To The Moon

$Marvell Technology(MRVL)$ 

#Marvell Surges 33% After Jensen Huang’s Trillion-Dollar Call — Revolutionary AI Winner or a Dangerous Chase?

#MRVL #NVDA #AVGO #AI #Semiconductors #CustomSilicon #ASIC #XPU #AIInfrastructure #DataCenters #TigerTrade

Marvell Technology’s explosive rally is understandable, but I would separate two very different questions:

1️⃣ Can Marvell become one of the most important semiconductor companies of the AI era?

2️⃣ Is $MRVL attractive immediately after a 32.52% one-day surge?

My answer is yes to the first question, but probably not at any price to the second.

NVIDIA CEO Jensen Huang calling Marvell the “next trillion-dollar company” is an extraordinarily powerful endorsement. It is especially meaningful because NVIDIA has also invested $2B in Marvell and is integrating Marvell’s custom XPUs and networking technology into the NVLink Fusion ecosystem.

However, Jensen’s statement should not be treated like an independent analyst price target.

@NVIDIA now has a direct financial and strategic interest in Marvell succeeding. Huang was not simply praising a random semiconductor company from the sidelines. He was promoting an important partner in which NVIDIA has invested substantial capital.

That does not make the thesis wrong. It simply means investors should analyse the underlying economics rather than buying solely because Jensen said the words “trillion-dollar company.”

Why Marvell’s AI Opportunity Is Real

The AI infrastructure market is evolving beyond a world where every workload runs on a general-purpose NVIDIA GPU.

Hyperscalers are increasingly designing custom processors optimised for specific workloads:

• Training large AI models

• Running inference at scale

• Recommendation systems

• Search and advertising

• Video processing

• Cloud networking

• Memory movement

• Internal data-centre workloads

These processors are often called ASICs, XPUs or custom accelerators.

The motivation is straightforward. NVIDIA GPUs remain the industry standard, but they are expensive, power-intensive and designed to serve many different workloads. A hyperscaler processing billions of similar AI requests may be able to reduce power consumption and cost per token by using a custom chip optimised for that specific workload.

The hyperscaler owns the workload knowledge and software requirements, but designing a leading-edge chip requires expertise in areas such as:

• Advanced semiconductor architecture

• High-speed SerDes

• Memory interfaces

• Packaging

• Optical connectivity

• Ethernet switching

• Power efficiency

• Verification

• Manufacturing coordination

• Integration with large-scale data-centre systems

That is where Marvell enters the picture.

Marvell helps customers turn an internal chip concept into a manufacturable product. It also supplies many of the technologies required to connect those processors across enormous AI clusters.

This is why Marvell is not merely a speculative “AI chip” story.

It potentially earns revenue from several layers of the AI factory:

CUSTOM COMPUTE

↓

XPUs and ASICs

↓

HIGH-SPEED INTERCONNECT

↓

OPTICAL AND COPPER CONNECTIVITY

↓

ETHERNET SWITCHING

↓

DATA-CENTRE INTERCONNECTION

The more AI accelerators hyperscalers deploy, the more difficult it becomes to move data efficiently between processors, memory, racks and separate data centres.

Compute receives most of the attention, but connectivity increasingly determines how much of that compute can actually be used.

A powerful accelerator that spends time waiting for data is an extremely expensive underutilised asset.

That makes Marvell’s expertise in custom silicon, optical DSPs, switching, interconnect and silicon photonics strategically valuable.

NVIDIA’s Investment Changes the Story

The NVIDIA–Marvell partnership is important because custom silicon was previously presented largely as a threat to NVIDIA.

The usual argument was:

Hyperscalers will replace expensive NVIDIA GPUs with internally designed ASICs.

NVIDIA’s response is more sophisticated.

Rather than fighting every custom chip, NVIDIA is expanding its ecosystem so selected XPUs can connect to NVIDIA CPUs, networking products, DPUs, GPUs and NVLink infrastructure.

Through NVLink Fusion, Marvell can help customers build custom accelerators that remain compatible with NVIDIA’s broader AI platform.

This creates a hybrid model:

NVIDIA GPU COMPUTE

+

CUSTOM MARVELL-ENABLED XPUs

+

NVIDIA/MARVELL NETWORKING

+

HIGH-SPEED OPTICAL CONNECTIVITY

NVIDIA remains inside the rack even when the hyperscaler introduces proprietary silicon.

Marvell benefits because customers gain access to NVIDIA’s architecture while retaining the ability to customise part of their compute infrastructure.

The partnership therefore transforms Marvell from a potential NVIDIA substitute into an important NVIDIA ecosystem partner.

That is strategically significant.

The Financial Momentum Is Strong

Marvell’s latest operating performance supports the bullish narrative.

The company reported Q1 fiscal 2027 revenue of approximately $2.42B, representing 28% YoY growth. Management guided the following quarter to approximately $2.70B, which would represent around 35% YoY growth.

Management also expects revenue growth to accelerate through fiscal 2027, supported by strong demand across:

• Custom XPUs

• XPU-attach solutions

• 800G and 1.6T optical products

• Ethernet switches

• Data-centre interconnect

• Scale-up networking

• Silicon photonics

Marvell’s full fiscal 2026 revenue reached approximately $8.20B, growing 42% YoY, while non-GAAP EPS increased to $2.84.

Most importantly, management now expects custom-silicon revenue alone to exceed $10B in fiscal 2029.

That means Marvell’s custom-chip business could become larger than the company’s entire fiscal 2026 revenue base.

The growth opportunity is unquestionably substantial.

Marvell Is More Than a Custom-Chip Designer

The market is currently focused on custom XPUs, but I believe Marvell’s connectivity portfolio may ultimately be just as important.

AI clusters are moving from thousands to tens of thousands and potentially hundreds of thousands of accelerators. That creates severe bottlenecks involving:

• Bandwidth

• Latency

• Power consumption

• Memory access

• Rack-to-rack connectivity

• Data-centre-to-data-centre communication

Marvell recently introduced its Teralynx T100, which the company describes as the first 102.4 Tbps switch specifically designed for AI and cloud infrastructure.

Marvell claims the product can deliver up to 25% lower power consumption than competing solutions while supporting a 512-port scale-out architecture.

Whether those exact competitive advantages are maintained in production remains to be proven, but the product highlights Marvell’s larger opportunity.

AI infrastructure spending is not only about purchasing GPUs.

For every dollar spent on accelerated compute, substantial additional investment is required in networking, optics, memory, storage, power and cooling.

Marvell is positioned across several of those categories.

The Trillion-Dollar Mathematics

This is where the excitement must meet reality.

At the latest regular-session close, Marvell traded around $263.47, representing a market capitalisation of approximately $235B.

To reach a $1T valuation, Marvell must increase by approximately:

$1,000B ÷ $235B = 4.25x

That would require the market value to compound at roughly:

📈 61.9% annually for three years

📈 33.6% annually for five years

📈 23.0% annually for seven years

📈 15.6% annually for ten years

A trillion-dollar Marvell is mathematically possible, particularly over a decade, but it would require far more than one successful custom-silicon programme.

It would require Marvell to become one of the dominant infrastructure suppliers across custom compute, optical connectivity, switching and AI networking.

The Valuation Is Already Pricing in Enormous Success

Marvell’s current market capitalisation is approximately 29x fiscal 2026 revenue and around 24x its latest quarterly annualised revenue run rate.

Its trailing P/E ratio is also roughly 90x.

These are extremely demanding valuation levels for a semiconductor company, even one growing rapidly.

The market is no longer valuing Marvell based primarily on its existing earnings. Investors are pricing the company according to what they believe it could earn several years from now.

To illustrate the challenge, suppose Marvell eventually receives a mature valuation of:

10x revenue → $100B revenue required for a $1T valuation

15x revenue → $66.7B revenue required

20x revenue → $50B revenue required

25x revenue → $40B revenue required

Maintaining the current near-29x sales multiple indefinitely would require approximately $35B of revenue to support a $1T market capitalisation.

However, valuation multiples normally compress as companies become larger and growth slows.

A more realistic trillion-dollar path may therefore require Marvell to generate somewhere between $50B and $100B of annual revenue, depending on its margins, free cash flow, competitive position and future market conditions.

Management’s target of more than $10B in custom-silicon revenue by fiscal 2029 is impressive, but it remains a long way from the revenue and profit base usually required to sustain a trillion-dollar valuation.

The Broadcom Comparison Matters

Marvell’s most important comparison is not NVIDIA. It is Broadcom.

@Broadcom has already established itself as a dominant supplier of custom AI accelerators and networking infrastructure.

Broadcom recently reported:

• Q1 revenue of approximately $19.3B

• AI revenue of approximately $8.4B

• AI revenue growth of 106% YoY

• Around $8B of quarterly free cash flow

• An adjusted EBITDA margin of approximately 68%

Broadcom’s quarterly AI revenue is already roughly equal to Marvell’s entire fiscal 2026 company revenue.

That does not mean Marvell cannot become a major winner. It means Marvell still has considerable execution ahead before it reaches Broadcom’s scale, profitability and cash-generation capacity.

Broadcom has:

✅ Larger custom-silicon programmes

✅ Deeper hyperscaler relationships

✅ Greater purchasing power

✅ Massive free cash flow

✅ Infrastructure software diversification

✅ Proven execution across multiple chip generations

Marvell may offer greater percentage upside because it is smaller, but it also carries greater execution and valuation risk.

In simplified terms:

$AVGO = Proven custom-silicon leader

$MRVL = Higher-growth challenger with greater upside torque

$NVDA = Dominant full-stack AI platform

The Risks Investors Cannot Ignore

1. Customer concentration

Marvell’s ten largest customers represented approximately 82% of fiscal 2026 revenue.

Three customers represented around 75% of gross accounts receivable at the end of the latest quarter.

This means one delayed product ramp, lost design win or reduced hyperscaler spending programme could materially affect growth.

Custom-chip projects are large, but they are also concentrated.

2. Design wins do not become revenue immediately

Custom silicon typically involves long development cycles.

A company may announce a design win years before the product reaches volume manufacturing. Delays in software, validation, packaging, foundry capacity or customer deployment can push revenue into later periods.

The market may be capitalising future programmes long before the cash actually arrives.

3. Hyperscalers have significant bargaining power

Marvell’s customers are among the largest and most technologically sophisticated companies in the world.

They can negotiate aggressively, develop more capabilities internally, use multiple suppliers or shift projects between semiconductor partners.

Marvell must continually prove that its technology and execution justify its role.

4. Custom silicon may carry different economics

Custom ASIC programmes can generate large and relatively predictable revenue once production begins, but the customer owns much of the product architecture and purchases enormous volume.

That can limit pricing power compared with highly differentiated merchant processors sold broadly across the market.

Revenue growth does not automatically produce NVIDIA-like margins.

5. Stock-based compensation and acquisition complexity

Marvell recorded approximately $208M of stock-based compensation in its latest quarter, up significantly YoY.

GAAP EPS was only $0.04, compared with non-GAAP EPS of $0.80, partly because of acquisition-related accounting, amortisation, stock compensation and changes in contingent consideration.

Non-GAAP earnings are useful for understanding operations, but investors should not ignore the substantial gap between adjusted and reported results.

6. Balance-sheet commitments

Marvell held approximately $3.84B in cash against roughly $4.96B in long-term debt at the end of the quarter.

It has also made significant acquisitions and entered agreements involving substantial future wafer and substrate capacity deposits.

These investments may strengthen Marvell’s long-term position, but they increase execution risk.

7. AI capital expenditure could slow

Marvell’s growth depends heavily on continued hyperscaler investment.

If AI infrastructure spending slows, shifts toward fewer suppliers, or becomes more focused on utilisation rather than new construction, Marvell’s growth expectations could be revised sharply.

A company trading near 29x trailing sales does not need a collapse in revenue to fall significantly.

It merely needs growth to come in slightly below extremely optimistic expectations.

The Technical Picture Is Already Warning Against Chasing

Marvell reached an intraday high around $310.32 before closing the latest session near $263.47, a daily decline of approximately 16.7%, with around 94M shares traded.

That price action matters.

The stock produced a huge breakout, attracted intense momentum buying, and then experienced equally aggressive profit-taking.

This is not normal low-volatility price discovery.

It is a parabolic market trying to determine how much the Jensen endorsement, NVIDIA investment and custom-silicon opportunity are actually worth.

Marvell’s upcoming addition to the S&P 500 before trading begins on June 22 provides another near-term catalyst. Index funds will need to purchase the stock, creating mechanical demand.

However, S&P inclusion can also become a sell-the-news event after passive buying is completed.

The combination of:

• A 33% one-day surge

• A record high

• Massive trading volume

• A sharp subsequent reversal

• S&P 500 inclusion demand

• Extreme social-media excitement

creates opportunity, but it also creates substantial short-term risk.

My Scenario Analysis

🟢 Bull Case — 30%

Marvell becomes a leading supplier of custom AI accelerators, optical interconnects and Ethernet switching.

Custom-silicon revenue exceeds management’s fiscal 2029 target, NVIDIA integration drives additional design wins, and Celestial AI strengthens Marvell’s silicon-photonics position.

Revenue compounds rapidly, margins remain near or above current levels, and Marvell approaches a $1T valuation within seven to ten years.

🟡 Base Case — 50%

Marvell becomes a major AI-infrastructure winner, but Broadcom retains leadership in the largest custom-silicon programmes.

Revenue and earnings grow strongly, yet valuation multiples gradually compress as Marvell becomes larger.

The company creates substantial long-term value, but the stock experiences repeated 25–40% corrections because expectations remain ahead of reported fundamentals.

A trillion-dollar valuation remains possible, but likely takes considerably longer than the current excitement implies.

🔴 Bear Case — 20%

One or more major custom programmes are delayed, hyperscaler capital expenditure slows, competitive pressure increases, or networking and optical revenue fails to meet expectations.

Marvell continues growing, but the market reduces its valuation from nearly 29x sales toward a more conventional semiconductor multiple.

The company remains fundamentally valuable while the stock experiences a severe drawdown.

This is the biggest danger in chasing: the business does not need to fail for the investment to lose money.

Would I Chase Marvell Here?

Personally, no.

I am bullish on Marvell’s long-term strategic position, but I would not chase a stock immediately after a vertical 33% move driven partly by a public endorsement.

The market has recognised the story. The question is now whether future earnings can catch up with the valuation.

For someone already holding $MRVL, I would consider keeping a core position while trimming a portion into extreme strength.

For a new investor, I would prefer one of three entry conditions:

1️⃣ A meaningful pullback that removes some speculative excess

2️⃣ A multi-week consolidation that allows moving averages and earnings expectations to catch up

3️⃣ Another earnings report confirming that custom-silicon, optics and switching growth are converting into revenue, margins and cash flow

Starting a very small position may make sense for an investor worried about missing the move entirely, but I would preserve most capital for volatility rather than buying a full position after a parabolic rally.

My Investment Ranking

For risk-adjusted exposure to this theme:

🥇 $NVDA — Best ecosystem, strongest competitive moat and broadest AI exposure

🥈 $AVGO — Most established custom-silicon and networking economics with exceptional free cash flow

🥉 $MRVL — Highest potential upside torque, but also the greatest valuation and execution risk

That does not mean Marvell will underperform.

It means Marvell currently requires the most aggressive assumptions.

Final Verdict

I believe Jensen Huang may ultimately be directionally correct.

Marvell has a credible path to becoming one of the most important companies in AI infrastructure. Its custom-silicon expertise, optical portfolio, Ethernet switching, silicon photonics and NVIDIA partnership place it directly inside the most capital-intensive technology buildout in history.

But “could eventually become a trillion-dollar company” is not the same as “must be purchased immediately after a 33% surge.”

At approximately $235B, Marvell still needs to increase more than fourfold to reach $1T. Achieving that outcome requires years of flawless execution, enormous revenue growth, expanding AI infrastructure investment and continued technological leadership.

I am bullish on the company.

I am cautious on the entry.

The strongest investment opportunities usually appear when a great long-term story temporarily loses momentum, not when every investor suddenly agrees that it is destined to become the next trillion-dollar company.

My strategy would be:

LONG-TERM BUSINESS: BULLISH

CURRENT VALUATION: EXTREMELY DEMANDING

SHORT-TERM PRICE ACTION: SPECULATIVE AND VOLATILE

EXISTING HOLDERS: HOLD CORE, CONSIDER TRIMMING EUPHORIA

NEW BUYERS: WAIT FOR A PULLBACK OR CONSOLIDATION

OVERALL VERDICT: WATCH CLOSELY, BUT DO NOT BLINDLY CHASE

@MarvellTechnology $MRVL

@NVIDIA $NVDA

@Broadcom $AVGO

@Alphabet $GOOGL

@Amazon $AMZN

@Microsoft $MSFT

@TSMC $TSM

#Marvell #NVIDIA #Broadcom #ArtificialIntelligence #AIStocks #CustomChips #CustomASIC #XPU #SemiconductorStocks #ChipStocks #DataCenter #CloudComputing #AIInfrastructure #Networking #SiliconPhotonics #OpticalNetworking #GrowthStocks #TechnologyStocks #StockMarket #MarketAnalysis #LongTermInvesting #EarningsGrowth #SNP500 #TigerTrade #Investi