opportunity: Nvidia's software ecosystem lock-in
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
The past month or so has not been kind to many chip stocks. Investors have grown increasingly worried that some have become overvalued and may not be worth holding. That's sent the share prices of Nvidia (NASDAQ: NVDA), Broadcom (NASDAQ: AVGO), and Cerebras (NASDAQ: CBRS) lower over the past several weeks.
But does this volatility among semiconductor stocks really mean you should get out of these three companies?
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
Cerebras just went public a few weeks ago, with an IPO price of $185 and an opening trading price of $350. However, after an initial share price surge, Cerebras' stock has shed about 18% since May 15 (the day after its IPO).
The company designs and manufactures massive wafer-scale chips -- each one about the size of a dinner plate -- that put the equivalent processing power of a cluster of GPUs into one large integrated chip. This new approach to semiconductors has put the company in competition with Nvidia, which (like most other chip companies) turns each silicon wafer it uses into hundreds of smaller processors, rather than one large one.
Cerebras says that its large-wafer technology is more efficient for artificial intelligence (AI) inference, making it a better fit for the next phase of the technology.
Shareholders who are thinking about selling right now may want to reconsider. While the company has its risks -- it's not profitable on a generally accepted accounting principles (GAAP) basis, and its shares are very expensive -- its unique approach to AI processing has quickly gained popularity among AI companies.
Consider that OpenAI says it will spend $20 billion on Cerebras' processors over the next few years, and Amazon is already integrating them into its AWS AI cloud services. Cerebras shares come with a premium price tag right now -- it trades at a price-to-sales (P/S) ratio of 97 compared to the tech sector's average of about 8 -- but having a small position in this novel AI company could pay off if more companies shift to its large wafer tech.
Broadcom stock has been on a strong run -- it's still up 55% over the past 12 months. But some investors changed their tune about the company after it reported its fiscal second-quarter results at the beginning of June.
Non-GAAP earnings per share of $2.44 beat Wall Street estimates, and AI semiconductor revenue surged by 143% to $10.8 billion. Broadcom's management also reiterated its previous estimate of $100 billion in AI chip sales for the year.
But investors were hoping that management would raise its AI chip guidance. And they didn't like that the company's total sales of $22.2 billion came in slightly below the analysts' consensus estimate of $22.27 billion.
But the sell-off that followed was likely an overreaction. Broadcom's application-specific integrated circuits (ASICs) still play a unique role in AI processing, allowing its customers to design processors to handle precisely the workloads that come from their AI models. Alphabet, Anthropic, OpenAI, and Meta Platforms are some of Broadcom's key customers.
Its semiconductor gross margins are high -- around 70% in the second quarter, which helped drive Broadcom's net income up 88% to $9.3 billion.
What's more, now that the stock has tumbled over the past month, it's carrying less of a premium. Broadcom's stock now trades at a price-to-earnings ratio of about 65, down from around 88 several weeks ago.
With its strong margins, long list of AI clients, and unique AI processors, owning Broadcom still looks like a good long-term bet.
Of course, Nvidia remains a leading semiconductor company, but it, too, is facing some investor skepticism right now.
Large tech companies have committed to investing hundreds of billions of dollars into AI infrastructure annually, and the pace at which they're laying out those funds has been increasing, but some onlookers have grown increasingly concerned that this spending spree could end soon. They're also aware that Advanced Micro Devices, Broadcom, and Cerebras offer AI chip alternatives to Nvidia.
They're not wrong to assume that this torrid spending won't last forever, and some investors have begun selling their shares of Nvidia to lock in their gains.
There's nothing wrong with that strategy, but it could be a mistake to assume Nvidia won't be able to hold on to its dominant position in the AI accelerator space. The company currently has a market share of 88% in data center GPU sales.
Nvidia is also looking toward the future of AI, in which it believes everything from PCs to cars and robots will have some level of autonomy built into them. If that turns out to be the case, then AI processing certainly still has more room to expand, and Nvidia's current lead will only be to its benefit.
And, with a price-to-earnings ratio of just 30 right now, Nvidia's stock is relatively inexpensive compared to many of its peers.
Before you buy stock in Nvidia, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Nvidia wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $438,283! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,257,427!
Now, it’s worth noting Stock Advisor’s total average return is 938% — a market-crushing outperformance compared to 206% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of June 12, 2026. *
Chris Neiger has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Amazon, Broadcom, Meta Platforms, and Nvidia. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Nvidia faces faster-than-expected share erosion to Broadcom ASICs and Cerebras as 2025 capex growth slows."
The article frames the chip sell-off as overdone, citing Nvidia's 88% share, Broadcom's 70% ASIC margins and $100B AI guidance, and Cerebras' OpenAI deal. Yet it glosses over hyperscaler capex sustainability: total AI infrastructure spend could plateau if ROI metrics disappoint in 2025. Cerebras' 97x P/S embeds flawless execution on unproven wafer-scale inference at scale, while Broadcom's missed revenue consensus and flat guidance hint at demand normalization. Nvidia's 30x P/E looks cheap only if growth stays above 40%; any share loss to custom ASICs would compress multiples faster than peers.
If enterprise AI adoption accelerates beyond data-center training and Nvidia's CUDA ecosystem locks in 80%+ share through 2027, the recent 20-30% drawdowns simply reset entry points before re-rating.
"Nvidia's data-center GPU dominance and ecosystem create a durable moat that justifies a long-term investment thesis even if near-term multiples compress."
The piece overemphasizes near-term jitters in chip names while underplaying Nvidia's secular AI demand and platform moat. Nvidia commands about 88% of data-center GPU share, with a pricing power moat and a rapidly expanding software/cloud ecosystem that could make AI capex a repeatable revenue stream, not a one-off spike. Cerebras remains a moonshot with GAAP losses and an eye-watering P/S around 97; OpenAI/AWS hype helps, but profitability isn’t assured. Broadcom looks solid on margins, yet a 65x forward P/E after a run implies ambitious optimism for AI-driven growth. The real risk is capex normalization failing to arrive as quickly as assumed, not just market sentiment.
Cerebras might prove its wafer-scale approach scales in practice if major customers commit long-term; Nvidia's moat could face pressure if customers seek cheaper, custom silicon or if AI capex slows.
"The market is currently mispricing the difference between Nvidia's proven high-margin dominance and the extreme execution risk inherent in Cerebras' unproven, capital-intensive manufacturing model."
The article conflates three distinct risk profiles. Nvidia (NVDA) remains a valuation anomaly; at a 30x forward P/E, it is arguably the 'value' play in AI, provided its data center moat holds against custom silicon. Broadcom (AVGO) is a different beast—its transition to a custom ASIC powerhouse makes it a play on infrastructure efficiency, not just raw compute. The real danger is Cerebras (CBRS). Trading at 97x sales is pure speculation on wafer-scale architecture. The article glosses over the massive execution risk: scaling manufacturing for dinner-plate-sized chips is exponentially harder than standard GPU production. Investors are confusing 'innovative tech' with 'investable equity' at these entry multiples.
If Cerebras successfully scales its Wafer-Scale Engine, it could render traditional multi-GPU clusters obsolete, making a 97x P/S ratio look like a bargain in retrospect.
"All three stocks are experiencing earnings deceleration masked by year-over-year comparisons, and the article's 'less of a premium' language obscures that multiples have compressed because growth forecasts have, not because valuations reached equilibrium."
The article frames a valuation reset as opportunity, but conflates three fundamentally different stories. Cerebras is pre-revenue speculation dressed up as innovation (P/S of 97 is not a discount—it's a warning). Broadcom's 143% AI revenue growth is real, but the article glosses over why guidance wasn't raised: saturation signals in custom ASIC demand. Nvidia's 88% GPU market share is genuine moat, but at 30x P/E it's priced for perpetual 25%+ growth in a market where capex cycles are already showing fatigue. The article's 'less premium' framing is misleading—these stocks fell because growth expectations contracted, not because valuations compressed to fair value.
If large-cap tech capex truly accelerates into 2025-26 (Nvidia guidance suggests it), and if Cerebras' efficiency gains actually shift workload allocation away from GPU clusters, then today's prices could look cheap in 18 months, making this a genuine dip-buying moment.
"Cerebras' OpenAI deal may generate revenue but doesn't mitigate the broader capex sustainability risks affecting Nvidia and Broadcom."
Claude correctly flags Cerebras' valuation risk but understates how an OpenAI deal could seed revenue, unlike pure pre-revenue plays. This connects to Grok's capex sustainability concern: if custom ASICs from Broadcom accelerate, Nvidia's moat erodes faster than 30x P/E implies. The unmentioned risk is delayed ROI feedback loops in 2025 causing broader hyperscaler pullback across all three names, amplifying drawdowns.
"Cerebras' 97x P/S is not a discount—it's a risk indicator until practical monetization, given manufacturing/integration hurdles and reliance on marquee customers."
Claude overstates Cerebras' potential without stressing the business model tightrope: wafer-scale is a massive manufacturing and integration challenge, not merely a hardware leap. Even with an OpenAI deal, ROI timing is uncertain, sales cycles are long, and reliance on a small number of marquee customers risks revenue volatility. If hyperscalers pull back capex or if software ecosystems don't commercialize workloads, 97x P/S becomes a mirage rather than a badge of disruptive scale.
"Nvidia's moat is defined by software ecosystem lock-in (CUDA), which makes the hardware-centric competition from Cerebras or custom ASICs less threatening to their market share than the panel suggests."
Gemini and Claude correctly identify the manufacturing absurdity of Cerebras, but both ignore the 'platform' trap. Nvidia isn't just selling chips; they are selling the entire software stack that defines the AI development cycle. Even if Cerebras creates a faster chip, the switching cost of moving away from CUDA is effectively infinite for most enterprises. The real risk isn't the hardware, it's the software ecosystem lock-in that makes Nvidia's 30x P/E look more resilient than Broadcom’s ASIC-dependent model.
"CUDA moat is real but not a valuation floor if custom ASICs achieve material efficiency gains at scale."
Gemini's CUDA lock-in argument is overcooked. Yes, switching costs exist, but they're not 'effectively infinite'—hyperscalers have proven willingness to build custom silicon (TPUs, Trainium) when ROI justifies it. The real constraint isn't software moat; it's that Cerebras and Broadcom ASICs must prove 2-3x efficiency gains *and* maintain software parity. Nvidia's 30x P/E survives if neither happens. If both do, the multiple compresses hard regardless of CUDA.
The panel consensus is bearish, with key risks including capex sustainability, custom ASIC acceleration, and delayed ROI feedback loops. The main opportunity lies in Nvidia's software ecosystem lock-in.
Nvidia's software ecosystem lock-in
capex sustainability and custom ASIC acceleration