On July 13, the semiconductor sector experienced a sharp decline, with escalating geopolitical tensions serving as one of the primary drivers. It is not surprising that many investors, concerned that further deterioration of the Middle East conflict could weigh on the broader equity market, chose to lock in profits. However, for capital with a long-term investment horizon, the current volatile period恰恰 presents a favorable opportunity to position in top-tier semiconductor stocks. Among them, Nvidia (NVDA) and Marvell Technology (MRVL) merit particular attention.
Relative to its growth potential, Nvidia’s current share price is surprisingly inexpensive. The stock trades at a price-to-earnings ratio (based on forward earnings) of 24.1 times, compared with the information technology sector average of 21.7 times. As the undisputed leader in the graphics processing unit (GPU) market, Nvidia benefits from a wide economic moat owing to high switching costs. With artificial intelligence infrastructure spending continuing to grow, the company still faces enormous opportunities, rendering the current price level quite attractive.
Although bears worry that competitors will eventually catch up, virtually no enterprise has so far made substantial progress in shaking its empire. Meanwhile, Nvidia is no longer merely a GPU company; its product portfolio covers multiple layers of the AI infrastructure stack. The company also sees broad opportunities in the CPU market, which is why it launched its standalone Vera CPU—and this is merely the tip of the iceberg.
Furthermore, Nvidia has recently significantly increased its per-share dividend and has committed to returning at least 50% of its free cash flow to shareholders on an ongoing basis through dividends and share repurchases. All these factors make the stock highly compelling. Even after the astonishing rally of the past several years, Nvidia’s upside potential is not yet exhausted.
Marvell is a leading manufacturer of application-specific integrated circuits (ASICs). These custom chips are developed for specific workloads and, while they may lack some versatility, they deliver high efficiency in designated tasks such as AI model training and are cost-effective when deployed at scale. A growing number of enterprises will rely on ASICs to capture the vast and expanding opportunities in the AI space.
Consider Amazon, for example. The company is exploring the possibility of selling its Trainium chips, which are designed by Marvell, to other data center operators. Amazon would not be investigating this potential if it did not see robust demand. Similarly, Alphabet has indicated that it will sell its custom AI chips to certain external customers. These developments are all positive signals for Marvell.
On the financial front, Marvell continues to deliver impressive results. In the first quarter of fiscal 2027, ended May 2, the company posted record revenue of $2.4 billion, up 28% year over year, with adjusted earnings per share of $0.80, a 29% increase from the same period last year. Revenue is expected to accelerate further in the coming quarters. As demand for custom AI chips surges toward the end of this decade and beyond, particularly from hyperscale data center customers, the company is poised to be a major beneficiary.
Marvell supports the UALink open interconnect standard, which is designed to provide AI chip manufacturers with an alternative to Nvidia’s proprietary NVLink architecture. In March 2026, Nvidia invested $2 billion in Marvell, while the two companies simultaneously established a strategic partnership covering custom AI chips, NVLink Fusion-compatible networking, optical interconnects, and silicon photonics technologies. Marvell had previously helped customers build alternative networking solutions; with Nvidia’s investment, it now ensures that it can also support Nvidia’s networking offerings. No public information currently indicates that Marvell has abandoned UALink support. What may be even more intriguing is that Marvell could become more valuable precisely because it is able to serve both ecosystems.
As Nvidia advances toward the Rubin generation, each increment in compute density places greater pressure on the surrounding network. This creates demand for technologies that Marvell has been developing for years, including SerDes technology for high-speed electrical signal transmission, switches for network traffic steering, optical DSPs for data transmission over fiber optics, and custom silicon businesses that help customers design processors tailored to specific workloads. For Marvell, Rubin is not just another product cycle for Nvidia; it also expands the connectivity challenges that Marvell’s solutions are designed to address.
In addition, Marvell acquired Polariton Technologies in April 2026 to gain plasmonic photonics modulation technology, aiming to advance optical connectivity toward 3.2T and beyond, proactively addressing the higher bandwidth and lower power consumption requirements of future AI systems.
