Marvell Technology: A Deep Long-Term Value Study

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

• QHow high is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?6/10

  A

  The ceiling is high, but the essence is that Marvell is taking a larger share of an existing pie that is expanding explosively, rather than creating a brand-new market from scratch. What Baillie Gifford LTGG cares about most is whether a market can hold a company capable of rising fivefold in ten years. The AI data center infrastructure markets where Marvell operates, including custom XPUs, optical interconnects, and switching/networking, are indeed large and durable enough. But Marvell is a key component supplier among the picks-and-shovels vendors. The pie is being created by others, namely hyperscale cloud companies' AI capital spending, and Marvell is capturing the silicon and optics slice of it.

  Start with the size of the pie. The report notes that management said data center revenue grew from about USD 200 million to more than USD 6 billion over ten years, accounted for about three quarters of company revenue in FY2026, and that optical interconnects made up about half of data center revenue while custom silicon made up about one quarter. The growth in this market is real. More importantly, after the report date of May 19, Q1 FY2027 results were disclosed: Marvell's official press release showed record quarterly revenue of USD 2.418 billion, up 28% year over year, with data center revenue of USD 1.8327 billion, or 76% of revenue. Management also raised full-year guidance sharply: according to Futurum's summary of the earnings call, the FY2027 revenue target was lifted to about USD 11.5 billion, or about +40%, and FY2028 to about USD 16.5 billion, or about +45%. In effect, the company itself is saying it can double its scale within two years. The ceiling is far from being reached.

  But it is important to distinguish between expanding an existing pie and creating a new market. AI compute demand, Ethernet/optical interconnect standards, and the trend of custom ASICs replacing general-purpose GPUs were not invented by Marvell. The pie has been enlarged by hyperscaler capital spending and the ecosystem defined by NVIDIA. What Marvell is truly creating is the commercialization of higher-speed optoelectronic solutions, from 800G to 1.6T and even the 3.2T path implied by the Polariton acquisition. This belongs to thickening content value along an existing demand curve, not opening a zero-to-one new category.

  So the conclusion is: the market ceiling is sufficient for a growth stock that could rise fivefold in ten years. This is the fundamental reason the market is willing to give it a near 87 times GAAP P/E multiple (stockanalysis, about USD 252.59 on June 10, market cap about USD 221 billion). But Marvell is a share taker in this large pie, not the creator of the pie. Its upside depends both on the pie itself, meaning whether AI capital spending can keep expanding for many years, and on how much share it can take from Broadcom, the NVIDIA ecosystem, Astera, Credo, and other competitors. The report's caution comes exactly from this point: the direction of demand is favorable, but how much Marvell can capture is much less certain than how large the market is.

  Jun 11, 2026
• QCan its revenue at least double over the next five years? Will growth be driven mainly by volume, price, or new businesses?7/10

  A

  Yes, and it is very likely to more than double. But the growth will come almost entirely from a combination of volume and higher content value per system, including price-like effects, rather than from price increases or a new business that does not yet exist. This is one of Marvell's strongest Baillie Gifford dimensions today. After the report date, management described the future path in quite specific terms, and the five-year doubling threshold looks low for the company.

  Start with the company's own numbers. The report was based on FY2026 revenue of USD 8.195 billion. After the report date, disclosed Q1 FY2027 revenue hit a record USD 2.418 billion, up +28% year over year. More importantly, guidance was raised: according to Futurum's earnings-call summary, FY2027 revenue is expected to be about USD 11.5 billion, or about +40%, and FY2028 about USD 16.5 billion, or about +45%. In other words, from FY2026 to FY2028 alone, revenue goes from about USD 8.2 billion toward about USD 16.5 billion, doubling in two years. The Baillie Gifford hard hurdle of doubling in five years is almost a floor under the company's own roadmap. Of course, guidance is management's view and AI capital spending is cyclical, so it cannot be extrapolated mechanically. Even with a discount, however, the probability of revenue doubling over five years is clearly higher than for most companies.

  Next, break down the growth drivers. This determines the quality of growth:

  • Volume, the main driver: continued buildout of hyperscale AI clusters drives shipment growth in custom XPUs and optical modules. The report notes that FY2026 fourth-quarter data center revenue rose +21% year over year, while Q1 FY2027 data center revenue rose another +27% to USD 1.833 billion. This is typical volume-led growth.
  • Higher content value per system, a healthy form of price: as optical interconnects move from 800G to 1.6T and then to 3.2T, the value per port rises with each generation. Custom silicon also expands from a single XPU to bundles of XPU, XPU-attach, and interconnect, lifting dollar content per customer. According to the earnings-call commentary, custom silicon is expected to grow more than 20% in FY2027 and more than double in FY2028, while DCI optical modules could reach USD 1 billion in annualized revenue in FY2028, twice FY2026. This is not raising prices on the same product. It is selling a higher-end, higher-ASP product mix.
  • New businesses and external expansion, incremental but risky: through the acquisitions of Celestial AI, a photonic fabric platform completed on February 2, 2026 for net cash consideration of USD 1.271 billion, and Polariton, in plasmonic photonics, Marvell is buying future scale-up interconnect and 3.2T capabilities. This is genuine incremental growth, but it is assembled through external M&A, so integration and dilution costs must be charged against it.

  The load-bearing judgment is therefore: revenue is highly likely to double over five years, driven mainly by volume growth plus content-value expansion from product generation upgrades. The contribution from pricing-power-style price increases is small. This fits the report's judgment that business quality is improving. The disagreement is only on price. The market, at about USD 252.59 and a market cap of about USD 221 billion according to stockanalysis, has clearly already priced in this path of doubling or even multiplying several times.

  Jun 11, 2026
• QFive years from now, what will take over as the next growth engine? Does this second curve exist today?5/10

  A

  The second curve does exist today, and it is growing from the same root inside the company. It is mainly silicon photonics/photonic fabric, including co-packaged optics, 3.2T and faster interconnects, plus the spread of custom silicon to more customers and more use cases. But it is highly homologous with the first curve, more an extension of the main curve than an independent new growth pole that could take over if the core business stalls. That distinction has to be made honestly when answering this question.

  Start with today's first curve: custom XPUs, current-generation optical interconnects at 800G/1.6T, and data center switching/networking. Data center already accounted for about three quarters of FY2026 revenue and 76% in Q1 FY2027. So who takes over five years from now? Judging by actions after the report date, the company is betting on three second-curve areas, all backed by real cash investment:

  1. Next-generation silicon photonics and photonic fabric. This is the clearest second curve. Before the report date, the company had completed the acquisition of Celestial AI on February 2, 2026 for net cash consideration of USD 1.271 billion, a photonic fabric platform for next-generation scale-up interconnects. In April, it also acquired Polariton to push optical performance toward 3.2T and beyond. This means that after 1.6T matures, 3.2T, co-packaged optics, and photonic fabric are the successors. They exist today, but are still in the early stage of industrialization.

  2. Custom silicon spreading to new customers and new workloads. According to the earnings-call commentary, custom silicon could more than double in FY2028, driven by both existing and new customer programs. Expanding from serving a few hyperscalers to more customers and more XPU/XPU-attach configurations is itself a second source of growth.

  3. AI entering telecom, AI-RAN, and other new scenarios. Through its NVLink Fusion collaboration with NVIDIA, the company has proposed turning telecom networks into AI infrastructure, including 5G/6G AI-RAN. This is farther out and less certain, currently closer to vision than revenue.

  The key honest judgment: when Baillie Gifford asks about a second curve, the ideal answer is that if the main business is disrupted, the company has another independent growth leg. Marvell's second curves, including photonic fabric, 3.2T, and broader custom silicon, share the same customers, namely hyperscale AI companies, the same demand source, namely AI capital spending, and the same engineering organization as the first curve. The benefit is strong synergy and reusable R&D. The drawback is lack of diversification. If the main pipe of AI data center capex tightens, the first and second curves will both come under pressure. They do not hedge each other. This is the other side of the customer concentration and common demand-source risk repeatedly emphasized in the report. My conclusion is therefore: the second curve is real in technology and investment, and the direction is right, but it is downstream of the same river. It is not the kind of regenerative growth pole in the Baillie Gifford sense that can independently support the company if the core business runs into trouble.

  Jun 11, 2026
• QWhat is its core competitive advantage? Will this moat widen or narrow over the next three to five years?5/10

  A

  The core competitive advantage is deep co-definition of chips with hyperscale customers, system-level engineering delivery across complex SoCs and optoelectronics, and switching costs once Marvell enters customers' multi-year product roadmaps. This is an engineering- and organization-driven moat. Over the next three to five years, I think it will widen slightly, but as it widens it will also become more dependent on external ecosystems, especially NVIDIA. It is a real moat, but not a very wide one, and its purity is being diluted. The report gives the moat 3/5 and calls it an emerging engineering moat. I agree with that characterization.

  Where is the moat? It is not in brand, network effects, or proprietary data. The report is clear: Marvell's real potential barrier is co-defining chips with customers, accumulating years of product development experience, executing at the system level, and creating switching costs after entering a platform. One quantifiable supporting point is the proxy data cited in the report: since 2018, the company's A0-to-production first-pass silicon success rate improved by about 90%, and average tape-out cycle time shortened by about 60%. This kind of delivery certainty is exactly why hyperscale customers are willing to entrust multi-year custom programs to it, although the data has a company-presented flavor. Switching costs are also real: once an XPU or optical interconnect solution enters a customer's rack roadmap, changing suppliers midway requires redesign, revalidation, and mass-production risk. Replacement is not easy. The report judges switching costs as moderately strong.

  Why do I think the moat will widen slightly?

  • Rising complexity works in its favor. From 800G to 1.6T and then 3.2T, and from single chips to bundles of custom silicon, optical interconnects, and switching, fewer suppliers can deliver all of this into mass production. Customer dependence on system-level players such as Marvell is deepening. After the report date, the acquisitions of Celestial AI and Polariton filled in next-generation photonics capabilities, actively reinforcing this moat.
  • NVIDIA's endorsement both widens the moat and sets limits. Before the report date, NVIDIA invested USD 2 billion in Marvell and brought it into the NVLink Fusion ecosystem, corresponding to the USD 2 billion preferred stock issuance in Q1 FY2027 financials. Being selected by the leader as a key partner for scale-up interconnects and custom silicon significantly widens the moat in the short term and improves confidence in design wins.

  Why is this moat not very wide, and why is its purity being diluted?

  • The capability can be replicated, and competitors are getting stronger. The report states clearly that replicating this engineering capability requires years and billions of dollars of R&D, but is not impossible. The most direct full-stack competitor, Broadcom, has stronger cash flow depth, platform breadth, and profit-pool positioning. Its market cap is about USD 1.76 trillion and P/E about 65 times, far larger than Marvell. In interconnect niches, Astera and Credo are also close behind.
  • Customers are both the moat and the biggest threat. The same hyperscale customers can foster second suppliers or even build in-house. The report lists customer concentration, with FY2026 annual report disclosures showing accounts receivable concentration at 73% across 4 customers, as a core risk.
  • Dependence on NVIDIA is a double-edged sword. NVLink Fusion widens Marvell's position inside the NVIDIA system, but it also means part of the moat is rented from someone else's ecosystem. Control is not entirely in Marvell's hands.

  In sum, this is an engineering moat with the right direction, currently widening, but limited purity. It is deeper than a pure connectivity IP player, and shallower than Broadcom's profit-pool-centered capability. The report therefore gives it a neutral-to-positive 3/5 while still insisting that Marvell has not yet proven a moat wide and durable enough to support the current valuation. That tension is one of the fundamental reasons it assigns a Watch rating rather than Buy.

  Jun 11, 2026
• QIf the core business is disrupted, does it have the DNA to reinvent itself? How does it handle mistakes and bad news?5/10

  A

  It has a fairly strong reinvention gene. This is one of the few hard qualifications Marvell has proven through real strategic transformation over the past decade. But its handling of bad news is more institutionalized and positively framed, lacking the founder-style habit of thorough self-criticism. This Baillie Gifford question asks whether the company can reinvent itself if its core business is disrupted, and how it faces mistakes. It is a question about resilience and an honest culture.

  First, the reinvention gene is supported by solid evidence. The report notes that CEO Matt Murphy reviewed in the 2026 proxy how Marvell completed a large-scale strategic remake over ten years: from a more consumer-oriented platform to data infrastructure and AI. Over that decade, it completed more than USD 24 billion of acquisitions and about USD 4.5 billion of divestitures, recently sold its automotive Ethernet business, and continued buying AI interconnect assets. This is not a slogan. A chip company once centered on mature businesses such as hard-drive controllers actively moved its focus to data centers and got data centers to three quarters of revenue, 76% in Q1 FY2027. That is a live case of a core business facing obsolescence by the times, followed by proactive reinvention. After the report date, this gene continued: the acquisitions of Celestial AI, in photonic fabric, and Polariton, in plasmonic photonics, are bets on post-3.2T photonic fabric before 1.6T is even mature. That is exactly the behavior pattern of reinventing before the core business is disrupted. In the Baillie Gifford sense, the answer to whether it can remake itself is positive.

  But the score on how it handles mistakes and bad news should be more cautious:

  • The governance baseline is sound, but institutionalized. The report notes that the proxy emphasizes pay-for-performance, has a clawback policy, prohibits executive pledging and hedging, has no golden-parachute excise tax gross-ups, and uses double-trigger change-in-control provisions. These show a framework for correction and accountability, and the integrity baseline looks acceptable.
  • Bad news is presented through a positive lens. The company's external communications, including proxy statements and press releases, carry a clear success-narrative tone. The report repeatedly reminds readers that this data has a company-presented flavor and that not every management vision should be treated directly as fact. Marvell is more like a large company that smooths and institutionalizes bad news than a founder culture that publicly reviews failures.
  • Accounting noise is a window into its honesty. Q1 FY2027 GAAP net income was only USD 34.5 million, heavily dragged down by expenses such as the USD 331.8 million fair-value change in contingent consideration related to Celestial. FY2026 net margin was also boosted by the USD 1.830 billion pre-tax gain from the automotive Ethernet sale. The company did fully disclose these one-off items in its financial statements, so it did not hide them. But frequent large acquisitions mean bad news such as integration trouble or impairment is more likely to appear in the future. The report correctly lists M&A-related impairment or integration trouble as a trigger for reassessment.

  Conclusion: Marvell's reinvention gene is one of its most tested strengths, and the past decade's transformation was not luck. But its way of facing mistakes is compliant, transparent, and positively narrated in an institutional style, not radically self-reflective. For Baillie Gifford, the former adds points; the latter does not subtract much but does not add points either. This is a company that can evolve and disclose, but is not especially likely to admit mistakes publicly.

  Jun 11, 2026
• QDoes management, especially the founder, have a long-term mindset and deep alignment with the company? Is it willing to sacrifice current profits for five to ten years out?5/10

  A

  Management has a clear long-term mindset and is plainly willing to sacrifice current profits for five to ten years out. But this is a professionally managed company without a founder holding a concentrated stake, so the hard alignment between management and shareholders is not strong. This is a real deduction for Marvell under the Baillie Gifford framework. The report gives management and capital allocation 3/5. I agree.

  First, the evidence for long-term vision and willingness to sacrifice the present is strong. CEO Matt Murphy led a ten-year strategic remake, including more than USD 24 billion of acquisitions and about USD 4.5 billion of divestitures, with a clear direction: focus on data infrastructure. More direct evidence of sacrificing current profit for the future appeared in the financials after the report date: Q1 FY2027 GAAP net income was only USD 34.5 million, R&D expense in one quarter was as high as USD 652.3 million, and the company had just spent USD 1.271 billion of net cash to acquire Celestial AI, plus acquired Polariton to bet on photonic fabric beyond 3.2T. It clearly did not cut R&D or external investment to make short-term GAAP profits look better. That is exactly the behavior Baillie Gifford values: putting firepower behind years 3-10.

  But the deep alignment point must be scored low honestly:

  • No founder and very low management ownership. The report cites the 2026 proxy: CEO Matthew Murphy beneficially owned about 412,871 shares, less than 1% of total shares outstanding. All current directors and executives together owned about 1,046,798 shares, also less than 1%. Compared with the company's roughly 875 million shares, shown on the Q1 FY2027 balance sheet, this is far from the structure Baillie Gifford prefers, where founders or management have their net worth tied to the company and share outside shareholders' upside and downside. Marvell is a mature company that only reorganized from a Cayman holding structure into a Delaware company in 2021. It does not have founder-led DNA.
  • Alignment comes through rules, not personal wealth. The report notes that the company does have stock ownership requirements: the CEO must hold shares equal to 6 times salary, other executives 3 times salary, and all have met the requirement. It also has clawback, no-pledging, and no-hedging policies. These are adequate institutional alignments, but they are not the same as concentrated ownership. The former is a compliance baseline; the latter is true economic resonance.
  • Stock-based compensation is a real shareholder cost, and it is increasing. The report already deducts FY2026 stock-based compensation of USD 591 million as a real shareholder cost. After the report date, Q1 FY2027 stock-based compensation alone reached USD 207.6 million, putting the annualized run rate clearly above FY2026. While management invests for the future, it is also continuing to dilute shareholders. This is another expression of insufficient alignment.

  Conclusion: management passes on thinking long term and being willing to spend for the future. The report also marks management trustworthiness as a pass in its Checklist. But deep alignment does not pass: there is no founder, ownership is below 1%, alignment relies on policies rather than personal wealth alongside shareholders, and stock-based compensation is high. Relative to Baillie Gifford's ideal of founder leadership and deep alignment with the company, Marvell is a trustworthy professional management team, not a captain in the same boat as you.

  Jun 11, 2026
• QIf it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable, and does it avoid harming society or relying on regulatory arbitrage?6/10

  A

  If Marvell disappeared tomorrow, a few hyperscale customers would miss it greatly in the short term on key custom programs. Its indispensability is moderately strong, but local rather than universal. Its growth model is clean and sustainable, and does not depend on harming society or stepping across regulatory red lines. This is one of its rare blemish-free dimensions. This Baillie Gifford question asks two things: indispensability, and whether growth is sustainable without harming society or regulation. They should be answered separately.

  First, how much would customers miss it? Indispensability is deep but not broad:

  • For awarded custom programs, replacement costs are high. The report judges switching costs as moderately strong: once a custom XPU or optical interconnect solution enters a customer's rack roadmap, switching suppliers midway requires redesign, revalidation, and mass-production risk. For the handful of programs in production, Marvell is indeed indispensable in the short term. This is also the moat foundation behind Q1 FY2027 data center revenue growing +27% year over year to USD 1.833 billion and reaching 76% of revenue.
  • But the missing is local and replaceable. The report repeatedly emphasizes that customers can foster second suppliers or even build in-house. The most direct full-stack competitor, Broadcom, can provide alternatives in many links of the chain. In other words, customers would miss Marvell's delivery on a specific program, but not in the way they would miss water, electricity, or gas. Marvell is not the industry's only solution. The other side of the report's FY2026 accounts receivable concentration of 73% across 4 customers is that Marvell depends more on customers than customers depend on Marvell.
  • NVIDIA's choice reinforces short-term indispensability. Before the report date, NVIDIA invested USD 2 billion and brought Marvell into the NVLink Fusion ecosystem, which amounts to a leader's endorsement of its strategic importance. But it also means part of Marvell's indispensability is borrowed from the NVIDIA ecosystem, so control is not entirely its own.

  Now consider whether growth is sustainable and whether it harms society or regulation. This part is very clean:

  • The source of growth is legitimate. Revenue comes from selling faster and more power-efficient chips and optical modules to AI data centers. In essence, it sells better engineering products. It is not monetizing users through exploitation, data abuse, addictive design, or regulatory arbitrage. The report contains no accusation point about social harm.
  • No major regulatory or litigation overhang. The report discloses that Deloitte issued an unqualified opinion, and the critical audit matter is inventory obsolescence/impairment estimation, a demand-forecasting issue, rather than revenue-recognition fraud or compliance risk. This is a company with an acceptable governance baseline and an above-board growth model.
  • The only external constraint is geopolitics/export controls, an industry-wide issue. As a semiconductor company serving global hyperscalers and depending on advanced foundry processes, it faces export controls and supply-chain geopolitical risk like the entire AI silicon chain. But that is a common industry macro constraint, not Marvell making money by harming others.

  Conclusion: customers' missing of Marvell is real but local, deep in individual programs and shallow in overall substitutability. Its indispensability is moderately strong, not top-tier. Its growth model is fully sustainable, does not harm society, and does not cross regulatory red lines. This is a clean positive in the Baillie Gifford framework and matches the report's lack of concern about the legitimacy of the business. The real risk has never been whether Marvell harms society. It is whether customers can live without Marvell while Marvell cannot live without customers.

  Jun 11, 2026
• QWhat are the unit economics of this business, including gross margin and incremental returns? Do they improve or deteriorate with scale? Where does the money it earns go?6/10

  A

  The unit economics are good and improving with scale, with gross margin rising and operating leverage beginning to show. But this is not an effortless money-printing machine. It is a high-quality hard-tech business that requires constant R&D reinvestment and consumes working capital as it grows. The money it earns mainly goes to R&D, M&A expansion, and buybacks. This question requires separating three issues: gross-margin quality, incremental returns, and the use of cash.

  Start with gross margin and scale. The direction is improving:

  • Gross margin is recovering as the AI mix upgrades. The report's financial table shows gross margin recovering from about 41% in FY2024/FY2025 to 51.0% in FY2026. After the report date, Q1 FY2027 GAAP gross margin was 52.1% and non-GAAP gross margin 58.9%, while Q2 guidance put GAAP gross margin even higher at 52.1%-53.1%. Larger scale plus a product mix shifting toward higher-end custom silicon and optical interconnects is improving gross margin rather than eroding it.
  • Operating leverage is starting to materialize. The report shows operating margin turning positive from -12.5% in FY2025 to 16.1% in FY2026. Q1 FY2027 operating income was USD 339.4 million on revenue of USD 2.418 billion, for a GAAP operating margin of about 14%. As revenue climbs toward about USD 16.5 billion in FY2028 according to earnings-call commentary, fixed R&D absorption should keep lifting operating margin. This is a positive signal that larger scale improves unit economics.

  But incremental returns deserve a discount because of three drags:

  • R&D intensity is high and rigid. The report notes FY2026 R&D expense of USD 2.075 billion, close to one quarter of revenue. Q1 FY2027 R&D alone was USD 652.3 million. To stay ahead, this business must keep reinvesting heavily. It is not like a software subscription where one investment can be harvested for many years.
  • Growth consumes working capital. The report specifically warns that FY2026 accounts receivable jumped from USD 1.028 billion to USD 2.187 billion, and inventory rose to USD 1.388 billion. The idea that faster growth automatically brings easier cash collection does not hold. This lowers incremental cash conversion.
  • Capital returns are not outstanding. The report roughly calculates GAAP ROIC in only the high single digits, while FY2026 GAAP ROE was heavily beautified by the asset-sale gain. The Checklist therefore marks excellent return on capital as a fail. Incremental returns are good, but not excellent.

  Finally, where does the money go? Three places:

  1. R&D, the largest and highest priority: maintaining engineering leadership, as above.
  2. M&A expansion: around the report date, Marvell spent USD 1.271 billion of net cash to acquire Celestial AI, and also acquired Polariton, buying next-generation photonics capability. Over ten years, cumulative acquisitions exceeded USD 24 billion. This is strategic spending, but it also raises goodwill, which reached USD 13.884 billion at the end of Q1 FY2027, and integration risk.
  3. Buybacks and dividends: the report says FY2026 buybacks were about USD 2.040 billion, reducing shares to 847 million. But after the report date, Q1 FY2027 share count rose due to Celestial share issuance and NVIDIA-related capital actions, quarterly buybacks were only USD 200 million, and the dividend was USD 0.06 per share. In this phase, dilution for expansion is outweighing share reduction.

  Conclusion: unit economics are improving with scale, with rising gross margin and visible operating leverage. That is a real strength. But high R&D intensity, working-capital use, and only high-single-digit capital returns make this a high-quality but not exceptional hard-tech unit-economic profile, not a software-like high-incremental-return machine. Cash mainly flows to R&D and M&A expansion, while buybacks take a back seat during the expansion phase. This is consistent with the report's judgment that business quality is improving, but Marvell is not yet a high-return cash cow.

  Jun 11, 2026
• QWhat conditions must all hold for it to rise fivefold in ten years? Are those conditions realistic? What expectations are embedded in today's share price?3/10

  A

  For today's Marvell to rise fivefold in ten years, several difficult conditions must all hold at the same time. Each condition is not absurd on its own, but getting every one of them right, starting from today's roughly USD 252.59 share price, is clearly harder than it was at the roughly USD 169 starting point when the report was written. The current share price already embeds a fairly optimistic script of about USD 16.5 billion of FY2028 revenue, sustained long-term high growth, and continuing margin expansion. The essence of this Baillie Gifford question is exactly the multiplication of conditions and what today's price implies.

  First calibrate where today's expectations already sit. When the report was written, MRVL was about USD 168.93, with a market cap of about USD 145.9 billion and a trailing P/E of about 59.3 times. On that basis, the report judged it to be at a significant premium to the relatively reasonable range of USD 60-90 and near the upper end of the optimistic range of USD 130-175. But after the report date, the share price rose further: according to stockanalysis, it closed around USD 252.59 on June 10, with a market cap of about USD 221 billion, a trailing GAAP P/E of about 87 times and forward P/E of about 55.7 times, and a 52-week range of USD 61.44-324.20. In other words, the current price is already about 40% above the upper end of the report's optimistic intrinsic value range of USD 130-175. The market is not merely pricing in a doubling. It has quite fully pulled forward the raised guidance of about USD 11.5 billion in FY2027 and about USD 16.5 billion in FY2028, or +40%/+45%, plus more distant sustained high growth and margin expansion.

  A fivefold return over ten years, from about USD 221 billion to about USD 1.1 trillion in market cap, requires all of the following conditions:

  1. AI data center capital spending keeps expanding for another ten years without peaking: hyperscalers keep building AI clusters at a steep rate, and cyclical swings do not interrupt Marvell's volume growth. This is the largest external uncertainty.
  2. Content value in custom silicon and optical interconnects keeps thickening: the move from 1.6T to 3.2T and the momentum for custom silicon to more than double in FY2028 must extend into years 3-10, not just be a two- or three-year benefit.
  3. Marvell keeps taking share from Broadcom, the NVIDIA ecosystem, Astera, and Credo: the report judges Broadcom still materially stronger in cash flow depth and profit-pool positioning. Marvell must prove it can become another Broadcom, rather than remain an unproven Marvell.
  4. Margins and cash conversion truly scale: the report repeatedly reminds readers that GAAP profit contains a lot of accounting noise, with FY2026 boosted by a USD 1.830 billion sale gain and Q1 FY2027 GAAP net income of only USD 34.5 million dragged down by USD 331.8 million of contingent consideration expense. Owner Earnings must keep moving from the report's roughly USD 800 million estimate to more than USD 2 billion to count as delivered.
  5. Customers become more diversified, and stock-based compensation and M&A do not keep eroding per-share value: Marvell must reduce 73% concentration across 4 customers and control dilution from stock-based compensation, which already reached USD 207.6 million in a single quarter.
  6. The valuation does not compress sharply: this is the most delicate condition. Even if the first five conditions are met, if the market rerates Marvell from a near 87 times GAAP multiple back to ordinary semiconductor multiples, market cap growth will be heavily offset by multiple compression.

  Are these conditions realistic? Each one is possible individually, but having all six hold simultaneously and continuously for ten years is a low-probability joint event. This is the report's core point: at the current price, investors are not buying already cheap cash flows; they are buying a growth story that must play out wonderfully over the next decade. At about USD 169, the report already judged the FCF yield to be only about 0.96%, far below the roughly 4.59% ten-year Treasury yield, and the margin of safety insufficient. The share price has risen another about 50%, so the margin of safety is only thinner. Conclusion: a fivefold gain in ten years is not impossible, but today's share price already prices doubling, sustained high growth, and margin expansion as the base case. Investors are not being offered a discount, but a high hurdle where every optimistic condition must be delivered. That is fully consistent with the report's judgment of a good company at a bad price, and the current price makes that judgment even more valid.

  Jun 11, 2026
• QWhy has the market not recognized all this yet? Is it because the market does not understand, looks down on it, or cannot look far enough? What will become the narrative inflection point?3/10

  A

  For Marvell, the classic Baillie Gifford question of why the market has not recognized this must actually be inverted. The market has recognized it, and may well have seen too much: the share price rose from about USD 169 when the report was written to about USD 252.59 on June 10, with a market cap of about USD 221 billion according to stockanalysis, and a near 87 times GAAP P/E. The truly scarce insight here is not that a cheap stock is being ignored, but whether the market is overestimating the speed and certainty of delivery. The narrative inflection point can move upward, confirming Marvell as the next Broadcom, or downward, disproving the quality of growth and compressing the multiple. This reversal has to be stated honestly when answering the question.

  Answering the three possibilities one by one:

  • Does the market not understand? Basically no longer true. After the report date, NVIDIA invested USD 2 billion and brought Marvell into the NVLink Fusion ecosystem, Q1 FY2027 revenue hit a record USD 2.418 billion, up +28%, and full-year guidance was raised sharply. Marvell has become one of the most closely watched names in AI interconnects. Sell-side coverage is dense and institutional ownership is broad. The valuation discount from not understanding has already been absorbed.
  • Does the market look down on it? The opposite: the market gives it a premium, not a discount. Under the report's relative valuation framework, even though Broadcom, with a market cap of about USD 1.76 trillion and about 65 times PE, has stronger cash flow depth and profit-pool positioning, Marvell's P/FCF is even higher. Credo and Astera also trade at high multiples. The market is looking highly on the whole AI connectivity track, not looking down on Marvell.
  • Can the market not look far enough? This is the only real cognitive disagreement. The market is looking very far: it has already priced in about USD 16.5 billion of FY2028 revenue, and even more distant 3.2T, photonic fabric, and a doubling of custom silicon. The report's counterview is that the market is treating the revenue and design-win story as if it were already thicker per-share cash flow, but there is still a gap of delivery quality, time, and valuation-compression risk between the two. In other words, the disagreement is not whether the market sees growth, but whether it sees the speed and certainty of that growth too optimistically.

  So what would the narrative inflection point be? Both directions matter:

  • Upside inflection point, reinforcing the bull narrative: Owner Earnings sustainably exceed USD 2 billion and are not driven by one-off items, ROIC moves into stable double digits, and customers/projects become clearly more diversified, no longer dependent on a few platforms. The report explicitly lists these as conditions that could justify upgrading the rating from Watch. If more hyperscalers award custom programs to Marvell and incremental orders from NVLink Fusion materialize, the market will further confirm the narrative that Marvell is the next Broadcom.
  • Downside inflection point, breaking the narrative: the report is very clear: data center revenue growth slows for consecutive periods, customer concentration worsens further, gross margin falls back to the low-to-mid 40% range for a long time, inventory impairments recur, or acquisitions such as Celestial/Polariton lead to impairment or integration trouble. Any of these could bring the market down from a near 87 times multiple to an ordinary semiconductor multiple. The report estimates that even if the business does not collapse, this multiple-compression scenario could create long-term capital losses of 50%-70%. A near-term validation window is how subsequent quarterly earnings deliver against the raised guidance.

  Conclusion: this stock is unusual because the Baillie Gifford discount opportunity of a market that cannot look far enough has inverted into the premium risk of a market that may be looking too fully. The real insight gap is not that nobody has discovered a cheap stock. It is whether high expectations can be caught by high-quality, sustainable cash flow. The upside narrative inflection depends on delivery quality exceeding expectations. The downside depends on growth quality being disproven and valuation reverting. The current price, already about 40% above the upper end of the report's optimistic range, means the risk balance is exactly as the report says: more tilted toward waiting for a better price than chasing higher.

  Jun 11, 2026