Overview Marvell Technology raised its fiscal 2028 revenue target to roughly $20 billion at its October 6 investor day, and the stock closed at $287.01, up 5.81%, after gaining close to 10% intraday, Overview Marvell Technology raised its fiscal 2028 revenue target to roughly $20 billion at its October 6 investor day, and the stock closed at $287.01, up 5.81%, after gaining close to 10% intraday,
Why Marvell Stock Is Surging: Google AI Chip Ties and the $20B Revenue Target
Overview
Marvell Technology raised its fiscal 2028 revenue target to roughly $20 billion at its October 6 investor day, and the stock closed at $287.01, up 5.81%, after gaining close to 10% intraday, leaving the company valued at about $251.7 billion according to StockAnalysis's MRVL data page. Broadcom rose about 4% in sympathy the same day.
The single headline number is not what reset expectations. Management also lifted its fiscal 2029 custom-chip revenue goal from $10 billion to more than $12 billion and framed a fiscal 2031 revenue range of $70 billion to $90 billion. Against a prior analyst estimate of roughly $46.85 billion for fiscal 2031, that target says management believes hyperscaler demand for in-house AI silicon is materially larger than sell-side models assume. Google sits at the center of that thesis, and the agreement the two companies signed in late July already ties order volume to equity through a warrant.
Key Takeaways
The raise is far wider than a normal beat. Reuters reported that the $20 billion fiscal 2028 target exceeds a Wall Street consensus of $18.2 billion, while the fiscal 2031 midpoint of $80 billion is roughly 1.7 times the analyst estimate.
Custom silicon targets moved up by a fifth. Per Investing.com's account of the investor day, the fiscal 2029 custom-chip goal rose from $10 billion to above $12 billion, with a fiscal 2031 earnings target of $30 or more per share.
Interconnect, not custom compute, is the largest bucket. Within the $80 billion midpoint for fiscal 2031, interconnect accounts for roughly $37.5 billion against about $30 billion for custom compute and $10 billion for switching and storage, so the AI story here is as much about connectivity as about ASICs.
The $120 billion figure has a precise origin. Marvell's 8-K filed with the Securities and Exchange Commission sets warrant vesting at one tranche per $500 million of custom product revenue across 240 tranches, which is exactly the cumulative revenue behind the widely quoted headline.
The gap to Broadcom remains wide. Broadcom's third-quarter results filed with regulators show $16.7 billion of AI semiconductor revenue in a single quarter, more than Marvell's full-year custom-chip target for fiscal 2029.
Customer concentration is the other side of the thesis. Marvell's fiscal 2026 annual report discloses one direct customer at 14% of net revenue and one distributor at 37%, together more than half the business.
The Numbers That Actually Moved Expectations
From Fiscal 2028 to Fiscal 2031
The event took place in New York, as set out in the August announcement. The target structure runs on several levels: roughly $20 billion of revenue in fiscal 2028, of which about $18 billion is data center; more than $12 billion of custom-chip revenue in fiscal 2029; and $70 billion to $90 billion of revenue in fiscal 2031 with earnings of $30 or more per share. Chief Executive Matt Murphy also put the addressable market at about $400 billion by 2030.
Those figures need the current base for context. Marvell's second-quarter filing for the period ended August 1, 2026 shows record revenue of $2.739 billion, up 37% year over year, with data center at $2.172 billion, or 79% of the total, up 46%, and third-quarter guidance of $3.15 billion plus or minus 5%. Annualizing the current run rate puts the business near $12 billion, so reaching $20 billion by fiscal 2028 implies something close to a doubling over two years, and the fiscal 2031 midpoint implies quadrupling again from there.
Targets are not results, but they move the valuation anchor. The market had been pricing Marvell against roughly $46.85 billion of fiscal 2031 revenue and about $19 of earnings per share. Management substituted $80 billion and $30. The move in the share price is the first instalment of the market's discount applied to that new anchor.
Market Reaction and the Broadcom Read-Across
The session was not one-directional. Investing.com records the stock down 3% as the meeting opened, up as much as 10% at the peak and closing 5.90% higher. StockAnalysis puts the close at $287.01, up 5.81%, against a 52-week range of $70.69 to $329.88, a spread that itself shows how violently this company has been repriced over the past year.
The spillover matters more than the move. Reuters noted Broadcom rose about 4% the same day. The two do not compete for the same sockets, but they share one assumption: hyperscalers will keep shifting AI workloads toward custom silicon rather than general-purpose GPUs. Marvell quantified that assumption over five years, and the market applied it to Broadcom by extension.
Why Custom Silicon Became the Main Story
How ASICs and General-Purpose GPUs Divide the Work
Custom silicon in this context means an application-specific integrated circuit, co-developed by a chip supplier and a cloud provider for one customer's workloads. It trades flexibility for performance per watt and lower unit cost on the target task. Google's tensor processing units are the most mature example. According to Google Cloud's official release notes, TPU7x, the first member of the seventh-generation Ironwood family, entered preview on November 24, 2025 and reached general availability on March 31, 2026, supporting training and inference for large language models, mixture-of-experts models and diffusion models.
The distinction worth preserving is division of labor rather than substitution. A general-purpose GPU earns its premium through adaptability, since architectures, frameworks and kernels change every few months, while an ASIC takes one to two years from specification to volume. For training runs against moving targets, the GPU remains the lower-risk choice. For inference at stable architectures running the same operation endlessly, the cost advantage of custom silicon fully lands. That is why Google expands its own accelerators while continuing to buy from Nvidia.
Inference Economics Did the Repricing
The variable behind this wave is the sheer scale of inference. Nvidia reported revenue of $96.2 billion for the quarter ended July 26, 2026, including $89.0 billion from data center, up 117% year over year, in its second-quarter results, where Chief Executive Jensen Huang described the moment as one where compute is revenue and demand is accelerating. At that volume of token generation, a single percentage point of unit cost compounds into billions of dollars, and the payback on a dedicated design becomes straightforward.
Capital budgets corroborate it. On Alphabet's second-quarter call, Chief Financial Officer Anat Ashkenazi raised full-year 2026 capital expenditure guidance to $195 billion to $205 billion from $180 billion to $190 billion, with roughly 60% directed at servers and 40% at data centers and networking equipment, per the transcript published by Investing.com. Management also noted that external TPU system sales already sit in the cloud backlog with most revenue expected in 2027. For suppliers, that is both the source of order visibility and the reason targets are set on fiscal 2028 rather than today.
What the Google Agreement Actually Binds
A Warrant That Converts Revenue Into Equity
The 8-K rewards a close read. Marvell and Google entered a commercial partnership on July 29, 2026 covering development of custom semiconductor products including AI inference accelerators, storage controllers and network interface controllers. As part of it, Marvell issued Google a warrant over 58,970,907 common shares at an exercise price of $206.58, expiring August 18, 2033.
Vesting splits in two. A time-based portion of 1,360,867 shares vests quarterly across the first year. The remainder vests against revenue, divided into 240 equal tranches, with one tranche vesting per $500 million of custom product revenue from the third quarter of fiscal 2027 through fiscal 2033.
Where the $120 Billion Comes From
Run the vesting schedule through simple arithmetic and the headline resolves itself. Two hundred and forty tranches at a $500 million revenue threshold each equals $120 billion, matching Reuters' description of a partnership that could generate up to roughly $120 billion in sales through fiscal 2033 if performance milestones are achieved. The number describes the cumulative revenue required for full vesting. It is a ceiling assumption, not a signed purchase commitment.
The structure cuts both ways for shareholders. It aligns the customer with the supplier's equity and lowers the odds of the socket being handed to a rival. It also means that the more revenue arrives, the more dilution arrives with it. At the October 6 close of $287.01 against the $206.58 strike, each tranche of roughly 240,000 shares carries about $19 million of intrinsic value, close to 4% of the $500 million of revenue that triggers it. That cost never shows up in gross margin, but it lives permanently in the denominator of earnings per share.
Where Marvell, Broadcom and Nvidia Actually Stand
The Distance to Broadcom
Broadcom still owns this category by a wide margin. Its third quarter ended August 2, 2026 delivered revenue of $29.6 billion, up 86%, including $16.7 billion of AI semiconductor revenue, up 221% year over year and 54% sequentially, with fourth-quarter guidance of $34.8 billion in total revenue and $21.7 billion from AI semiconductors. Chief Executive Hock Tan said demand for custom AI accelerators and networking remains very strong.
Placed side by side, the scale gap is unmistakable: Broadcom books more AI semiconductor revenue in one quarter than Marvell targets from custom silicon in all of fiscal 2029. The investor day is therefore less a challenge to Broadcom than a declaration that Marvell has moved from single-customer supplier to credible second source across several hyperscalers.
Substitution or Diversion for Nvidia
Reading custom silicon as a replacement for general-purpose GPUs is the most common error on this topic. Nvidia's $89.0 billion of quarterly data center revenue exceeds the $80 billion midpoint of Marvell's full-year fiscal 2031 target, and that Marvell figure still includes interconnect, switching and storage. Even if Marvell and Broadcom hit every goal, custom silicon takes a share of incremental budgets rather than the installed base.
Diversion describes it better. Training clusters stay anchored on general-purpose GPUs while custom parts take a growing share of inference and of architectures that have stopped moving. Between those two pools of compute, connectivity is the common requirement. Marvell placing its single largest fiscal 2031 bucket in interconnect rather than custom compute is a statement about exactly that: whoever supplies the accelerators, data still has to move between racks and clusters, and demand for optics and network interfaces scales with total compute rather than with who wins the socket.
AI silicon was never a single-ticker trade. Follow the whole chain on one US stocks board
Risks, Scenarios and What to Watch
Concentration and Dilution
Marvell disclosed in its fiscal 2026 annual report that for the year ended January 31, 2026 one direct customer accounted for 14% of net revenue and one distributor for 37%. The company's own risk language flags dependence on a few customers for a significant portion of revenue, notes that major customers represent an increasing percentage of revenue, and points to concentration in the data center end market along with the gain or loss of design wins. For a company projecting five years of growth from a handful of hyperscaler relationships, that risk factor carries more weight than usual.
Dilution is the second layer, and the revenue-linked vesting makes it mechanical: the better the business performs, the more certain the share count grows. Execution is the third, because custom silicon takes years from design win to volume and any slip in process nodes, advanced packaging capacity or optical supply pushes the schedule right. The fourth is the assumption set itself, since a $400 billion addressable market and the fiscal 2031 range both rest on AI capital spending continuing to compound, which in turn rests on cloud providers earning a return on it.
Three Scenarios
In the delivery case, custom products ramp on schedule at Google and other hyperscalers, warrant tranches vest steadily, fiscal 2028 revenue of $20 billion becomes a verifiable checkpoint, and the market gradually migrates its valuation anchor toward management's framework.
In the delayed case, the orders are real but volume slips a quarter or two. The shape of the revenue curve changes while the endpoint does not, and the damage concentrates in a single guidance miss rather than in the long-term thesis.
In the broken-assumption case, AI capital spending decelerates or customers pull more design work in-house, and order visibility deteriorates. The warrant then loses its alignment function at the same moment customer concentration stops being a paper risk. The low-probability tail is the loss of a flagship design win to a competitor, which has precedent in this industry.
The Watchlist
The nearest test is Marvell's third-quarter report against guidance of $3.15 billion plus or minus 5%, where the commentary on the custom ramp matters as much as the print. Broadcom's fourth quarter follows, and whether its $21.7 billion AI semiconductor guide lands will set sentiment for the whole custom category. Then comes Alphabet's capital expenditure guidance and the pace at which external TPU system revenue converts during 2027. For readers tracking price action directly, the MRVL stock token contract market is one way to follow it, alongside the broader US equities section on MEXC.
Exclusive View from James Mitchell
For James Mitchell, the most informative line of the investor day is not $20 billion but the fiscal 2031 mix, where interconnect outweighs custom compute. The market trades Marvell as a custom silicon proxy while the company itself places its largest revenue bucket in connectivity. If the thesis rests on how large the Google order grows, it rests on what management ranks second. Connectivity demand scales with total AI compute regardless of which vendor wins the accelerator socket, which makes it a steadier exposure than the fight over ASIC share.
Three misreadings look likely. The first treats $120 billion as a contract value. The 8-K is explicit that 240 tranches at a $500 million revenue threshold produce that number, so it describes the condition for full warrant vesting, a revenue ceiling assumption rather than a signed order book. The second treats custom silicon as a GPU replacement. Nvidia books $89.0 billion of data center revenue in a quarter, more than the midpoint of Marvell's entire fiscal 2031 revenue target, which makes substitution the wrong frame and diversion of incremental inference budgets the right one. The third ignores the economics of the warrant. At the October 6 close, every $500 million of custom product revenue carries roughly $19 million of warrant intrinsic value, close to 4% of that revenue, a cost that bypasses gross margin and lands on the share count instead.
The variable worth tracking from here is vesting progress rather than the targets. Because tranches vest per $500 million of custom product revenue, the disclosed custom revenue line in coming quarters doubles as a yardstick for how fast the orders actually convert, and it is harder to dress up than any guidance range. The companion metric is gross margin, since custom ASIC work typically carries a lower margin structure than standard products. A widening gap between revenue growth and profit growth is normal during a mix shift toward custom, and the market usually starts repricing the first time it sees that gap rather than when it is announced.
The cross-asset implication is that profit distribution across the AI supply chain is rebalancing. For two years the compute narrative concentrated on one supplier, while custom accelerators, optical interconnect, network interfaces and storage controllers were treated as supporting roles. They are becoming distinct profit pools. For investors holding both crypto and technology equities, that structural shift is more actionable than picking a single winner: when profits spread from one node to a chain, exposure to the links generally beats concentration in the champion. These judgments rest on currently disclosed targets and data, and a single quarter of execution can change them.
FAQ
Why did Marvell stock jump on October 6?
The investor day did it. Management raised the fiscal 2028 revenue target to roughly $20 billion against a consensus of $18.2 billion, and framed fiscal 2031 revenue of $70 billion to $90 billion with earnings above $30 per share, both well ahead of prior analyst estimates. The stock closed at $287.01, up 5.81%, after trading nearly 10% higher intraday. Broadcom rose about 4% the same session, showing the market read the targets as a signal for the entire custom silicon category.
What exactly is the Marvell and Google arrangement?
Per Marvell's 8-K, the two companies entered a commercial partnership on July 29, 2026 covering development of custom semiconductor products including AI inference accelerators, storage controllers and network interface controllers. Marvell also issued Google a warrant over 58,970,907 shares at $206.58, expiring August 18, 2033. A small portion vests on time, and the rest vests against custom product revenue, one tranche per $500 million across 240 tranches.
Is the $120 billion figure an order value?
No. It comes from the vesting schedule: 240 tranches multiplied by the $500 million revenue threshold equals $120 billion. Reuters described the partnership as capable of generating up to roughly $120 billion in sales through fiscal 2033 if performance milestones are achieved. The number represents the cumulative revenue required for the warrant to vest in full, which makes it a ceiling assumption rather than a committed purchase contract.
What is custom silicon and how does it differ from an Nvidia GPU?
Custom silicon usually means an application-specific integrated circuit co-developed by a chip supplier and a cloud provider for one customer's workloads, trading flexibility for better performance per watt and lower unit cost on the target task. General-purpose GPUs earn their place through adaptability when model architectures and frameworks keep changing. In practice training stays anchored on GPUs while stable, high-volume inference suits custom parts, which is why Google expands its own accelerators and keeps buying from Nvidia.
Will custom chips replace Nvidia?
Not at current scale. Nvidia reported $89.0 billion of data center revenue in the quarter ended July 26, 2026, up 117%, and that single quarter exceeds the $80 billion midpoint of Marvell's full-year fiscal 2031 revenue target, which also includes interconnect, switching and storage. Diversion is the better description than substitution, since custom silicon competes for the inference-heavy slice of incremental budgets while optics and network interfaces grow with total compute regardless of vendor.
Is Marvell or Broadcom bigger in custom chips?
Broadcom leads clearly for now. Its third quarter ended August 2, 2026 produced $16.7 billion of AI semiconductor revenue, up 221% year over year, with fourth-quarter guidance of $21.7 billion. Marvell's target for all of fiscal 2029 is more than $12 billion. In other words, Broadcom books more in a quarter than Marvell aims to book in a year three years out, which positions Marvell as the credible second source to hyperscalers rather than the category leader.
What risks come with this story?
Customer concentration sits at the top. Marvell's fiscal 2026 annual report discloses one direct customer at 14% of net revenue and one distributor at 37%, and the company flags rising dependence on a few customers in its own risk factors. Warrant dilution follows, and because vesting tracks revenue, success makes dilution more certain. Beyond that are execution risk, since custom designs take years to reach volume, and assumption risk, since the long-term targets depend on AI capital spending continuing to compound.
Disclaimer
The information above is provided for general market information and analysis only and does not constitute investment advice, financial advice, legal advice, tax advice or a recommendation to trade. Equities, crypto assets and other related financial assets can move sharply, and company targets, guidance, past performance and industry data do not guarantee future results. The financial figures, regulatory filings, prices and market reactions cited here reflect publicly available information at the time of publication and may change, so the latest company announcements and regulatory filings should be treated as authoritative. Readers should conduct their own research and make decisions based on their own financial circumstances, investment objectives and risk tolerance, consulting a qualified professional where appropriate. The MEXC Crypto Pulse team accepts no liability for any direct or indirect loss arising from the use of this information.
About the Author
James Mitchell specializes in technical analysis, market trends, and trading strategies for both Bitcoin and altcoins. Based in London, he has over 10 years of experience in financial markets. Before joining MEXC Learn, James worked as a senior analyst at a leading European investment firm, where he developed expertise in risk management and quantitative trading. His transition to cryptocurrency markets began in 2017, and he has since become recognized for his data-driven approach. He holds a Master's degree in Financial Economics from the London School of Economics. His analytical approach combines traditional technical analysis with on-chain metrics to provide readers with actionable insights.
Areas of Expertise: Technical Analysis, Market Trends and Cycles, Trading Strategies, Bitcoin and Altcoin Analysis, Risk Management.
Research References
Want the fastest access to MEXC's latest updates? Join our official Telegram group now!
Account Verification: Understand KYC | How to Complete KYC
Market Opportunity
USD.AI Price(CHIP)
$0.05089
$0.05089$0.05089
USD
USD.AI (CHIP) Live Price Chart
The articles shared on this page are sourced from public platforms and are provided for reference only. They do not represent the position or views of MEXC. All rights belong to James Mitchell. If you believe any content infringes upon the rights of a third party, please contact service@support.mexc.com for prompt removal. MEXC does not guarantee the accuracy, completeness, or timeliness of any content and is not responsible for any actions taken based on the information provided. The content does not constitute financial, legal, or other professional advice, nor should it be interpreted as a recommendation or endorsement by MEXC. For expert insights and in-depth analysis, visit MEXC Learn.
Latest Updates on USD.AI
View More
Breaking: Samsung Dramatically Postpones Massive Strike, Korean Chip Stocks Surge, MEXC Launches Samsung and SK Hynix Futures
In the late hours of May 20, 2026, Samsung Electronics reached a tentative wage agreement with its union, averting an originally planned 18‑day strike. On May 21, the KOSPI surged 8.42%, with Samsung Electronics closing up 8.51% at an all‑time high and SK Hynix rising 11.17%. Both companies posted record first‑quarter earnings: Samsung reported revenue of 133.9 trillion won, while SK Hynix surpassed 50 trillion won in quarterly revenue for the first time, driven by strong HBM demand. Positive catalysts also included NVIDIA’s earnings report and South Korean semiconductor policy initiatives. MEXC has launched perpetual contracts for Samsung and SK Hynix, supporting USDT trading and 24/7 operation.
2026/05/21
NVIDIA Annual Meeting 2026: Jensen Huang’s AI Factory Thesis Faces Its Next Market Test
NVIDIA held its 2026 Annual Meeting of Stockholders online on June 24, 2026, at 9:00 a.m. Pacific Time. The formal agenda included the election of directors, advisory approval of executive compensation, auditor ratification, and several shareholder proposals. However, for the market, the meeting was about more than annual governance. It was another test of whether NVIDIA can continue to defend the AI infrastructure trade after a historic year of revenue growth.
Although some Asia-Pacific investors followed the NVIDIA annual meeting on June 25 local time, the official meeting schedule was June 24 Pacific Time. According to NVIDIA’s 2026 Annual Meeting of Stockholders page, the event was held virtually, with shareholders able to participate online.
The key question for investors is no longer simply whether AI chip demand is strong. That has already been proven by NVIDIA’s fiscal 2026 results. The bigger question is whether AI data center spending can keep converting into lower token costs, higher inference activity, and durable earnings power. In other words, the 2026 annual meeting was not just a shareholder vote. It was a fresh checkpoint for Jensen Huang’s “AI factory” thesis.
2026/06/25
Are Memory Giants Becoming AI’s New Toll Booth? Korea’s Expansion Plan, DRAM Lawsuit and Micron’s Wild Reversal Show the Trade Is Changing
The AI supply chain debate is shifting from GPUs to memory.
South Korea has announced a massive semiconductor and AI investment push, with Samsung Electronics and SK Hynix each set to build two new large-scale chip fabrication facilities in the country’s southwest region. The broader project is designed to create a chip production ecosystem valued at roughly 800 trillion won, or about $518 billion, as Seoul tries to strengthen its position in the global AI infrastructure race.
At almost the same time, Samsung, SK Hynix and Micron Technology were hit with a U.S. class-action antitrust lawsuit alleging that the three memory giants coordinated to restrict conventional DRAM supply while shifting capacity toward high-bandwidth memory, or HBM. The allegations have not been proven in court, and Micron has denied the claims, saying it competes fairly and lawfully.
Micron stock became the market’s live stress test. MU fell as low as about $1,024.86 intraday on June 29 before closing at $1,145.28, slightly above the previous close. The move suggests that investors are not abandoning the AI memory bull case, but they are starting to price its side effects: legal risk, downstream cost pressure and the possibility that memory inflation could become a problem for the wider AI economy.
2026/06/30
You May Also Like
HOT
Currently trending cryptocurrencies that are gaining significant market attention
Crypto Prices
The cryptocurrencies with the highest trading volume
Newly Added
Recently listed cryptocurrencies that are available for trading