Marvell Drops 10%, Nvidia Decouples as Fed Fear Returns

Tech stocks pulled back on August 30 as Federal Reserve speculation resurfaced, dragging the Nasdaq down 0.52% and hitting chip stocks hard. Marvell took the worst of it, while Nvidia once again moved to its own beat.

Key points

• Marvell Technology (MRVL) fell 10.3% intraday after B. Riley Financial cut its price target from $345 to $315, though the firm kept a buy rating. The stock opened at $225.26, down from a prior close of $241.45.
• The Nasdaq lost 0.52% as Fed Governor Warsh revived rate-hike trade concerns. Software held up relatively well; chip names absorbed the bulk of the selling.
• Nvidia’s correlation to other chip stocks has dropped to just 0.03, a striking decoupling from semiconductor peers.
• Applied Materials (AMAT) slid 4.3% on insider selling activity, with around 6.4 million shares changing hands, roughly 22% below average daily volume.
• Nvidia CEO Jensen Huang publicly shrugged off OpenAI’s in-house “Jalapeño” chip, saying “lots of projects get started, lots of projects get canceled.”

What happened to Marvell today?

The selloff in MRVL is sharp enough to warrant context. B. Riley Financial cut its price target by $30 to $315 but held its buy rating, which means this is a trimming of expectations rather than a reversal of conviction. The stock trading at $216.62 intraday puts it well below even the reduced target, suggesting the market is pricing in something more pessimistic than the analyst’s base case.

This follows an already bruising stretch for Marvell. Prior coverage from August 28 flagged that timing concerns around its Google custom silicon deal had already weighed on the stock. Today’s gap-down open compounds that pressure. Investors sitting on MRVL positions are dealing with back-to-back negative sessions driven by different catalysts: deal skepticism first, analyst repricing now.

The macro backdrop made things worse. Fed rate-hike speculation, attributed to commentary from Kevin Warsh, pushed yields higher and compressed multiples across growth-heavy tech. Chip stocks, which carry elevated forward earnings expectations, are acutely sensitive to that kind of repricing.

Why is Nvidia moving differently from the rest?

With a 30-day correlation to semiconductor peers of just 0.03, Nvidia is effectively trading as its own asset class right now. That is an unusual reading. For context, a correlation of 1.0 would mean it moves in perfect lockstep with peers; 0.03 is near zero.

Part of the explanation is structural. As Livemint noted, overall AI compute demand is growing fast enough that even as hyperscalers build custom silicon, Nvidia’s total addressable volume keeps expanding. The custom chip narrative that is weighing on peers (including Marvell, Broadcom, and ASIC designers) does not bite Nvidia the same way because it is not a contract chip designer. Its risk profile is different.

Huang’s comments on the OpenAI Jalapeño chip are consistent with this posture. He framed the in-house chip effort as routine industry behavior, neither threatening nor surprising. Whether that confidence is warranted depends on how quickly OpenAI and others can actually displace Nvidia training clusters at scale, which remains an open question over the next two to three years.

What else is moving in the AI ecosystem?

Apple and Corning announced plans for a new joint innovation center in central Kentucky. The details available are limited, but the partnership reflects Apple’s ongoing push to deepen its domestic manufacturing and materials supply chain. Corning is a key supplier of specialty glass used in Apple devices. For AI investors, the angle is indirect: Apple’s hardware roadmap and any AI-enabled device features depend on the underlying materials and manufacturing ecosystem this kind of collaboration supports.

OpenAI, meanwhile, announced it will end its partnership with Cursor following SpaceX’s acquisition of the AI coding tool, citing concerns over contract compliance. The move signals that OpenAI is drawing sharper boundaries around its partnerships as its competitive landscape shifts post the SpaceX deal.

On the retail investor side, South Korean individual investors bought a net $7.04 billion in U.S. stocks over July and the first part of August, rotating out of a correcting KOSPI. Their top pick was the SOXL semiconductor 3x leveraged ETF at $3 billion in net purchases. They also added $514.96 million in SpaceX shares. Notably, they were net sellers of Nvidia ($748.9 million) and Palantir ($615.4 million) over the same period, a reminder that retail flows can cut against headline momentum names even during strong runs.

Today’s session is a useful reminder that AI stocks are not monolithic. A single analyst note moved Marvell 10%. Nvidia barely flinched. The divergence between custom silicon plays and vertically integrated GPU platforms is becoming one of the defining fault lines in AI stock performance heading into fall.