By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel discusses the recent sell-off in tech stocks, particularly Broadcom, Micron, and Marvell. While some attribute it to weaker earnings and fading AI-chip demand, others see it as a 'valuation reset' or temporary derisking after an outsized rally. The key debate revolves around the potential shift in hyperscalers' behavior, such as Google accelerating in-house silicon or diversifying away from Broadcom's custom ASICs, which could impact Broadcom's margins and market share.
Risk: Potential shift in hyperscalers' behavior towards in-house silicon, which could erode Broadcom's margins and market share.
Opportunity: Temporary pullback presenting a buying opportunity, as underlying demand for high-bandwidth memory and custom ASIC infrastructure remains intact.
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 →
Technology stocks fell in premarket trading Thursday, with chipmakers Broadcom, Micron Technology and Marvell Technology leading the losses as investors fled the semiconductor space.
Broadcom shares were down 15.1% ahead of the market open after the company, which designs and makes customized AI chips for other technology names, reported weaker-than-expected earnings on Wednesday.
The fall was part of a broader retreat from chipmakers and other tech stocks ahead of Thursday's market open.
Micron Technology was down 7.1% in premarket trading, while Marvell Technologies had fallen 7.5%.
Elsewhere, Qualcomm shed 3.9%, with Intel falling 4.1%, and AMD down 4.3% ahead of the opening bell.
Futures tied to the tech-heavy Nasdaq were last seen 1.4% lower.
AI memory chips have been a key theme in the equity market rally in recent weeks. Thursday's retreat led equities markets lower, with the S&P 500 futures last seen 0.7% down.
In a note Thursday, HSBC analysts led by Max Kettner, chief multi-asset strategist, flagged a slide in chip prices, coupled with a slowdown in AI spending and rollout, as among their "biggest worries."
John Vinh, equity research analyst at Keybanc Capital Markets, said the growing pressure on Broadcom and other semiconductor stocks is warranted.
"These stocks have all had very strong runs," Vinh told CNBC's "Squawk Box" on Thursday, pointing to repeated upward revisions, especially on the AI front. Vinh suggested that the Broadcom reversal indicates that market expectations have caught up with the chip sector run.
He noted that Broadcom has suffered a degree of share loss within its largest customer, Google, which has started diversifying toward other chip suppliers. "The near-term pull-back makes sense," Vinh explained, but added he remains optimistic on Broadcom.
Keith Lerner, CIO and chief market strategist at Truist Wealth, said that a sell-off was normal after a strong run, adding that "we're due for a rest."
"We've come a long way. Fundamentals are solid," he said on CNBC's "Closing Bell" on Wednesday.
"Bull market still deserves a benefit of the doubt, but often markets are two steps forward, one step back. We've had three steps forward, so maybe at least a mini step back, or at least some sideways chop."
Four leading AI models discuss this article
"Near-term downside momentum in semiconductors is a risk-off re-rating rather than a fundamental AI demand collapse; if AI spending holds and data-center capex proves persistent, the decline could reverse as investors reprice the sector."
Premarket losses hit Broadcom, Micron, and Marvell as tech stocks retreat; the article pins it on weaker Broadcom earnings and a fading AI-chip rally. But the read is not ironclad: the declines may reflect a near-term derisking after an outsized rally, not a collapse in AI demand. The crosscurrents matter: Google's diversification for Broadcom hints at competitive risk; Broadcom's mix and margin trajectory could be idiosyncratic for one quarter. Missing context includes concrete guidance on AI-related orders, memory-price trends, and broader data-center capex signals. If AI capex remains resilient and memory pricing stabilizes, the downside could prove temporary rather than durable.
The sell-off could be signaling a durable shift lower if AI demand cools and margin pressures intensify; however, the strongest counter is that this looks like a standard top-to-down re-rating after a rally, and if AI budgets stay intact, these names often bounce back quickly.
"The current semiconductor pullback is a healthy valuation correction driven by over-exuberant expectations rather than a fundamental collapse in AI-driven demand."
The sell-off in Broadcom (AVGO), Micron (MU), and Marvell (MRVL) is a classic 'valuation reset' masquerading as a fundamental breakdown. While John Vinh highlights Google's diversification as a headwind, the market is overreacting to the transition from 'AI hype' to 'AI execution.' Broadcom’s 15% drop ignores the long-term stickiness of their custom silicon business and networking dominance. We are seeing a rotation out of overextended semiconductor multiples into more defensive tech, but the underlying demand for high-bandwidth memory and custom ASIC infrastructure remains intact. I expect the 'sideways chop' Keith Lerner mentions to be a buying opportunity for investors who differentiate between cyclical laggards like Intel and secular winners.
If the slowdown in AI capital expenditure mentioned by HSBC is structural rather than temporary, these chipmakers face a multi-quarter margin compression that current consensus earnings estimates have not yet priced in.
"Broadcom's miss is a valuation and customer concentration problem, not proof that AI spending is rolling over; contagion selling in peers creates tactical opportunity if fundamentals hold."
Broadcom's 15% premarket drop is real and warrants scrutiny, but the article conflates a single earnings miss with sector-wide deterioration. AVGO's problem appears company-specific: Google diversification and customer concentration risk, not macro AI spending slowdown. Micron (MU) and Marvell (MRVL) down 7%+ on contagion, not their own news. The HSBC note on 'chip price slides' is vague—which chips, which segments? Memory pricing has cyclical pressure, but AI accelerator ASPs remain elevated. Keybanc's Vinh is right: valuations got ahead of themselves after 'three steps forward,' but that's valuation reset, not demand destruction. S&P 500 futures down only 0.7% despite Nasdaq weakness suggests rotation, not panic.
If Google's diversification signals a broader customer shift away from custom chips toward in-house silicon, and if AI capex growth is actually decelerating (not just normalizing), then this isn't a healthy pullback—it's the start of a demand cliff that will cascade through the entire supply chain.
"Broadcom's miss and Google share loss point to near-term re-rating risk for AI-exposed semis, but the move looks like healthy digestion after outsized gains rather than a demand cliff."
Broadcom's 15.1% premarket drop after missing estimates, alongside 7%+ slides in Micron and Marvell, reflects valuation compression after AI-driven runs rather than outright demand collapse. HSBC's cited worries on falling chip prices and slower AI rollout are credible near-term risks, especially with Broadcom losing Google share. Yet the piece omits that these names still guide to multi-year AI tailwinds and that analysts like Vinh see the pullback as expectations catching up, not fundamentals breaking. Futures imply a sector-wide de-rating that may overshoot if Q2 data confirms sustained spending.
The selloff could prove the start of a sustained AI capex slowdown if hyperscalers truly diversify away from Broadcom-style custom silicon, validating the price and rollout concerns HSBC flagged.
"The risk is not just a valuation reset, but potential hyperscaler migration away from Broadcom's custom silicon, threatening longer-term order momentum and margins."
Gemini’s 'valuation reset' framing risks underplaying a potential shift in hyperscaler behavior. If Google and peers accelerate in-house silicon or diversify away from Broadcom's custom ASICs, margin durability weakens and AI capex may prove more cyclical than advertised. A simple P/E catch-up ignores sustained orders attrition from large cloud customers and a multi-quarter margin squeeze, embedding slower recovery even if AI demand remains intact.
"The sell-off reflects a structural erosion of Broadcom's competitive moat due to hyperscaler in-house silicon initiatives, not just a valuation reset."
Claude and Gemini are underestimating the 'in-house' risk. Broadcom's valuation isn't just a multiple reset; it is a fundamental repricing of their moat. If hyperscalers like Google successfully internalize ASIC design, Broadcom shifts from a high-margin growth engine to a commoditized networking vendor. The market isn't just rotating; it is pricing in a permanent reduction in Broadcom's addressable market share. This is a structural threat to the 'AI infrastructure' thesis that goes well beyond temporary capex volatility.
"In-house silicon is a real headwind to Broadcom's custom ASIC margins, but networking and interconnect durability limits the structural damage to a subset of revenue, not the whole business."
Gemini's 'permanent moat erosion' thesis assumes hyperscalers successfully internalize ASIC design at scale—unproven. Google's custom TPU exists, but Broadcom still dominates high-bandwidth interconnect and networking layers that are harder to replicate. The real risk isn't commoditization; it's margin compression on *custom silicon only*, while core networking revenue remains sticky. That's a 20-30% revenue hit, not a structural collapse. Conflating two different threats inflates the downside.
"Broadcom's networking layers create a buffer that limits any ASIC share loss to partial margin pressure rather than full structural collapse."
Gemini overstates the structural threat by treating ASIC internalization as an existential moat breach. Even if Google captures more custom silicon, Broadcom retains high-margin networking and high-bandwidth interconnect layers that hyperscalers have not replicated at scale. The 20-30% revenue exposure Claude flags is real but contained; it does not automatically convert a growth compounder into a commoditized vendor. This distinction keeps the pullback closer to valuation reset than the permanent addressable-market contraction Gemini describes.
The panel discusses the recent sell-off in tech stocks, particularly Broadcom, Micron, and Marvell. While some attribute it to weaker earnings and fading AI-chip demand, others see it as a 'valuation reset' or temporary derisking after an outsized rally. The key debate revolves around the potential shift in hyperscalers' behavior, such as Google accelerating in-house silicon or diversifying away from Broadcom's custom ASICs, which could impact Broadcom's margins and market share.
Temporary pullback presenting a buying opportunity, as underlying demand for high-bandwidth memory and custom ASIC infrastructure remains intact.
Potential shift in hyperscalers' behavior towards in-house silicon, which could erode Broadcom's margins and market share.