Higher data center expectations support long-term growth, but Marvell remains Equal-weight due to its high valuation

Overview

The report assesses Marvell's second-quarter results, third-quarter guidance, and medium- to long-term AI data center outlook. Morgan Stanley views the results and guidance as solid and the long-term growth drivers as increasingly diversified, but most of the improvement remains consistent with management's previously outlined trajectory, while the Google partnership was already largely incorporated into FY28 expectations. With the valuation still elevated, the price target was raised to $246 while the Equal-weight rating was maintained.

Core views

Marvell delivered solid second-quarter results, but they did not constitute a significant near-term surprise. Revenue was $2.74 billion, up 37% year over year and 13% quarter over quarter, above the midpoint of guidance and slightly ahead of Morgan Stanley and market expectations. Data center revenue was $2.17 billion, up 46% year over year and 18% quarter over quarter, accounting for approximately 79% of total revenue; Communications & Other revenue was $568 million, up 10% year over year and down 3% quarter over quarter. Non-GAAP gross margin was 58.9%, slightly above expectations; EPS was $0.94, also slightly above Morgan Stanley and consensus expectations. The report therefore characterizes the quarter as “good, but largely consistent with management's prior expectations.” Third-quarter guidance indicates that growth will continue to accelerate. The midpoint of revenue guidance was $3.15 billion, representing growth of approximately 52% year over year and approximately 15% quarter over quarter, above the market expectation of $3.020 billion and Morgan Stanley's prior forecast of $3.009 billion. Data center revenue is expected to grow more than 20% quarter over quarter and approximately 75% year over year, with further sequential acceleration in the fourth quarter. The midpoint of EPS guidance was $1.10, above the market expectation of $1.07 and Morgan Stanley's forecast of $1.05. However, the midpoint of gross margin guidance was 58.0%, below the market's 58.5% and Morgan Stanley's 58.4%, indicating that rapid growth has not translated into a corresponding gross-margin beat. The medium- to long-term guidance was the principal positive element of this update. The company now expects FY27 revenue of approximately $12 billion, up 45% year over year and above the prior expectation of approximately $11.5 billion. The main reason is that the expected data center growth rate was raised from approximately 50% to approximately 60%, a change that Morgan Stanley estimates could contribute approximately 10% EPS upside. The FY28 revenue outlook was raised from $16.5 billion to approximately $18 billion, implying growth of approximately 50%, with data center revenue expected to grow more than 60% and custom business revenue expected to more than double. Importantly, the revenue contribution through FY28 from the recently expanded Google partnership is already incorporated into this custom business outlook, so the market cannot simply treat the partnership as incremental to the original guidance. Morgan Stanley raised its forecasts accordingly: FY2027 revenue, gross margin, and EPS were revised from $11.498 billion, 58.5%, and $3.98 to $11.979 billion, 58.3%, and $4.18, respectively; FY2028 estimates were revised from $16.161 billion, 57.2%, and $5.93 to $18.033 billion, 57.9%, and $6.68. Quarterly EPS forecasts were also increased, with third- and fourth-quarter estimates raised from $1.05 and $1.21 to $1.10 and $1.33, respectively. The report views the direction of the forecast revisions as clearly positive, although the near-term earnings trajectory remains broadly consistent with the information management provided before the earnings release. The growth mix is becoming broader and less dependent on a single large custom ASIC project. Recent momentum has primarily come from optical DSPs, scale-out switching, and the ramp-up of the custom business in the second half, while 1.6T optical products are expected to accelerate further in FY28. Longer-term opportunities include XPU attach products such as NICs, storage controllers, memory interface controllers, and near-memory computing, as well as scale-up connectivity. FY28 scale-up optical expectations have increased significantly due to NPO and CPO activity, and the company also participates in the UALink, ESUN, and NVLink Fusion switching ecosystems. Morgan Stanley believes this portfolio offers greater visibility than the model 18 months ago, which emphasized large, monolithic custom ASIC projects. It could reduce the impact of any single project's success or failure on overall growth and increase confidence in the sustainability of data center growth in FY29 and beyond. This strategic shift also addresses the competitive challenges facing the custom chip business. The report believes that competing directly with NVIDIA is not easy and that the related projects may not always succeed, while the current forecasts may still face headwinds from the Trainium relationship. In comparison, increasing optical, networking, storage, and memory connectivity content around XPU configurations can diversify the sources of growth. As AI systems become increasingly dependent on networking and memory, the report remains constructive on the revenue potential of optics, scale-up connectivity, and XPU attach products and believes Marvell could outgrow overall data center spending, led by its optical business. The near-term share-price outlook remains constrained by expectations and valuation. Investor expectations rose significantly after the Google partnership was announced, but management explicitly stated that the associated FY28 contribution had largely been incorporated into prior guidance. The report also notes that Marvell tends to provide optimistic long-term forecasts first and then rely on subsequent execution to deliver them, while other AI companies trade more around near-term earnings beats and upward estimate revisions. This makes the October investor day an important validation point. Morgan Stanley broadly agrees with the company's long-term optimism but would prefer long-term messaging that leaves more room for sustained near-term beats and raises. If the share price pulls back before the investor day, the report states that its tactical stance would become more constructive. On valuation, the price target was raised from $224 to $246, still based on 45 times CY27 ModelWare EPS of $5.47, equivalent to approximately 37 times CY27 non-GAAP EPS of $6.68. The report's bull, base, and bear cases are $376, $246, and $143, respectively. The bull case applies 60 times CY27 ModelWare EPS of $6.27 and assumes CY27 revenue exceeds $17 billion; the base case uses 45 times and $5.47; and the bear case uses 30 times and $4.77. Despite attractive long-term growth drivers, the current valuation remains above AI peers, and elevated stock-based compensation expense diminishes the stock's appeal. Morgan Stanley therefore prefers other AI names and maintains its Equal-weight rating. A more constructive stance would require continued upward estimate revisions, additional evidence of networking share gains and higher content per system, or greater certainty that custom and XPU attach growth will exceed the current framework.

Analysis framework

The report first compares quarterly actuals item by item with company guidance, Morgan Stanley forecasts, and market consensus expectations, and then assesses whether the next-quarter guidance represents a new upside surprise. It subsequently translates management's FY27 and FY28 revenue and data center growth outlooks into revisions to revenue, gross margin, and EPS forecasts. The long-term section breaks down growth sources across optics, switching, custom chips, scale-up connectivity, and XPU attach products, and evaluates whether business diversification can reduce the risks associated with a single ASIC project. Finally, it constructs bull, base, and bear valuation scenarios using CY27 earnings and forward P/E multiples, and determines the rating and price target based on the valuation relative to AI peers.

Methodology notes

• Valuation methodologyP/E and PEG valuation

  Forward P/E valuation based on CY27 earnings

  The report calculates its $246 base-case price target using 45 times CY27 ModelWare EPS of $5.47 and constructs bull and bear scenarios using multiples of 60 times and 30 times, respectively; it also uses approximately 37 times CY27 non-GAAP EPS as a comparable measure.

• Event-driven analysis and behavioral financeExpectation gap/expectation management

  Comparison of actual results, original company guidance, institutional forecasts, and market consensus expectations

  Rather than only assessing whether the figures grew, the report compares quarterly results and third-quarter guidance against management's prior commentary, Morgan Stanley forecasts, and market expectations, and uses those differences to determine whether the Google partnership has already been reflected in market expectations and the company's outlook.

• (Out-of-vocabulary methodology)

  Morgan Stanley ModelWare framework

  The report uses the ModelWare framework to prepare earnings forecasts and financial statements and distinguishes between ModelWare EPS and non-GAAP EPS; the price target is based directly on ModelWare EPS.

• (Out-of-vocabulary methodology)

  Bull, base, and bear scenario analysis

  The report applies different assumptions to revenue and earnings outcomes, then assigns different CY27 ModelWare EPS figures and valuation multiples to each scenario to derive a risk-reward range of $376, $246, and $143.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Marvell Technology Group Ltd (MRVL.O / MRVL US)

  Rising networking and memory demand in AI systems creates growth opportunities for its optical DSPs, switching, scale-up connectivity, custom chips, and XPU attach products.

  Strengths

  Data center revenue already accounts for approximately 79% of total revenue; optics, scale-out switching, the custom business, and multiple XPU attach products form a more diversified growth portfolio.

  Weaknesses

  The valuation is above AI peers, stock-based compensation expense is elevated, and portions of the storage, enterprise data center, and networking businesses remain weak.

  Comparison

  The report believes its base-case valuation is close to that of high-growth AI semiconductor companies, but the current valuation remains elevated and states that it prefers other AI companies at the current share-price level.

  Risks

  AI and custom projects may fall short of expectations, networking inventory and enterprise data center weakness may persist, and the Trainium relationship may continue to create headwinds.

Key data

• Second-quarter revenue$2.74bnUp 37% year over year and 13% quarter over quarter, slightly above Morgan Stanley and market expectations.
• Second-quarter data center revenue$2.17bnUp 46% year over year and 18% quarter over quarter, accounting for approximately 79% of total revenue.
• Second-quarter Communications & Other revenue$568mnUp 10% year over year and down 3% quarter over quarter.
• Second-quarter non-GAAP gross margin and EPS58.9%; $0.94Both were slightly above Morgan Stanley and market expectations.
• Third-quarter revenue guidance$3.15bnThe midpoint represents growth of approximately 52% year over year and approximately 15% quarter over quarter, above the market forecast of $3.020bn and Morgan Stanley's forecast of $3.009bn.
• Third-quarter gross margin guidance58.0%The midpoint is below the market forecast of 58.5% and Morgan Stanley's forecast of 58.4%.
• Third-quarter EPS guidance$1.10The midpoint is above the market forecast of $1.07 and Morgan Stanley's forecast of $1.05.
• Company FY27 revenue outlookApproximately $12bn, +45% year over yearPreviously approximately $11.5bn; data center growth expectations were raised from approximately 50% to approximately 60%.
• Company FY28 revenue outlookApproximately $18bn, approximately +50%Previously $16.5bn; data center revenue is expected to grow more than 60%, while the custom business is expected to more than double.
• Morgan Stanley FY2027 forecastRevenue $11.979bn / gross margin 58.3% / EPS $4.18Previously $11.498bn, 58.5%, and $3.98, respectively.
• Morgan Stanley FY2028 forecastRevenue $18.033bn / gross margin 57.9% / EPS $6.68Previously $16.161bn, 57.2%, and $5.93, respectively.
• Base case$24645 times CY27e ModelWare EPS of $5.47.
• Bull case$37660 times CY27e ModelWare EPS of $6.27, assuming CY27 revenue exceeds $17bn.
• Bear case$14330 times CY27e ModelWare EPS of $4.77.
• Current share price and price target$241.45 / $246.00The share price is the August 27, 2026 closing price; the price target was raised from $224.

Impact & implications

The report believes that visibility into Marvell's data center growth is improving, with optics, switching, scale-up connectivity, and multiple categories of XPU attach products reducing dependence on a single large custom ASIC project and enhancing the sustainability of growth in FY29 and beyond. However, the results and outlook largely followed management's previously outlined trajectory, and the Google partnership was already broadly included in the FY28 framework. The high valuation means that optimistic long-term commentary alone is insufficient to support a more constructive rating; continued upward earnings estimate revisions and evidence of share gains and higher content will be needed.

Risks

• The current valuation remains above AI peers; if earnings estimates do not continue to rise, the rating and price target may come under pressure.
• Elevated stock-based compensation expense reduces the report's enthusiasm for the stock.
• The AI opportunity may be smaller than expected, while the scale and timing of cloud custom chip or XPU attach projects may also fall short of the current framework.
• The storage and enterprise data center businesses remain weak, with limited visibility into the timing of a recovery.
• Customer and channel inventory may continue to weigh on the networking business and offset growth in the custom ASIC business.
• The report believes that the current forecasts still face headwinds from the Trainium relationship.
• If the Storage and Networking businesses underperform the report's expectations, the enterprise data center and networking businesses may continue to weigh on overall results.
• In an extreme bear-case scenario, the share price could fall excessively toward tangible book value.

What to watch

• Watch whether the October investor day provides new evidence of growth beyond the existing FY28 framework.
• Track whether data center revenue achieves third-quarter growth of more than 20% quarter over quarter and approximately 75% year over year, followed by further acceleration in the fourth quarter.
• Monitor the FY28 ramp-up progress of 1.6T optics, NPO, CPO, and scale-up connectivity projects.
• Track whether networking share and content per system continue to increase.
• Watch whether the custom chip and XPU attach businesses can exceed the current growth framework.
• Monitor whether earnings estimates continue to rise and whether new incremental opportunities emerge beyond the Google partnership.
• Track the recovery of the Storage, Networking, and enterprise data center markets.