Unlocking Hidden Gems: Undervalued AI Infrastructure Partners in Nvidia's Ecosystem

The AI revolution is no longer confined to research labs-it has become the backbone of enterprise software, driving everything from predictive analytics to autonomous systems. At the heart of this transformation lies a critical question for investors: Which infrastructure partners are positioned to benefit from Nvidia's dominance while trading at a discount to their intrinsic value?

The Nvidia Effect: A New Era of Open Collaboration

Nvidia's recent NVLink Fusion initiative has redefined the AI infrastructure landscape. By opening its high-speed interconnect technology to partners like Qualcomm, Fujitsu, and MediaTek, Nvidia is fostering a hybrid ecosystem where third-party silicon can integrate seamlessly with its Blackwell GPUs. This shift from a closed to a semi-custom model has created opportunities for companies to innovate while leveraging Nvidia's leadership in AI compute.

However, not all partners are created equal. While some, like Intel, are securing multi-billion-dollar investments NVIDIA and Intel partnership, others, such as Marvell and Qualcomm, are quietly building out AI infrastructure with compelling valuation metrics.

Marvell Technology: The Undervalued Workhorse

Marvell's forward P/E ratio of 46.3x and price-to-sales ratio of 9.6x, according to a Marvell financial analysis, suggest it trades at a discount to peers like AMD and Nvidia. This is striking given its role in supplying custom AI chips to hyperscalers like Amazon and Google. In Q3 FY2025, 73% of Marvell's revenue came from its custom silicon business, as noted in a Futurum report, driven by a five-year agreement with Amazon AWS.

The company's 2nm SRAM technology and UALink interconnect are highlighted in a TradingNews profile that positions it to capture 25% of the data center accelerator chip market by 2028, according to an AI Chips report. Yet, its valuation remains anchored to legacy metrics. With AI-related revenue projected to hit $2.5 billion in FY2026 (TradingNews profile), Marvell appears undervalued relative to its growth trajectory.

Qualcomm: A Strategic Shift to Edge AI

Qualcomm's P/E ratio of 15.74, according to Qualcomm ratios, is among the lowest in the semiconductor sector, reflecting skepticism about its transition from mobile to AI infrastructure. However, its Snapdragon X series and data center processors, designed to work with Nvidia GPUs, were discussed in an article titled processors team up with Nvidia, which signals a pivot to edge AI-a $154 billion market by 2030 (AI Chips report).

The company's Q2 FY2025 revenue of $10.8 billion, shown in Q3 2025 slides, and $8.89 billion in R&D spending, detailed in a Qualcomm analysis, underscore its commitment to innovation. With discussions underway for server infrastructure silicon (as reported, Qualcomm says it's ready), Qualcomm's undervaluation may be a temporary discount on its long-term AI ambitions.

(Note: subsequent references to the AI Chips report and Qualcomm's server plans are presented without hyperlinks.)

Fujitsu: A Quiet Powerhouse in Japan

Fujitsu's P/E ratio of 30.68 is listed on Fujitsu PE ratio, higher than its historical averages, but its collaboration with Nvidia on Monaka CPUs and AI agents is detailed in a Fujitsu collaboration that positions it as a leader in Japan's AI infrastructure market. The company's full-stack solutions, combining x86 and Arm architectures, are tailored for healthcare and robotics-sectors expected to grow at 30% CAGR, according to a Gartner report.

While its valuation isn't as attractive as Marvell's, Fujitsu's $24.54 billion in trailing revenue is shown in Fujitsu statistics and its strategic focus on industry-specific AI applications make it a compelling long-term play.

The Valuation Gap: Why These Stocks Are Overlooked

The AI infrastructure market is projected to grow at 30.4% CAGR through 2030, according to an AI infrastructure market report, yet investors are still pricing many partners based on legacy business models. For example:
- Intel's P/E of 0.00 is documented in the historical Intel PE ratio data and reflects ongoing financial struggles, despite a $5 billion Nvidia investment (NVIDIA and Intel partnership).
- MediaTek's P/E of 19.9 is shown in the MediaTek PE ratio and, while reasonable, lacks the AI-specific revenue streams of Marvell or Qualcomm.

The key differentiator is custom silicon adoption. Companies like Marvell and Qualcomm are securing multi-year contracts with hyperscalers, creating recurring revenue streams that traditional metrics fail to capture.

Conclusion: Where to Allocate Capital

For investors seeking undervalued exposure to AI infrastructure, Marvell and Qualcomm stand out. Marvell's custom silicon expertise and Qualcomm's edge AI pivot offer asymmetric upside in a $1.5 trillion AI spending market (Gartner report). Fujitsu, while more mature, provides stability in a region where AI adoption is accelerating.

Nvidia's ecosystem is no longer a closed loop-it's a launchpad for partners willing to innovate. The question now is: Which of these partners will outperform as AI becomes the new electricity?