Marvell: The AI Infrastructure Boom Is Getting Bigger – But Can Earnings Keep Up?
The AI trade is evolving, and one company increasingly positioned to benefit is $MRVL(Marvell Technology)$.
While Nvidia dominates the AI accelerator market, Marvell is building a different kind of opportunity: the custom chips, high-speed networking and optical connectivity that allow enormous AI data centres to function.
And the latest developments suggest this opportunity could be considerably bigger than investors initially anticipated.
1. Wall Street’s earnings expectations
Marvell reports Q3 FY2027 results on 1 December 2026, and expectations are already elevated.
Current analyst consensus points to:
* Revenue: $3.15 billion
* Adjusted EPS: $1.10
* Year-on-year revenue growth: approximately 52%
* Year-on-year EPS growth: approximately 45%
For comparison, last year’s corresponding quarter delivered $2.07 billion in revenue and $0.76 adjusted EPS.
These are substantial growth expectations. The challenge is that simply beating estimates might not be enough. Investors will want evidence that growth is accelerating beyond current forecasts.
2. The Google opportunity could be enormous
One of the biggest developments is Marvell’s expanded collaboration with Alphabet’s Google on custom AI chips.
The agreement could generate up to $120 billion in sales through fiscal 2033, subject to commercial milestones.
Google is also eligible to acquire approximately $12.2 billion worth of Marvell shares through a performance-linked warrant arrangement.
This is important because hyperscalers increasingly want their own specialised AI processors rather than relying exclusively on expensive general-purpose GPUs.
Custom silicon could become a major earnings driver as AI inference workloads expand.
However, management has indicated that the larger financial contribution from the Google agreement is expected from FY2029 onwards, rather than immediately.
3. The Lumentum acquisition changes the investment story
Another major development is Marvell’s agreement to acquire Lumentum in an all-stock transaction announced in November.
The deal has an implied enterprise value of approximately $115 billion.
Why does this matter?
AI data centres increasingly need faster connections between processors, servers and entire computing racks. Traditional electrical connections face limitations involving bandwidth, power consumption and heat.
Lumentum brings optical technology that could complement Marvell’s existing networking and custom silicon businesses.
The combined company aims to create a much broader AI connectivity platform, with management identifying a potential $550 billion total addressable market.
The companies also expect more than $1 billion in annual cost synergies.
The opportunity is significant, but the acquisition introduces integration risks, share dilution and execution challenges.
4. Management has raised its long-term ambitions
Perhaps the most interesting development is Marvell’s October investor update.
Management raised its FY2028 revenue outlook to approximately $20 billion, exceeding the previous $18 billion target.
It also increased its FY2029 custom-chip revenue target to approximately $12 billion, up from more than $10 billion previously.
Looking further ahead, management outlined a potential FY2031 revenue range of $70–90 billion.
These figures demonstrate just how aggressively management sees the AI infrastructure opportunity expanding.
Of course, these are forward-looking targets, not guaranteed earnings.
5. What could surprise investors positively or negatively?
Bullish catalysts:
* Stronger-than-expected demand for custom AI accelerators.
* Google and other hyperscalers increasing their chip commitments.
* Faster adoption of optical connectivity in AI data centres.
* Higher revenue contributions from Microsoft and Amazon-related programmes.
* Improving operating leverage as revenue scales.
Risks:
* AI infrastructure spending slowing or being delayed.
* Competition from Broadcom and in-house chip development.
* Customer concentration among a handful of hyperscalers.
* Lumentum integration costs and acquisition-related dilution.
* Expectations becoming so high that even strong earnings disappoint.
My investment view
Marvell is an interesting way to participate in the next phase of AI infrastructure spending without investing exclusively in GPU manufacturers.
The combination of custom silicon, networking and optical connectivity gives the business exposure to several important areas of AI infrastructure.
However, there is an important distinction between having an enormous addressable market and actually converting that opportunity into sustainable earnings and free cash flow.
For the December earnings report, I’ll be watching three things particularly closely: custom-chip revenue growth, forward guidance and management’s commentary on Google-related demand.
If execution matches the ambitious long-term targets, Marvell could have a substantial growth runway. If expectations run ahead of actual revenue conversion, volatility could be considerable.
The big question: Is Marvell becoming one of the most important suppliers behind the AI revolution, or is Wall Street already pricing in too much of its future success?
@Tiger_comments @TigerStars @TigerClub
Comments