Marvell Technology (MRVL) lifted its fiscal year 2028 revenue projection to approximately $20 billion on Tuesday, driven by strong demand for its data center chips. Shares rose nearly 9 percent in early trading.

The company designs and supplies the critical high-bandwidth plumbing that allows GPUs and AI accelerators to operate at scale within data centers. Marvell currently generates about $9.45 billion in revenue from integrated circuits, with roughly $3.9 billion sourced from China and the remainder from Taiwan, the United States, and other regions.

What separates Marvell from the field is its vertically integrated control of four critical layers: custom ASIC design, 1.6T optical DSP, silicon photonics through its Celestial AI acquisition, and CXL switching. No single competitor replicates this full-stack position today. This architectural advantage translates directly into pricing power and customer lock-in as hyperscalers build out their AI infrastructure.

Seaport analyst Jay Goldberg initiated coverage with a Buy rating and $270 price target. At current levels, that target implies meaningful upside and reflects conviction in Marvell's ability to capture outsize share of the AI infrastructure buildout.

Marvell's 2021 investor day strategy centered on custom and cloud-optimized silicon engineered specifically for data centers. The $20 billion fiscal 2028 target represents roughly 2x the company's current revenue run rate—a CAGR that assumes sustained AI capital spending but is achievable given Marvell's embedded position in hyperscaler procurement.

Astera Labs (ASRA), another pure-play AI infrastructure name, develops semiconductor connectivity chips that link GPUs, accelerators, memory, and networking within data centers. The company generates approximately $1.2 billion from semiconductor connectivity products, primarily serving cloud and AI customers in China, Singapore, and Taiwan.

Astera targets physical bottlenecks inside AI clusters—the signal integrity, wiring, and system orchestration that determine whether expensive accelerators run efficiently. Management expects to expand its Scorpio X customer base beyond its current lead hyperscaler by year-end. Execution here matters: a transition from single-customer concentration to a true platform story would justify a higher multiple.

Seaport also assigned Buy ratings to ON Semiconductor and Semtech as beneficiaries of the AI infrastructure cycle.

For equity portfolios with AI infrastructure exposure, the key distinction is between companies selling picks-and-shovels (Marvell, Astera) and those riding adoption curves elsewhere in the stack. Marvell's combination of custom silicon, optical DSP, and switching creates a defensible moat that justifies premium multiples to the broader chip sector. Monitor quarterly gross margins and hyperscaler customer concentration as leading indicators of execution risk.