AT&S and Marvell Technology Deepen Collaboration for Next-Generation AI Infrastructure

Powering Tomorrow's Intelligence: AT&S and Marvell Technology Deepen Collaboration for Next-Generation AI Infrastructure

The relentless march of artificial intelligence is reshaping industries, economies, and societies at an unprecedented pace. As AI models grow in complexity and scope, the demand for robust, high-performance infrastructure to support them has skyrocketed. Today, September 22, 2026, marks a pivotal moment in this evolution as two industry leaders, AT&S and Marvell Technology, announced a significant expansion of their strategic collaboration, set to accelerate the development of next-generation AI infrastructure. This deepened partnership underscores the critical role of advanced semiconductor solutions in building the foundation for an increasingly intelligent world.

The AI Revolution Demands Advanced Infrastructure

The promise of artificial intelligence, from autonomous systems to hyper-personalized experiences, hinges entirely on the underlying hardware and software that enable its computational feats.

What is AI Infrastructure?

AI infrastructure refers to the comprehensive ecosystem of hardware and software components specifically designed and optimized to build, train, deploy, and manage artificial intelligence models and applications efficiently. This includes high-performance servers, specialized processors like GPUs and custom ASICs, advanced networking solutions, and the critical interconnects that bind these elements into a cohesive, high-speed system.

It provides the necessary computational power, memory bandwidth, and low-latency connectivity required to handle the massive datasets and complex algorithms inherent in modern AI workloads. Without sophisticated AI infrastructure, the potential of cutting-edge AI innovations would remain largely untapped.

The scale of this undertaking is immense. According to Gartner, worldwide spending on AI is forecast to reach a staggering $2.7 trillion in 2026, marking a significant 49.5% increase year-over-year. A substantial portion of this, approximately 56%, is allocated directly to AI infrastructure, which is projected to hit $1.48 trillion this year. John-David Lovelock, Distinguished VP Analyst at Gartner, succinctly summarized the situation, stating that "The buildout of AI data center capacity is the largest infrastructure project humanity has ever undertaken." Data from PwC further reinforces this outlook, projecting global investment in AI infrastructure to reach a record $31.6 trillion through 2050, with annual data center capital expenditure expected to rise from approximately $800 billion in 2026 to $1.8 trillion by 2050.

AT&S and Marvell: A Strategic Alliance Deepens

The expanded collaboration between AT&S, a leading manufacturer of high-end IC substrates and printed circuit boards, and Marvell Technology, a data infrastructure semiconductor solutions powerhouse, is strategically timed to address these burgeoning demands. The partnership aims to significantly enhance IC substrate capacity specifically for AI and cloud infrastructure, ensuring that the foundational components for advanced AI systems are readily available and continuously innovated.

A History of Innovation: Complementary Strengths

Both companies have independently, and in previous engagements, demonstrated a strong commitment to advancing the technologies crucial for AI. AT&S has been a key player in the advanced packaging space, with its solutions enabling higher performance and lower power consumption for next-generation microchips. The company’s strategic focus on AI infrastructure and advanced computing systems has resulted in a remarkable 53% growth in its Microelectronics business unit for the fiscal year 2025/2026. This growth is driven by strong demand and successful cost optimization efforts, contributing to AT&S’s overall revenue reaching €1.8 billion in FY 2025/2026, a 21% increase in constant currency.

Marvell Technology, on the other hand, has consistently led in custom silicon and interconnect solutions. The company's offerings are central to hyperscalers building out AI infrastructure, providing custom silicon tailored for various applications and an extensive portfolio of interconnects and network switch products. Marvell reported record fiscal 2026 revenue of $8.195 billion, a 42% year-over-year increase, largely fueled by robust AI demand. Its Q2 Fiscal Year 2027 revenue further demonstrated this momentum, hitting $2.739 billion, up 37% year-over-year, with data center revenue accelerating by 46%.

The Catalyst for Expansion: Why Now?

The decision to deepen this collaboration on September 22, 2026, is a direct response to several critical industry trends:

• Explosive AI Growth: The insatiable demand for AI processing power requires a corresponding increase in the underlying hardware, particularly high-performance logic and memory chips.
• Shift to Chiplet Architectures: The industry is moving away from monolithic chips towards chiplet designs, which integrate multiple specialized dies onto a single package. This architectural shift places immense pressure on advanced packaging and IC substrates, requiring larger, more complex, and higher-layer-count substrates.
• Bottlenecks in the Supply Chain: The manufacturing of advanced substrates, particularly those utilizing Ajinomoto Build-up Film (ABF), faces significant supply constraints. Ajinomoto Fine-Techno, for instance, controls over 95% of the upstream dielectric film supply, making supply allocation and manufacturing yield paramount. AT&S is one of the few global manufacturers capable of producing these critical ABF substrates for AI accelerators.
• Performance and Efficiency Imperatives: As AI data centers grow, power consumption and heat dissipation become major challenges. Advanced packaging and optimized interconnects are crucial for achieving higher performance per watt.

Unpacking the Expanded Collaboration: Next-Generation AI Infrastructure

This expanded partnership focuses on boosting the capacity and technological capabilities of IC substrates, a fundamental component in advanced semiconductor packaging. Specifically, the agreement leverages AT&S’s cutting-edge manufacturing prowess and Marvell’s custom silicon expertise.

AT&S's Kulim, Malaysia, manufacturing site is at the heart of this expansion. The plan involves equipping its second plant and establishing new manufacturing facilities to significantly ramp up production of high-end IC substrates. These investments, ranging from €1.5 to €2.0 billion, are fully supported and financed by long-term customer commitments, signaling strong market confidence and a clear roadmap for capacity growth. This strategic move has already led AT&S to increase its outlook for the fiscal year 2026/2027, now expecting a currency-adjusted revenue growth of 45-55% and an EBITDA margin of 32-37%.

Michael Mertin, CEO of AT&S, emphasized the significance of this milestone, stating, "This expanded collaboration with Marvell Technology is a testament to our shared vision for innovation and value creation in the AI era. Our advanced IC substrates are becoming increasingly vital for the complex semiconductor designs that power AI."

Marvell's role in this collaboration is equally crucial. The company relies on AT&S's advanced substrates to realize its next-generation custom silicon designs and interconnect solutions. Marvell is at the forefront of enabling AI infrastructure, designing custom chips tailored for specific applications and offering a comprehensive portfolio of interconnects and network switch products. Vinay Krishna, a spokesperson for Marvell Technology, highlighted that "securing necessary manufacturing capacity through partners like AT&S is paramount as demand for AI infrastructure continues its explosive growth."

Key technological areas of collaboration and impact include:

• Advanced Packaging Platforms: Marvell has been delivering advanced packaging platforms for custom AI accelerators, including a modular RDL interposer design that offers an alternative to traditional silicon interposers and provides critical supply chain flexibility. This relies heavily on high-quality substrates from partners like AT&S.
• Glass Core Substrates: AT&S is actively advancing glass core substrates from research toward industrial use in AI, high-performance computing, and photonics. These substrates provide a stable foundation for larger, more capable chip packages and enable new functionalities, including optical links directly within the package. This is a crucial area for future AI scalability.
• Optical Interconnects: Marvell is a pioneer in optical interconnect technologies, showcasing industry-first 2-nanometer optical interconnect solutions at ECOC 2026 (European Conference on Optical Communication) in Malaga, Spain, from September 20-24, 2026. These include 2nm 400G/lane optical PAM4, 800G ZR/ZR+ pluggable, and 1.6T ZR/coherent-lite technologies, optimized for ultra-high-speed and energy-efficient communication within and between AI data centers.
• Custom Silicon for AI: Marvell's custom ASIC offerings specialize in addressing the complex, high-speed, and high-performance silicon requirements of next-generation cloud data centers and enterprise applications. Their broad IP portfolio, now extending to 3nm process nodes, is central to creating optimized custom compute and interconnect solutions that demand the highest quality substrates.
• Photonic Fabric™ Technology: Marvell's Photonic Fabric™ technology platform is purpose-built for the emerging scale-up interconnect market, designed to enable more tightly integrated, high-bandwidth, and power-efficient solutions for data center customers. This technology, aimed at memory sharing across multiple racks, inherently benefits from advanced substrate capabilities.

The Impact on the AI Ecosystem and Industry

This expanded partnership between AT&S and Marvell is poised to have a ripple effect across the entire AI ecosystem:

• Accelerated Innovation: By ensuring a robust supply of advanced IC substrates, the collaboration will help mitigate potential bottlenecks in AI chip manufacturing, accelerating the development and deployment of more powerful AI accelerators.
• Enhanced Performance and Efficiency: The synergy between AT&S's substrate technology and Marvell's custom silicon and optical interconnects will lead to AI systems with higher computational density, lower latency, and significantly improved energy efficiency. This is crucial for managing the immense power demands of hyperscale AI deployments.
• Greater Supply Chain Resiliency: In an era of increasing geopolitical and supply chain uncertainties, securing long-term capacity commitments for critical components like ABF substrates enhances the stability and predictability of the AI hardware supply chain.
• Fueling AI Adoption: More efficient and powerful AI infrastructure will enable broader and more sophisticated AI applications across various sectors, from scientific research and healthcare to finance and autonomous vehicles.

Market Outlook: Growth in AI Hardware

The collaborative efforts of companies like AT&S and Marvell are directly contributing to the exponential growth projected for the AI hardware market. The focus on specialized components and infrastructure highlights the capital-intensive nature of building out AI capabilities.

Here’s a snapshot of current and projected AI spending, emphasizing infrastructure:

Source: Gartner, September 2026 data. Note: Figures are rounded and percentages are approximate based on reported growth rates.

These figures underscore the massive investment flowing into the foundational elements of AI, with infrastructure leading the charge in terms of sheer dollar volume. IDC corroborates this trend, forecasting AI infrastructure hardware spending to reach $487 billion in 2026, a 53% increase from $318 billion in 2025. This significant growth trajectory validates the strategic importance of collaborations like that between AT&S and Marvell.

The Road Ahead: Challenges and Opportunities

While the future of AI infrastructure is bright, challenges remain. The intense demand for specialized materials like ABF, the complexities of advanced packaging, and the need for continuous innovation in power efficiency are ongoing hurdles. However, the proactive steps taken by industry leaders through strategic partnerships demonstrate a clear commitment to overcoming these challenges. The collaboration between AT&S and Marvell Technology represents a powerful fusion of expertise – advanced substrate manufacturing meeting cutting-edge custom silicon and optical interconnects. This synergy is essential for laying the robust, scalable, and efficient groundwork upon which the next generation of artificial intelligence will be built.

Frequently Asked Questions

What is the primary goal of the AT&S and Marvell collaboration?

The primary goal of the expanded collaboration is to boost the manufacturing capacity and technological capabilities of advanced IC substrates for next-generation AI and cloud infrastructure. This ensures a stable and sufficient supply of critical components that Marvell Technology requires for its high-performance custom silicon and interconnect solutions, ultimately accelerating the deployment of more powerful and efficient AI systems globally.

How will this collaboration impact AI data centers?

This collaboration will significantly impact AI data centers by enabling higher computational density, lower latency, and improved energy efficiency. By securing advanced IC substrates, the partnership facilitates the production of more sophisticated AI accelerators and optical interconnects, which are vital for handling the massive data loads and power demands of hyperscale AI operations. This will lead to more scalable and cost-effective AI deployments.

What specific technologies are central to this expanded partnership?

Central to this expanded partnership are AT&S's advanced IC substrates, particularly their ABF (Ajinomoto Build-up Film) based solutions and emerging glass core substrates, crucial for high-density chiplet designs. From Marvell's side, key technologies include their custom silicon (ASICs) optimized for AI, advanced packaging platforms, and cutting-edge optical interconnects, such as their newly demonstrated 2-nanometer optical technologies.

Why is advanced packaging crucial for AI infrastructure?

Advanced packaging is crucial for AI infrastructure because it allows for the integration of multiple specialized chiplets—such as compute, memory, and networking dies—into a single, high-performance package. This enables shorter signal paths, lower electrical losses, and more efficient heat dissipation, all of which are critical for achieving the extreme computational power and energy efficiency demanded by modern AI processors and data centers.

Featured image by Tima Miroshnichenko on Pexels