JPMorgan Research Report Analysis: Marvell CY28 EPS Outlook Raised to $11, 14% Above Consensus

marsbitPublicado em 2026-08-25Última atualização em 2026-08-25

Resumo

JPMorgan raises Marvell's CY28 EPS estimate to $11, representing a 14% premium over the $9.64 consensus, following its Q2 FY27 preview report which reaffirms an Overweight rating. The report highlights Marvell's robust data center growth drivers, including strong demand for optical DSPs (1.6T/800G), customer adoption of Teralynx 10 switch chips, and the ramp of AWS's 3nm Trainium 3 ASIC. Microsoft's Maia XPU order book is noted as solid, with design work already underway for the next generation. JPMorgan expects Q3 FY27 revenue guidance to surpass consensus, nearing $3.1 billion, driven by these data center segments. The firm believes the market underappreciates Marvell's full-stack capabilities from optical DSPs to XPU/ASICs and switch silicon, further validated by its strategic partnerships with major cloud providers (Google, AWS, Microsoft). While risks include potential AI capex slowdowns and competitive dynamics, JPMorgan sees current valuation as factoring in caution, with upside potential outweighing downside risks. The price target is maintained at $240.

Author: Rita

While the market debates AI ASIC competition and optical communication market share, JPMorgan has already raised its earnings per share (EPS) estimate for Marvell Technology Group Ltd. for CY28 to $11, which is 14% higher than the market consensus of $9.64. On August 24, JPMorgan released a preview report for Marvell's fiscal 2027 second quarter (July quarter), reiterating its Overweight rating, noting that the company's data center growth trajectory has only strengthened over the past 90 days.

Google's custom chip partnership has been secured, AWS Trainium 3 is ramping up, and Microsoft's Maia XPU order book is robust—all three major cloud providers are now on Marvell's client list. JPMorgan believes the market has not yet fully priced in the company's comprehensive capabilities, spanning from optical DSP to XPU ASIC and switch chips.

July Quarter Results Expected to Meet or Slightly Exceed Expectations

Marvell's fiscal 2027 second quarter (July quarter) results are expected to meet or slightly exceed market expectations. Demand for optical DSPs remains strong, with both 1.6T and 800G generations ramping simultaneously. The Teralynx 10 switch chip continues to gain customer traction, and the custom chip business is beginning to benefit from the early ramp of AWS's 3nm Trainium 3 project.

JPMorgan expects Trainium 3 shipments to accelerate significantly in the second half of the year, while Trainium 2 phases out concurrently, creating a generational transition. The combined effect of both product generations is expected to significantly boost Marvell's data center revenue. The growth trends for optical DSPs and switch chips are also seen as solid.

October Quarter Revenue Guidance Expected to Beat Consensus

The market consensus for fiscal 2027 third quarter (October quarter) revenue is $3.028 billion. JPMorgan expects Marvell's official guidance to be around $3.1 billion, implying a quarter-on-quarter total revenue growth of 13% to 14%. Growth in the data center segment is anticipated to be even faster, projected at a 16% to 18% sequential increase.

This growth is driven by four factors: a more substantial ramp in Trainium 3 shipments, sustained strong demand for optical DSPs and switch chips, expansion of XPU-related opportunities, and the planned progression of the Maia 3nm project. The order book for the Maia 3nm project in calendar year 2027 is already quite firm, and design work for the next-generation Maia 2nm has commenced.

Data Center Long-Term Outlook Likely to Be Revised Upward

The market's focus is on the calendar year 2027 data center revenue outlook. The company's current expectation is year-over-year growth of 55%, but JPMorgan views this figure as conservative. The combined effect of six factors—continued ramp of optical DSPs, customer expansion for Teralynx 10, ramp of storage and CXL controllers, scale of Trainium 3, visibility for Maia, and broadening pipeline of XPU companion chips—suggests that the revenue and EPS outlook for calendar years 2026 to 2028 should be higher than the current consensus.

The public confirmation of the Google partnership further validates Marvell's strategic value within the custom chip ecosystem of cloud providers. JPMorgan emphasizes that this is not just a win for a single XPU project but represents the company's elevated role as an independent chip partner for hyperscale customers.

Recommendation and Risk Factors

JPMorgan reiterates its Overweight rating with a $240 price target. Achieving an EPS of $11 in calendar year 2028 implies a 14% premium to the current market consensus of $9.64, and this path is becoming increasingly clear.

Key risks center on the following: a potential marginal slowdown in AI capital expenditure could directly impact the data center business; changes in ASIC competitive dynamics could affect market share; fluctuations in optical communication market share require ongoing monitoring; and high customer concentration remains a structural topic of discussion. JPMorgan believes these risk factors have been extensively discussed by the market, and the current valuation already incorporates a significant degree of cautious expectation, suggesting greater upside potential than downside risk.

Disclaimer

This article is a compilation and interpretation by ChaoXiang Research of a third-party brokerage research report (JPMorgan, August 24, 2026), combined with publicly available market information. The ratings, price targets, earnings forecasts, and related judgments cited herein are the views of the brokerage's analysts, representing the stance of their institution only. They do not represent the views of ChaoXiang Research nor constitute any investment advice.

The market carries risks, and investment decisions should be made independently. This article should not be used as a basis for buying or selling any securities.

Perguntas relacionadas

QWhat is J.P. Morgan's new CY28 EPS estimate for Marvell Technology, and how does it compare to the consensus?

AJ.P. Morgan's new CY28 EPS estimate for Marvell Technology is $11, which is 14% higher than the market consensus of $9.64.

QWhat are the main growth drivers for Marvell's data center business cited in the report?

AThe main growth drivers for Marvell's data center business are: strong demand for optical DSPs (1.6T/800G), customer adoption of Teralynx 10 switch chips, ramping volume of the AWS 3nm Trainium 3 project, and a solid order book for the Microsoft Maia 3nm/2nm XPU.

QWhat is the expected revenue guidance for Marvell's Q3 FY2027 (October quarter), according to J.P. Morgan's analysis?

AJ.P. Morgan expects Marvell's official revenue guidance for Q3 FY2027 (October quarter) to be around $3.1 billion, exceeding the market consensus of $3.028 billion.

QWhat risks to Marvell's outlook does the J.P. Morgan report highlight?

AThe highlighted risks include: potential marginal slowdown in AI capital expenditures impacting the data center segment; changes in the ASIC competitive landscape affecting market share; fluctuations in optical communications market share; and the structural issue of high customer concentration.

QWhat is J.P. Morgan's investment rating and price target for Marvell Technology?

AJ.P. Morgan reiterates an Overweight rating for Marvell Technology with a price target of $240.

Leituras Relacionadas

Meta Faces a $1.4 Trillion Penalty: Algorithmic Recommendation in the Dock. Will the Rules Change in the Second Half of the Internet Era?

A landmark federal trial in Oakland, California, has begun against Meta, with 29 states accusing the company of harming children and teens through its social media platform designs. The states seek a potential maximum penalty of $1.4 trillion, calculated from alleged repeated violations of consumer protection laws and the Children's Online Privacy Protection Act (COPPA) involving millions of underage users. The core legal strategy bypasses the traditional shield of Section 230 by targeting Meta's own platform features—like its recommendation algorithms, infinite scroll, and "like" buttons—rather than user-generated content. The plaintiffs argue these designs are addictive and deceptive. Meta denies the claims, calling the penalty "unprecedented" and disputing the methodology. While the astronomical $1.4 trillion figure is seen as a starting point for negotiations, even a significantly reduced penalty in the hundreds of billions could establish a critical precedent. The case focuses on holding platforms legally responsible for their algorithmic designs. A ruling against Meta could force product changes, such as removing "likes" or imposing usage limits, and provide a legal template for similar suits against other tech giants like TikTok and YouTube. The trial's outcome may redefine accountability for algorithm-driven business models across the internet.

marsbitHá 6m

Meta Faces a $1.4 Trillion Penalty: Algorithmic Recommendation in the Dock. Will the Rules Change in the Second Half of the Internet Era?

marsbitHá 6m

A Precedent for the Prediction Markets: CFTC Aims to Prove Authority Over Polymarket via Criminal Case

A U.S. Army soldier, Gannon Ken Van Dyke, who earned $400,000 using non-public information on the crypto event-prediction platform Polymarket, is contesting the U.S. Commodity Futures Trading Commission's (CFTC) attempt to intervene in his criminal case. The CFTC, which initiated a separate civil suit against Van Dyke but has not pursued it to trial, seeks to submit arguments in the ongoing criminal proceeding. Van Dyke's lawyers filed a motion opposing this, arguing the CFTC is improperly trying to influence the criminal case as an outside expert rather than through its own lawsuit. They contend that event contracts traded on platforms like Polymarket are not swaps and thus fall outside the CFTC's jurisdiction, calling the regulator's tactics a procedural maneuver. Van Dyke faces criminal fraud charges for allegedly trading based on confidential information about potential regime change in Venezuela. His civil case with the CFTC is stayed pending the criminal trial, expected in late 2026 or early 2027. The court's decision on whether to allow the CFTC's intervention could set a precedent for how such event contracts are regulated. An AI analysis notes inconsistency in the CFTC's approach compared to a similar case involving a Google employee, raising questions about the legal strategy for the prediction market industry, which includes platforms like Kalshi. The outcome may define future regulatory standards for the entire market.

cryptonews.ruHá 11m

A Precedent for the Prediction Markets: CFTC Aims to Prove Authority Over Polymarket via Criminal Case

cryptonews.ruHá 11m

Nvidia's Jensen Huang, Fei-Fei Li, and Bin Lin All Invest in the Same Robotics Company

Generalist, a San Francisco-based AI startup focused on developing a "universal brain" for robots, has rapidly raised significant funding, attracting high-profile investors including Nvidia, Jeff Bezos' Bezos Expeditions, AI pioneer Fei-Fei Li, and Xiaomi co-founder Bin Lin. Founded in 2024 by former Google DeepMind researchers Pete Florence, Andy Zeng, and former Boston Dynamics engineer Andrew Barry, the company has secured approximately $750 million across four rounds, reaching a valuation surpassing $2 billion. The company's core innovation is its GEN-1.5 model, which aims to give robots powerful generalization capabilities. Unlike traditional systems requiring extensive reprogramming for each new task, GEN-1.5 can learn from a single, brief human demonstration (3-12 seconds) without needing retraining or fine-tuning. This "physical prompting" allows a robot to perform tasks like opening jars or using new tools after seeing them done once. While current single-demo success rates average 59% on simple tasks (improving to 83% with minor fine-tuning), the technology promises to drastically reduce robot deployment time from months to seconds. Generalist's strategy is to provide the AI intelligence layer that can run on various robotic hardware, rather than building robots themselves. The founding team combines DeepMind's expertise in AI model generalization with Boston Dynamics' practical robotics experience, targeting the industry bottleneck where robot "brains" lag behind advanced "bodies." Capital sees potential in this approach for creating a versatile, low-margin-cost robotic base applicable across factories, warehouses, and labs. However, GEN-1.5 remains a work in progress. Its capabilities are currently demonstrated on short, simple tasks, and significant challenges remain for reliable, long-term operation in complex, unpredictable real-world environments. Commercial viability also depends on hardware costs, maintenance, and system integration. Nevertheless, Generalist represents a shift toward "generalist" robots that can quickly adapt to new tasks, potentially redefining the robotics landscape and posing future questions about the role of human labor.

marsbitHá 16m

Nvidia's Jensen Huang, Fei-Fei Li, and Bin Lin All Invest in the Same Robotics Company

marsbitHá 16m

Trading

Spot
活动图片