Marvell Stock Surges 25% on Jensen Huang's 'Next Trillion-Dollar Company' Call: Nvidia's $2B Vote, the Teralynx T100, and the Custom Silicon War With Broadcom

What You'll Learn

• What Marvell reported in the first quarter of fiscal 2027, including revenue, earnings, margins and operating cash flow.
• What the NVIDIA NVLink Fusion partnership and $2 billion investment do and do not establish.
• How the Teralynx T100 and Celestial AI expand Marvell’s AI connectivity strategy.
• Which execution, customer, competition, acquisition and valuation risks should be monitored after the market reaction.

Marvell stock became a major AI-infrastructure discussion after a series of company announcements and a strong quarterly report. The safer way to read the story is to divide it into four layers: historical operating results, strategic announcements, management projections and market interpretation. Mixing those layers can make a product launch look like revenue or turn an executive’s forecast into a valuation target.

This article uses Marvell’s official investor-relations releases as the main evidence. The Marvell fiscal 2027 first-quarter release covers the reported financial period. The NVIDIA and Marvell NVLink Fusion announcement, the Teralynx T100 release and the Celestial AI acquisition release are used for the strategic sections.

Quick answer: what did Marvell report in fiscal 2027 Q1?

Marvell’s first quarter of fiscal 2027 ended May 2, 2026. Revenue of $2.418 billion was a record and grew 28% from the year-earlier quarter. The company said the result was $18.0 million above the midpoint of the guidance it had provided on March 5, 2026. Data center revenue was $1.833 billion, representing 76% of total revenue, while communications and other revenue was $585.1 million.

The GAAP and non-GAAP figures show why definitions matter. GAAP net income was $34.5 million, or $0.04 per diluted share. Non-GAAP net income was $718.0 million, or $0.80 per diluted share. GAAP gross margin was 52.1%, while non-GAAP gross margin was 58.9%. Marvell’s reconciliation excludes items such as stock-based compensation and amortization of acquired intangible assets from selected non-GAAP measures. The adjusted number should not replace the GAAP number.

Cash flow from operations was $638.8 million. Marvell also reported $3.844 billion of cash and cash equivalents at May 2, 2026, compared with $2.639 billion at January 31, 2026. The cash balance and operating cash flow provide useful context, but they do not remove debt, acquisition or semiconductor-cycle risk.

For context on AI infrastructure demand, readers can review our NVIDIA earnings analysis and Dell AI server revenue review. Those posts discuss different companies and should not be treated as independent confirmation of Marvell’s projections.

What the NVIDIA partnership changes

On March 31, 2026, NVIDIA and Marvell announced a strategic partnership that connects Marvell to NVIDIA’s AI factory and AI-RAN ecosystem through NVIDIA NVLink Fusion. The announcement says Marvell will provide custom XPUs and NVLink Fusion-compatible scale-up networking. NVIDIA will provide supporting technologies that include Vera CPUs, ConnectX networking, BlueField DPUs, NVLink interconnect and Spectrum-X switches.

NVIDIA also announced a $2 billion investment in Marvell. That investment is a strategic signal from a major AI-compute company. It may improve ecosystem access and increase the potential addressable market for Marvell’s custom silicon and connectivity products. It does not guarantee that every partnership opportunity will become a customer design win, a production shipment or recurring revenue.

The distinction is important for Marvell stock analysis. A partnership announcement describes a relationship and intended technology integration. It does not disclose the revenue contribution of each future product, the margin profile of a future program or the timing of customer adoption. The commercial outcome depends on design wins, product qualification, manufacturing, pricing and the customer’s own roadmap.

Marvell’s role can span custom silicon, high-performance analog, optical digital signal processing, silicon photonics and networking. NVIDIA’s ecosystem can create an opportunity for those products, but it also places Marvell inside a demanding platform and supply-chain environment. Readers should therefore track subsequent filings for product revenue, customer concentration, inventory and capital requirements rather than relying on the announcement alone.

What is the Teralynx T100?

On June 1, 2026, Marvell introduced the Teralynx T100 switch silicon for AI and cloud data-center infrastructure. Marvell describes it as a 102.4 Tbps product purpose-built for AI workloads. The company says the T100 uses a 3nm process, supports up to a 512-port scale-out radix, operates at under 1000W of typical power and can deliver up to 25% lower power than competitive solutions.

Those specifications are company claims from a product announcement. They are not the same as independently tested performance across a complete customer system. The T100 was scheduled to begin customer sampling during the quarter. Sampling is an important product milestone, but it does not establish volume shipments, recognized revenue or a particular customer win.

The technical rationale is straightforward. Large AI clusters move data among many accelerators, and the network can become a bottleneck for utilization, latency and power. A high-radix switch can reduce the number of network tiers in some architectures. The real result depends on optics, cables, packaging, software, interconnect protocols, power delivery and the customer’s deployment design.

Marvell says the T100 supports scale-out and scale-up fabrics and can work with emerging Ethernet and Ultra Ethernet Consortium requirements. It also describes programmable pipeline architecture, telemetry and traffic-management features. These capabilities support the product strategy, but the commercial test remains adoption. A product specification should not be rewritten as proof that Marvell will dominate AI networking.

Our HBM and AI-memory analysis provides another example of why the AI hardware chain has several layers. Memory, networking, optics, custom silicon and compute each have different customers, cycles and bottlenecks.

How Celestial AI expands the interconnect strategy

Marvell completed its acquisition of Celestial AI on February 2, 2026. The acquired company brought Photonic Fabric optical interconnect technology designed for high-bandwidth and low-latency connectivity across large-scale AI deployments. Marvell said the technology would become part of its Data Center Group.

Marvell’s acquisition release projects initial revenue contributions from Celestial AI in the second half of fiscal 2028. It projects a $500 million annualized revenue run rate in the fourth quarter of fiscal 2028 and a $1 billion annualized run rate by the fourth quarter of fiscal 2029. These are management projections. They are not reported Celestial revenue and should be evaluated against integration, customer adoption, production and timing risk.

The transaction also has financial consequences. Marvell said the acquisition would add approximately $50 million in annual non-GAAP operating expenses to the current run rate. It said the completion reduced cash by $1 billion and could reduce annual other income by approximately $38 million. Marvell also said equity issued for the acquisition increased diluted weighted-average shares by approximately 27 million. These figures belong in a transaction analysis, not in a simple product-growth headline.

Photonic interconnect may help move data between components over demanding distances and bandwidth levels. It can also create manufacturing, packaging, qualification and integration challenges. The acquisition broadens Marvell’s portfolio, while increasing the burden on management to integrate the technology and convert the opportunity into customer revenue.

For a wider explanation of autonomous systems and their infrastructure requirements, see our agentic AI guide. For an example of a high-growth technology company where projections must be separated from current results, see our Anthropic funding and IPO analysis.

What Marvell guided for fiscal 2027 Q2

Marvell’s official fiscal 2027 Q1 release guided to second-quarter net revenue of $2.700 billion plus or minus 5%. The release also gave a GAAP gross-margin range of 52.1% to 53.1% and a non-GAAP gross-margin range of 58.25% to 59.25%.

Marvell guided to GAAP diluted net income per share of $0.37 plus or minus $0.05 and non-GAAP diluted net income per share of $0.93 plus or minus $0.05. It guided to basic weighted-average shares of 899 million and diluted weighted-average shares of 915 million. Readers should compare those figures with the next reported quarter and preserve the distinction between GAAP and adjusted results.

Guidance is useful because it reveals management’s expectations at a particular date. It is not a guarantee. Actual results can change with customer schedules, semiconductor demand, supply availability, product qualification, acquisition integration, pricing and macroeconomic conditions. The Marvell quarterly-results page should be checked for subsequent disclosures before making a new statement about the company’s outlook.

Key risks after the AI-infrastructure reaction

Marvell’s release lists risks including demand estimation, reliance on major customers, the data-center end market, customer-developed solutions, design wins, supply-chain disruption, acquisitions, inventory, macroeconomic conditions, debt and cybersecurity. Those risks are not predictions that a problem will occur. They are the categories readers should review in filings and later earnings materials.

The market reaction also creates valuation risk. The original title and legacy body used a dramatic share-price move and a trillion-dollar framing. This article does not use a current price, market capitalization, P/E ratio or analyst target because those require a dated market-data pull. Jensen Huang’s public comment, as described in market coverage, should be treated as a forecast or opinion rather than a price target or an achieved valuation.

Readers can compare the macro angle with our Treasury-yield and AI-rally review and our AI-stocks research framework. Neither article establishes that Marvell’s products will meet a future valuation outcome.

How to evaluate the next Marvell update

A disciplined follow-up should begin with the fiscal period. Confirm whether the number is for the quarter ended May 2, 2026 or a later quarter. Next, separate revenue from bookings, product announcements and management outlook. A 102.4 Tbps product specification is not recognized revenue, and a projected Celestial AI run rate is not a reported operating result.

Then review the mix. Marvell’s Q1 fiscal 2027 data-center revenue was $1.833 billion and represented 76% of total revenue. Communications and other revenue was $585.1 million. The mix makes data-center demand important, but it also means a change in one end market can have an outsized effect on the company’s consolidated results.

Finally, read the reconciliation. Marvell’s non-GAAP measures exclude stock-based compensation, acquired-intangible amortization and other items. Adjusted results can be informative, but GAAP income, cash flow, balance-sheet obligations and dilution remain part of the economic picture. A good update should show both where the company delivered and where the evidence is still forward-looking.

Our NVIDIA results review and Dell infrastructure analysis can be used for context, but Marvell should be assessed on its own reporting definitions, product cycle and customer evidence.

Conclusion: strategic opportunity, execution still decisive

Marvell’s fiscal 2027 first quarter showed record revenue of $2.418 billion, up 28% year over year, alongside a 52.1% GAAP gross margin, $34.5 million of GAAP net income and $638.8 million of operating cash flow. The NVIDIA partnership and $2 billion investment provide strategic context. The Teralynx T100 and Celestial AI acquisition expand the company’s AI-connectivity portfolio.

The harder question is conversion. Product specifications must become qualified deployments. Partnerships must produce customer programs. Celestial AI projections must be tested against integration and revenue disclosures. Guidance must be compared with actual results. The market reaction may be important, but it does not replace filings, customer evidence or a dated valuation method.

This article explains reported results, company announcements, management projections and risk factors. It does not tell readers to buy, hold or sell MRVL. This is research and analysis only, not personalized financial advice.