Marvell Surges 32% as AI Data Center Demand Drives Growth

icon MarsBit
Share
Share IconShare IconShare IconShare IconShare IconShare IconCopy
AI summary iconSummary

expand icon
Marvell surged 32.5% on June 2, reaching $290.79, fueled by AI and crypto news alongside support from NVIDIA CEO Jensen Huang at Computex. Huang emphasized Marvell’s custom ASICs and optical interconnects as critical for AI data centers. Founded in 1995 by Weili Dai and Sehat Sutardja in Silicon Valley, the company has grown into a global leader in semiconductors. The Dai family’s network encompasses six companies, two IPOs, and four acquisitions. Recent inflation data and strong tech demand continue to drive investor interest in AI-powered infrastructure.

Author: Ada, Shenchao TechFlow

On June 2, Marvell surged 32.5% in a single day, closing at $290.79, setting a new all-time high—a 265% increase over the past 12 months. The direct catalyst was Jensen Huang naming Marvell’s custom ASICs and optical interconnects as “the core of AI data center architecture” at Computex.

It’s rare enough for a company to be personally endorsed by NVIDIA’s CEO at a conference. This company was founded in 1995 by Weili Dai and her husband Sehat Sutardja in their living room. Weili Dai is the youngest of the three Dai siblings from Shanghai and one of the driving forces behind the family’s thirty-year footprint in the semiconductor industry.

Brother Dai Weimin is based in Shanghai and serves as Chairman of Verisilicon, the first company in China’s A-share market specializing in semiconductor IP. Verisilicon’s market capitalization is expected to reach approximately RMB 150 billion by 2026, with AI ASIC orders setting new records for six consecutive quarters. Brother Dai Weijin is currently a Director and General Manager of the IP Division at Verisilicon; he founded Vivante in 2007, which Verisilicon acquired for $575.3 million.

Over the past three decades, the six companies founded by the three Dai siblings have seen two go public and four acquired. But this is only half the story. What truly underlies it all is a vast network in the semiconductor industry woven by the union of two Chinese families—Dai and Sutardja—spanning from the United States to China, from EDA tools to advanced chiplet packaging facilities, and from IP licensing to AI SuperNICs.

The three siblings placed six bets over thirty years, each time timing it perfectly.

The three siblings from the Dai family all graduated with degrees in electrical engineering from UC Berkeley, and their entrepreneurial timelines precisely aligned with three paradigm shifts in the semiconductor industry.

In 1995, Wei-Li Dai and her husband Sehat Sutardja, along with Sehat’s brother Pantas Sutardja, founded Marvell in Silicon Valley, focusing on hard disk drive controllers using a fabless business model. That same year, their older brother Wei-Min Dai founded Ultima in Silicon Valley to develop EDA tools. At the time, as PC adoption was nearing its peak, the fabless design model and the automation of EDA tools were defining characteristics of the semiconductor industry’s first major restructuring—and the Dai family successfully bet on both. Ultima was acquired by Cadence in 2000, and Marvell went public in 2000.

In 1996, Wei-Jin Dai co-founded Silicon Perspective, which specialized in digital implementation EDA tools, and was acquired by Cadence for approximately $500 million in 2002. Around the same time, his older brother, Wei-Min Dai, turned his attention to China and returned to Shanghai in 2001 to found VeriSilicon, betting on the “IP licensing + one-stop chip customization” model to provide semi-finished products to China’s emerging SoC design companies. That same year, China joined the WTO, causing the number of domestic chip design companies to surge from over 100 to thousands. VeriSilicon became a key supplier of resources for this wave of growth.

In 2007, Dai Weijin founded Vivante, specializing in embedded GPU IP for automotive and IoT applications, at a time just before the mobile internet boom, when all devices began requiring graphics capabilities. In 2016, Dai Weimin’s Verisilicon fully acquired Vivante for $57.53 million, and Dai Weijin transitioned from CEO of Vivante to General Manager of Verisilicon’s IP Division. This family-led acquisition connected the two strands of “China’s leading IP company” and “embedded GPU IP.”

In 2019, Wei-Li Dai launched her third venture. After leaving Marvell, she co-founded Dream Big Semiconductor in Silicon Valley with Sehat and former Marvell executive Sohail Syed, focusing on an open chip platform and AI SuperNIC. In 2021, Wei-Li Dai, Sehat, and South Korean semiconductor veteran Han Byung Joon co-founded Silicon Box in Singapore to establish an advanced chiplet packaging facility. Chiplets represent the industry’s only viable path to further enhancing single-chip performance following the slowdown of Moore’s Law, positioning them as a key bet for the post-Moore era.

In August 2020, VeriSilicon, founded by Wei-Min Dai, listed on the STAR Market, raising RMB 1.862 billion and was hailed by the market as the "first semiconductor IP company to go public." In October 2025, Wei-Li Dai's Dream Big was acquired by Arm for $265 million in cash.

Thirty years, six companies, two went public, four acquired by top buyers. It’s an impressive record—but looking only at this record misses the other half of the story.

image

An industrial foundation built by two families

In 1995, Marvell’s founding trio consisted of Weili Dai, Sehat Sutardja, and Sehat’s brother Pantas Sutardja. Sehat was born in Jakarta, Indonesia, and became a licensed radio technician at age 13. He earned his Ph.D. in Electrical Engineering from Berkeley in 1988, where he met and married Weili Dai. From its very first day, Marvell has been a joint venture of the Dai and Sutardja families—not merely a husband-and-wife startup.

This advantage has gradually grown over the past thirty years.

The Dai family has deep roots in China’s semiconductor ecosystem. The eldest brother, Dai Weimin, founded Verisilicon, China’s leading IP provider, whose partnerships with SMIC and Hua Hong Semiconductor date back to SMIC’s founding. Verisilicon developed SMIC’s first 0.18-micron standard cell library, resolving IP export restrictions SMIC faced at the time. The second brother, Dai Weijin, transitioned from Silicon Perspective to Vivante and back to Verisilicon, progressively building layers of EDA, GPU IP, and IoT customer networks into China’s domestic SoC ecosystem.

Now consider the Sutardja family: since the Marvell era, their network of engineers has extended across Southeast Asia and Europe. In 2021, Weili Dai and Sehat founded Silicon Box in Singapore with Han Byung Joon; by early 2024, the company had crossed the unicorn threshold. It has built a state-of-the-art semiconductor packaging facility in Tampines, Singapore, valued at approximately $2 billion, and is constructing a new $3.6 billion plant in Italy. Both facilities are supported by industrial policy alignment between Singapore’s Economic Development Board and the Italian government. This East Asia–Europe production coordination capability cannot be unlocked through the Dai family’s connections in mainland China.

More noteworthy is the combined investment portfolio of the Dai and Sutardja families. Publicly identifiable, they have invested in or co-founded at least 15 companies centered around the chiplet ecosystem: Alphawave, which develops high-speed SerDes interconnect IP, was acquired by Qualcomm for $2.4 billion in December 2025; Expedera, focused on NPU IP; BlueCheetah, specializing in UCIe die-to-die interface IP; Nubis, working on optical interconnects; Ventana, developing RISC-V server CPUs; and FLC, pioneering DRAM alternatives. Together with the previously mentioned Verisilicon, Vivante, Dream Big, and Silicon Box, these companies cover nearly every layer required in the chiplet era—semiconductor IP, interconnect standards, packaging facilities, and application-specific computing chips. The two families have collectively built the foundational industrial infrastructure for the post-Moore era.

Replicate Marvell's upward momentum logic

What is the logic behind Marvell's recent price increase?

Over the past year, the bottleneck in AI data centers has quietly shifted. GPU compute shortages defined the story of 2023–2024. By the second half of 2025, after large-scale training and inference are fully operational, the real constraints have become three things: custom ASICs (to reduce customers’ reliance on purchasing NVIDIA’s general-purpose GPUs), high-speed interconnects between chips, and advanced packaging capabilities to integrate all these components into a single package.

Marvell has simultaneously mastered the first two: designing custom ASICs like TPUs for Google and Amazon, and developing optical communication chips for high-speed data transmission. This is the real reason for its 265% surge in value over the past year—and the true reason NVIDIA invested $2 billion in Marvell in March. Even Jensen Huang needs this critical interconnect backbone.

Replacing the same image with the "Dai + Sutardja Family" logo completely changes the scene.

Dream Big is betting on the chiplet platform and AI SuperNIC (800 Gbps bandwidth for horizontal interconnects between GPUs). In October 2025, Arm announced a cash acquisition of $265 million. Arm’s intent is clear: it aims to evolve from selling CPU IP to becoming a “full-stack data center architect,” mirroring NVIDIA’s 2019 move of acquiring Mellanox for $6.9 billion.

Alphawave, which specializes in high-speed SerDes interconnect IP, went public in London. On December 18, 2025, Qualcomm completed its acquisition for $2.4 billion. The Dai + Sutardja family is Alphawave’s second-largest shareholder, and Dai Weili realized approximately $237 million from this transaction.

VeriSilicon in China operates an "IP + one-stop ASIC customization" business model, identical to Marvell’s in the U.S., but with customers shifted to Chinese AI chip buyers such as Alibaba, ByteDance, and Cambricon. In 2025, AI computing accounted for 73% of new orders; in the first four months of 2026, AI made up 91% of the newly signed contracts worth RMB 8.24 billion. With a market capitalization of approximately RMB 147.7 billion (USD 20.5 billion), it is roughly 8% the size of Marvell but growing at a steeper rate.

Silicon Box invests in factories specializing in advanced chiplet packaging. In early 2024, its valuation surpassed $1 billion. The company is neither publicly listed nor for sale, representing the戴+Sutardja family’s most significant bet in the critical production layer of AI infrastructure.

In addition, over ten companies we have invested in and incubated—such as Expedera (NPU IP), BlueCheetah (UCIe die-to-die interface IP), Nubis (optical interconnect), Ventana (RISC-V server CPU), and FLC (DRAM alternative)—each precisely addresses one of the three bottlenecks in the AI data center outlined above.

Adding these assets together, a conservative estimate puts the combined asset portfolio directly related to the AI wave at over $22 billion for the two families. This figure will not appear on any ranking list, as it is dispersed across five jurisdictions, four types of corporate structures, and more than a dozen companies—but it exists.

From the perspective of their family investment portfolio, the two families have made at least six independent bets in this AI data center wave, each of which aligns closely with the logic behind Marvell’s current rise. Marvell is their most prominent symbol, but far from their only entry point in this cycle.

Third path: Implement critical components at the standard switching point

Currently, there are two dominant narratives in the AI semiconductor industry.

One is large platform companies reaping the benefits: NVIDIA sells GPUs and the CUDA ecosystem, while Broadcom and Marvell sell custom ASICs and interconnects—this is the domain of players with market valuations exceeding $150 billion.

One approach is for ASIC-focused startups to go public independently. Companies like Tenstorrent, Cerebras, Groq, and Etched are bypassing NVIDIA to develop specialized compute acceleration solutions, betting on establishing an alternative to GPUs.

The Sutardja family has taken a third path: developing key components for open standards, building their own wafer packaging facilities, and either waiting to be acquired by large corporations or becoming a domestic IP leader. This approach is particularly logical in the chiplet era, as chiplets are inherently a product of resistance to closed vertical integration. With open standards in place, critical IP and packaging capacity become scarce assets—offering a much shorter path than independently developing a full ASIC and pursuing an IPO.

But the cost is clear: this path is destined never to produce the next NVIDIA. It may allow founders to exit gracefully multiple times and maintain long-term influence within the industry ecosystem, but it will not place anyone on the final podium of AI infrastructure.

When Wei-Li Dai co-founded Marvell in a living room in Silicon Valley in 1995 with her husband, the company had little recognition. Today, Marvell is a $254 billion market cap star in AI data centers. Although most of Dai’s equity from 30 years ago has been exited at various stages, she and her family still hold cash proceeds from exits including芯原 (VeriSilicon), Silicon Box, Alphawave, and the sale of Dream Big to Arm, as well as equity stakes in over a dozen chiplet ecosystem companies.

Marvell was her most prominent battle, but not her only one, nor her last.

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.