# Interconnect Related Articles

HTX News Center provides the latest articles and in-depth analysis on "Interconnect", covering market trends, project updates, tech developments, and regulatory policies in the crypto industry.

After Marvell's 32% Surge, the Chinese Chip Family Behind It Emerges

The stock price of Marvell Technology surged 32.5% on June 2nd, driven by NVIDIA CEO Jensen Huang highlighting its custom ASICs and optical interconnects as core to AI data center architecture. This event brought attention to the Chinese semiconductor family behind Marvell: the Dai siblings. The story centers on three siblings, all UC Berkeley graduates, whose three-decade entrepreneurial journey aligns with major semiconductor industry shifts. In 1995, youngest sister Dai Wei Li co-founded Marvell with her husband Sehat Sutardja and his brother, focusing on storage controllers. Eldest brother Dai Wei Min founded EDA company Ultima, later sold to Cadence, and later founded VeriSilicon (芯原) in China, becoming a leading semiconductor IP provider. Second brother Dai Wei Jin co-founded EDA firm Silicon Perspective (sold to Cadence) and GPU IP company Vivante, later acquired by VeriSilicon. The combined "Dai-Sutardja" family network extends beyond Marvell. Their ventures and investments form a comprehensive ecosystem for the post-Moore's Law, chiplet era. Key holdings include: Dream Big Semiconductor (AI SuperNICs, acquired by Arm), Alphawave (high-speed SerDes IP, acquired by Qualcomm), and Silicon Box (a chiplet advanced packaging foundry). VeriSilicon itself thrives on the AI ASIC and IP boom in China. Collectively, the family's AI infrastructure-related portfolio is estimated at over $22 billion. Their strategy represents a distinct path: building critical components for open standards and key manufacturing capacity in the chiplet era, rather than pursuing standalone AI chip dominance. While this path may not create the next NVIDIA, it has enabled repeated successful exits and sustained influence within the global semiconductor industry.

marsbit06/03 11:16

After Marvell's 32% Surge, the Chinese Chip Family Behind It Emerges

marsbit06/03 11:16

'ASIC Giant' Marvell Sets Record Quarterly Revenue, Raises Guidance Again, CEO Says Data Center Business Is 'On Fire'

Marvell Technology, a leading player in custom AI chips and data center connectivity, reported record revenue for its fiscal Q1 2027, driven by explosive demand in its data center business. Revenue reached $2.418 billion, slightly surpassing expectations, though GAAP net income fell year-over-year due to acquisition-related costs. Crucially, data center revenue hit $1.83 billion, making up 76% of the total and growing 27% YoY. The company significantly raised its full-year and next-year guidance, citing "exceptionally strong AI-related orders." Revenue is now projected at ~$11.5 billion for FY2027 and ~$16.5 billion for FY2028. CEO Matt Murphy emphasized that growth in the data center segment is accelerating. The AI Interconnect business, now expected to grow over 70% annually, saw its forecast lifted again due to rising network demands in complex AI models. Additionally, Marvell's custom chip (XPU) business is on a steep growth path, with FY2028 revenue anticipated to double and a target of over $10 billion by FY2029. The company also expanded its strategic collaboration with NVIDIA, focusing on silicon photonics, system integration, and AI-RAN solutions. To secure supply for surging demand, Marvell plans about $1 billion in supplier prepayments this fiscal year, highlighting its long-term capacity planning. Despite the strong results, the stock dipped slightly in after-hours trading.

marsbit05/28 04:09

'ASIC Giant' Marvell Sets Record Quarterly Revenue, Raises Guidance Again, CEO Says Data Center Business Is 'On Fire'

marsbit05/28 04:09

Bernstein's 97-Page Report Decoded: The Battle for AI Data Center Connectivity, Who Will Be the True Winner by 2026?

Bernstein's 97-page report analyzes the AI data center connectivity landscape. It argues that the bottleneck is shifting from raw compute (GPU) to the systems connecting GPUs, crucial for cluster efficiency. Copper and optical interconnects are not in a simple replacement cycle but will coexist long-term, with copper dominating short-distance "scale-up" connections and optics favored for longer "scale-out" scenarios. While Co-Packaged Optics (CPO) is the long-term direction for power and cost savings, its widespread adoption faces manufacturing and reliability hurdles, with mass deployment unlikely before 2028. Transitional technologies like Linear Pluggable Optics (LPO) and Near-Packaged Optics (NPO) are seen as near-term leaders. A key insight is that CPO will fundamentally reshape the value chain, shifting profits from traditional optical module suppliers towards chip designers (e.g., NVIDIA, Broadcom), advanced packaging (e.g., TSMC), and system integrators. For 2026, the report highlights more immediate and certain investment opportunities in the essential "infrastructure" enabling this connectivity shift. This includes upgrades for PCBs, ABF substrates, and CCLa driven by new AI server/switch platforms, alongside demand for 1.6T optical modules, LPO/NPO, and the testing/validation equipment required for future CPO scale-up.

marsbit05/19 03:16

Bernstein's 97-Page Report Decoded: The Battle for AI Data Center Connectivity, Who Will Be the True Winner by 2026?

marsbit05/19 03:16

Understanding CPO (Co-Packaged Optics) in One Article: Why Nvidia Is Willing to Spend $3.2 Billion on a Fiber?

NVIDIA and Corning announced a multi-year strategic partnership on May 6, 2026, with NVIDIA committing up to $3.2 billion to support Corning's U.S. expansion. This investment will triple Corning's manufacturing plants and significantly boost its optical fiber and communications production capacity. The core driver behind this massive investment is the fundamental shift from copper to optical interconnect technology within AI data centers. As GPU clusters scale, copper wires face critical limitations: severe signal attenuation over distance, high energy consumption for signal integrity, and excessive heat generation. Optical fiber, transmitting light instead of electrical signals, solves these issues with minimal loss, near-light speed, and lower power needs. The article outlines a three-stage evolution of data center interconnect: 1. **Traditional Copper Interconnects:** The mainstream solution of the 2010s, now being phased out due to scaling bottlenecks. 2. **Pluggable Optical Modules:** The current mainstream, where modules convert electrical signals to light externally. This process still introduces energy loss and latency. 3. **CPO (Co-Packaged Optics):** The next-generation technology where the optical engine is integrated directly with the GPU chip package. This drastically reduces the electrical signal travel distance to mere millimeters, slashing power consumption and latency while boosting data density. NVIDIA CEO Jensen Huang has identified CPO as an essential core technology for AI infrastructure. NVIDIA's investment signifies a strategic shift from being a buyer to actively controlling its supply chain for critical components. With demand for specialized optical fiber far outstripping supply—evidenced by soaring prices—securing long-term manufacturing capacity has become a competitive necessity. While Corning's expansion may pressure some suppliers, a projected global fiber supply gap of 5-15% over the next few years creates a significant opportunity window, particularly for Chinese manufacturers competitive in optical preforms, chips, and modules. Ultimately, NVIDIA's move is not about chasing a trend but an engineering imperative. The transition to light-based interconnects like CPO is driven by the physical limits of copper, marking a definitive step in the ongoing AI computing revolution.

marsbit05/11 10:07

Understanding CPO (Co-Packaged Optics) in One Article: Why Nvidia Is Willing to Spend $3.2 Billion on a Fiber?

marsbit05/11 10:07

活动图片

Hot Categoriesright-arrow

Hot Tagsright-arrow

Market Analysis

Regulatory Policy