Marvell's 2025 Underperformance and the 2026 AI Storage Catalyst

The core investment question for Marvell in 2025 is stark: why did a company with a booming AI business see its stock crushed? The answer lies in a brutal portfolio split. While the broader semiconductor market soared, Marvell finished the year down approximately 22% as the PHLX Semiconductor Index (SOX) surged over 30%. This wasn't a sector-wide slump; it was a company-specific collapse in its legacy segments.

The "AI Paradox" defines the year. On one side, the data center division grew by a staggering 88% year-over-year. On the other, the company's traditional anchors imploded. Carrier infrastructure revenue plunged 73% as telecoms cut 5G spending. Enterprise networking, once a stable pillar, saw nearly 40% of its revenue vanish as corporate budgets fled to AI servers. The result was a stock price that ignored the high-flying AI segment and focused solely on the collapsing legacy business.

This performance gap was the direct result of a severe, multi-quarter inventory correction. The market's focus was on the immediate revenue holes left by the carrier and enterprise slump, overshadowing every "beat and raise" from the data center team. The strategic pivot to AI, symbolized by the $2.5 billion sale of its automotive ethernet business to Infineon in August, was a necessary step to streamline the portfolio. Yet the market's tepid reaction showed it remained fixated on the near-term pain of clearing that inventory mountain, not the long-term promise of custom silicon.

The bottom line is that 2025 was a year of contradictions. Marvell successfully built a high-growth AI business, but the stock paid for the painful transition out of its cyclical past. The performance gap quantifies the market's verdict: until the legacy drag fully clears, the AI story will struggle to drive the stock higher.

The Q3 Catalyst: Record Revenue Meets Weak Guidance

The event that defined Marvell's recent volatility was a stark disconnect between a stellar quarterly report and a disappointing outlook. In its fiscal third quarter, the company delivered a record $2.075 billion in net revenue, a 37% year-over-year jump that beat its own guidance. This strength was driven by robust demand for its data center products, a clear sign of its momentum in the AI infrastructure race. Yet the market's reaction was brutal, with shares plunging 18% after the report.

The catalyst for the sell-off was the weak guidance for the current quarter. Management explicitly forecasted "overall data center revenue in Q3 to be flat sequentially", a significant slowdown from the previous quarter's growth. CEO Matt Murphy attributed this to "nonlinear growth" and "lumpiness" in demand from cloud providers, a pattern he called normal as hyperscalers build out infrastructure. For investors, however, this lack of clarity on near-term visibility was a major red flag. The guidance implied that the strong AI demand fueling the record revenue might not be a smooth, predictable pipeline but a series of large, unpredictable orders.

This moment highlighted the market's tactical focus on near-term execution over long-term potential. The stock's plunge showed that even a record quarter could be overshadowed by uncertainty about the immediate future. The guidance failed to provide the "bottoms-up granularity" analysts were seeking to validate Marvell's ambitious 20% data center market share target. The result was a sharp reset in expectations, turning a story of acceleration into one of temporary chop.

The Amazon Contract Overhang: A High-Stakes Uncertainty

The catalyst creating material volatility for Marvell is a stark clash between a bearish analyst thesis and the company's own bullish guidance. Benchmark analyst Cody Acree delivered the blow, downgrading the stock to 'Hold' on Monday with a "high degree of conviction that the company has lost both Amazon's Trainium 3 and 4 designs" to Taiwan's Alchip. This is a direct challenge to the narrative that Marvell is the undisputed partner for Amazon's next-generation AI chips, a relationship that has been central to its recent stock surge.

The market's reaction was immediate and severe, with shares falling 7% on the news. This drop shows the overhang is material, as investors are pricing in the risk of losing a key future revenue stream. Acree argues the company's own projections imply the expectation of lost business, suggesting the market has misread Marvell's recent signals as a sign of Amazon stability.

Yet this downgrade sits in direct contradiction with Marvell's own multi-year outlook. Management just gave a powerful forecast, projecting 25% Data Center growth in the next fiscal year and an acceleration to 40% growth the following year. This guidance, which analysts like JP Morgan's Harlan Sur have called a reason to raise price targets, implies continued strength from Amazon and other hyperscalers. Acree contends this growth will come from older Trainium 2 volumes and other engagements, not the coveted new Trainium 3 and 4 custom XPU designs.

The bottom line is a high-stakes uncertainty. The stock's 7% drop shows the market is deeply concerned about the Amazon relationship. But the company's own guidance suggests the risk may be overstated or already priced in. For now, the rumor mill is the overhang, and investors are left to weigh a bearish conviction against a bullish forecast.

The 2026 AI Storage Catalyst

The path to a re-rating for Marvell hinges on its ability to resolve near-term execution uncertainty and demonstrate leadership in the next frontier of AI infrastructure: storage and optical interconnects. The company's forward-looking product cycles are now explicitly designed to capture value as the AI bottleneck shifts from compute to data movement.

The immediate catalyst is its AI storage business, which is positioned to benefit from the industry's pivot to high-speed, low-latency storage. As AI models transition from training to inference and Retrieval-Augmented Generation (RAG), the need for advanced SSD controllers and memory fabrics becomes critical. Marvell's Bravera™ SSD controllers and Structera™ CXL products are central to this shift. The company's recent demonstration with Samsung and Liqid showed that its Structera A memory accelerator can deliver 5.3X more vector search queries per second than a standard system, highlighting the tangible performance gains from its CXL technology. This positions Marvell not just as a supplier, but as an enabler of the next generation of AI workloads.

A key strategic move to capture this value is the acquisition of Celestial AI for its "Photonic Fabric" technology. This deal is a direct play on the future of "million-GPU" clusters, where optical interconnects are essential for scaling beyond the limits of copper. The acquisition gives Marvell a unique edge in the race for scale-up fabrics, enabling high-bandwidth, low-latency connections within and across racks. As the evidence notes, this technology provides more than two times the power efficiency of copper interconnects and allows for co-packaging with high-power XPUs, a significant thermal and space advantage. This positions Marvell to lead the "plumbing" for the largest AI systems, a market it expects to expand significantly.

Finally, the company's long-term reacceleration is tied to a new hyperscaler engagement anticipated in fiscal 2028. While near-term concerns linger-particularly around visibility with Amazon's upcoming Trainium 3 and 4 designs-Marvell's management has signaled that its fiscal 2028 data-center revenue growth could "accelerate meaningfully". This projection, combined with the Celestial AI integration, provides a clear forward-looking thesis. The bottom line is that Marvell is executing a multi-year plan: leverage its storage and connectivity leadership to de-risk its legacy business, then capture the next wave of AI infrastructure demand. For investors, the catalyst is clear: watch for the ramp of its CXL and optical interconnect products in 2026, which will determine whether the stock can finally decouple from its cyclical anchors and re-rate on its AI infrastructure story.

Valuation and Near-Term Setup

The tactical setup for Marvell is defined by a stark contrast between near-term execution and longer-term potential. The stock's recent 7% drop reflects a sharp reassessment of its most immediate risk: the stability of its relationship with Amazon, its top customer for custom AI chips. Analyst Cody Acree's downgrade to a hold, citing a "high degree of conviction" that Marvell has lost the design for Amazon's Trainium 3 and 4 chips, has created a clear overhang. The concern is that Amazon is moving toward more economical designs and sourcing key technology like SerDes from competitors, which could materialize into a tangible revenue loss if Marvell fails to secure new orders for these next-generation chips. This uncertainty directly challenges the company's own guidance, which implies stability from continued Trainium 2 volumes and a Kuiper satellite project, not a transition to the newer designs many had anticipated.

The key near-term catalyst is the Q4 earnings report, which will provide the first concrete data point on this risk. The company has guided for net revenue of $2.200 billion +/- 5% for the quarter ending January 31, 2026. Investors should watch for any clarification on the Amazon relationship within that guidance. The report will also offer a snapshot of the pipeline for new customer engagements, which is critical for validating the company's long-term growth thesis beyond its largest client.

What investors should watch for is a clear signal on two fronts. First, management must provide more than just a statement of intent; it needs to offer tangible evidence of ongoing Amazon business or a credible plan for replacing any lost volume. Second, progress on the acquisition of Celestial AI and the integration of its photonic interconnect technology into the product roadmap will be a key indicator of execution on its strategic pivot. The acquisition is a bet on Marvell's ability to lead in optical interconnects for AI data centers, a market it believes will be worth billions. Any update on how quickly this technology can be commercialized will be a major driver for the stock's valuation beyond the next fiscal year.

The bottom line is a high-stakes test of credibility. The premium valuation is built on the promise of a reacceleration in data-center revenue, with CEO Matt Murphy pointing to a "meaningful" acceleration in fiscal year 2028. But for that promise to be believed, Marvell must first prove it can navigate the near-term turbulence with Amazon and demonstrate that its new strategic bets, like the Celestial AI integration, are moving from announcement to revenue. The next earnings report is the first checkpoint.