OCP Sparks Innovation in Southeast Asia—Next Stop: Taiwan !

Original Articles By SemiVision Research (2025 OCP , Ranovus , Lightmatter , EXPO, Alphawave Semi, MediaTek , AMD , Cerebras )

OCP (Open Compute Project) 2025 “South East Asia Tech Day”

Event Overview

• Date & Venue:
  July 9–10, 2025, from 8:00 AM to 5:00 PM
  Marina Bay Sands Expo & Convention Centre, Singapore

• Expected Attendance:
  Approximately 400+ participants, including:

  • 40% from hyperscale data center operators

  • 40% technical architects

  • 20% strategic decision-makers

Core Themes and Key Topics

This event will focus on a wide range of current and future-critical issues, including:

• Regulatory and Policy Developments
  Impacting data center growth across Southeast Asia

• GPUs and AI Accelerators
  Challenges in deployment for compute-intensive applications

• High-Density Compute Architectures
  Strategies for management and performance optimization

• Open Collaboration and Innovation
  In system design, thermal management, networking, and more (e.g., Open Rack V3, dynamic power management)

• OCP Certification & Sustainability
  Standards such as OCP Ready™ for Hyperscale and sustainable architecture guidelines

Event Highlights & Networking Opportunities

• Tech Talks & Exhibition:
  Industry speakers and companies will showcase cutting-edge developments in AI, server design, and open infrastructure—encouraging cross-disciplinary technical exchange.

• Networking Reception & Sundown Drinks:
  A valuable platform to connect with technology leaders, hardware vendors, and system integrators.

• Regional Forums & Strategic Roundtables:
  Sessions tailored to local issues such as regulatory frameworks, market demand, and long-term sustainability planning.

Introduction to the Open Compute Project (OCP)

The Open Compute Project (OCP) is a global, non-profit organization dedicated to driving the adoption and standardization of open hardware design across data centers, cloud computing, high-performance computing (HPC), and AI infrastructure. OCP’s mission is to break the limitations of proprietary systems by fostering innovation and efficiency through open collaboration in compute, storage, networking, and data center infrastructure.

Background and Development

OCP was founded in 2011 by Facebook (now Meta) to address the growing demands of its rapidly expanding data center operations. By designing its own servers and racks, Facebook achieved significant efficiency gains, and later chose to openly share its designs—sparking a revolution in open hardware. To scale this initiative and build a broader ecosystem, Facebook joined forces with Intel, Rackspace, Goldman Sachs, and Andy Bechtolsheim, among others, to establish the OCP Foundation.

Since its founding, OCP has evolved from a platform focused on open server designs into a comprehensive standards organization spanning servers, storage, network switches, power systems, rack architecture, liquid cooling, time synchronization, edge computing, and AI accelerators.

Technical Community and Contributor Network

OCP operates as an open, community-driven platform with over 250 member organizations and research institutions worldwide. These include hyperscale cloud service providers, OEMs/ODMs, chip designers, materials companies, manufacturers, academic institutions, and government agencies.

Key members include:

• Meta (founding member)

• Microsoft (major promoter of OCP-certified data centers)

• Google, AWS, Tencent, Alibaba Cloud

• Intel, AMD, NVIDIA, Broadcom

• Dell, HPE, Supermicro, Wiwynn, Inventec, Foxconn, UfiSpace, and others

OCP encourages all members to actively contribute to its technical working groups and releases all certified designs in open formats—such as specifications, CAD files, and firmware—promoting supply chain transparency, reusability, and rapid innovation.

Global Activities and Impact of the Open Compute Project (OCP)

The Open Compute Project (OCP) hosts multiple global and regional events each year to drive collaboration and showcase innovations in open hardware design. These events play a vital role in expanding the reach and influence of OCP's open ecosystem.

Key Events and Engagement Platforms

• OCP Global Summit (California, USA)

  

  The largest open hardware conference in the world, attracting thousands of professionals from across the globe. It serves as the primary platform for major announcements, technical discussions, and ecosystem building.

• OCP Regional Summits (APAC, Europe, Middle East)
  Focused on region-specific industry needs, these summits—such as OCP China Day, OCP Taiwan/APAC Summit, and OCP Regional in Prague—promote local engagement and strategic collaboration.

• Engineering Workshops & PlugFests
  Regular hands-on sessions that support technical validation, interoperability testing, and the development of implementation standards.

• OCP Experience Centers
  Located in Taiwan, Singapore, Amsterdam, Texas, and other regions, these centers provide real-world demonstrations of OCP designs and deployment environments.

  

Core Values and Standardization Framework

OCP operates based on five foundational design principles:
Openness, Modularity, Reusability, Scalability, and Sustainability.
To ensure the quality and applicability of open designs, OCP has developed a set of robust certification and standardization mechanisms:

• OCP Accepted / OCP Inspired Certification
  These certifications verify that products meet OCP's open design specifications, allowing vendors to build upon and innovate from these designs.

• OCP Ready™ Program
  A facility-level certification for data centers, ensuring that infrastructure aligns with OCP’s deployment and operational requirements.

• Focused Subprojects
  OCP drives technical advancement in specific domains through initiatives like:

  • OpenRMC (Remote Management)

  • OpenEdge (Edge computing systems)

  • SONiC (Software for Open Networking in the Cloud)

  • Time Appliance Project (Precision timing for distributed systems)

OCP’s Expanding Role in the Global Infrastructure Landscape

As a global leader in open hardware, OCP is reshaping the design logic of data center and computing infrastructure. By leveraging an ecosystem of industry contributors, OCP not only reduces total cost of ownership (TCO) for infrastructure but also drives the industry toward higher efficiency, lower energy consumption, and long-term sustainability.

Looking forward, OCP will continue to expand its global footprint—enabling more enterprises, governments, and institutions to adopt and benefit from open innovation.

---

Introduction to OCP Taiwan

OCP Taiwan (Open Compute Project Taiwan) is one of the key regional communities under the global Open Compute Project (OCP) organization in the Asia-Pacific region. It is dedicated to advancing the development and international integration of open data center infrastructure design in Taiwan. As cloud computing, AI, and high-performance data centers rise rapidly, OCP Taiwan actively serves as a bridge between Taiwan’s local supply chain, technical communities, and global standards. Its mission is to support local companies in participating in the global open hardware ecosystem and promote the adoption of open standards in the domestic market.

Core Mission and Activities

OCP Taiwan’s primary objectives include:

• Encouraging Taiwanese companies to participate in OCP specification development and contribute open hardware designs.

• Facilitating collaboration between Taiwan’s supply chain and global hyperscalers such as Meta, Google, and Microsoft.

• Hosting technical workshops and seminars to enhance knowledge sharing and community engagement.

In terms of physical presence, OCP Taiwan has partnered with UfiSpace in New Taipei City to establish the “OCP Experience Center.” This facility showcases OCP-certified systems such as servers, network switches, and power modules, serving as a crucial platform for real-time engagement between local suppliers and international buyers.

Growing International Role

Since 2019, OCP Taiwan has actively organized a range of regional technical forums. In 2025, it was elevated to the host of the OCP APAC Summit, a flagship event for the Asia-Pacific region. The summit drew participation from over 100 international and local technology leaders, covering essential topics such as:

• Server architecture

• Open networking

• Liquid cooling technologies

• High-density power systems

This demonstrates Taiwan’s increasing strategic importance in the global data center supply chain.

Vision and Global Integration

Through deepening industry participation, advanced technology adoption, and international collaboration, OCP Taiwan aims to position Taiwanese enterprises at the core of the global open data center technology ecosystem.

OCP Taiwan 2025 Annual Technical Summit – You’re Invited!

As generative AI and large-scale data center demands grow, interest in Open Compute Architectures is surging worldwide. Hosted by the Open Compute Project Foundation and co-organized by the OCP Taiwan regional community, the OCP APAC Summit 2025 will be held from August 5 to 6, 2025, at the Taipei Nangang Exhibition Center Hall 2 (TaiNEX 2).

This summit will be the largest OCP technical event in Asia-Pacific, bringing together over 500 global experts, industry leaders, and open hardware innovators. The agenda will span cloud computing, AI chips, data center design, energy-efficient systems, and advanced packaging, making it a must-attend event for those shaping the future of open infrastructure.

Event Highlights

Global Keynote Sessions

Hear directly from Meta, Google, Microsoft, Intel, AMD, Broadcom, Supermicro, UfiSpace, and other core OCP members and partners. Topics include strategies and real-world implementations of next-generation open data centers, GPU-accelerated platforms, and sustainable computing technologies.

Technical Forums & Breakout Sessions

The conference will feature in-depth technical tracks focusing on key topics such as:

• High-Density AI Servers & Liquid Cooling Thermal Management

• Innovations in OCP Open Rack V3 & Power Supply Chain

• Open Networking Architectures (SONiC, ORv3, Time Appliance Project)

• OCP Certification & Modular Data Center Deployment Case Studies

• How Taiwan’s Supply Chain Is Adopting & Contributing to OCP Standards

Live Demonstrations & Supplier Booths

Explore OCP-certified products and solutions firsthand, including:

• Server motherboards

• Network switches

• Rack systems

• Cooling modules

• Power subsystems

Experience real-world applications and engage directly with leading hardware providers.

Networking & Collaboration Opportunities

Meet face-to-face with hyperscalers, data center operators, and system integrators. This is a prime opportunity to build strategic relationships and explore technical or commercial partnerships.

Event Information

• Event Name: OCP APAC Summit 2025 (OCP Taiwan Annual Technical Summit)

• Dates: August 5–6, 2025 (Tuesday–Wednesday)

• Venue: Taipei Nangang Exhibition Center Hall 2 (TaiNEX 2), Taiwan

• Organizer: Open Compute Project Foundation

• Co-organizer: OCP Taiwan Community

• Registration Fee: USD $100 (Discount available for OCP members)

• Registration Link: https://www.opencompute.org/summit/2025-ocp-apac-summit

Who Should Attend

This event is highly recommended for:

• Data center owners and operators

• AI chip and system developers

• Suppliers of packaging, power, thermal, and server components

• OCP-certified solution and product teams

• Technologists interested in open hardware ecosystems and standards

Gain early access to the latest OCP standards and ecosystem developments. Strengthen your global industry connections and enhance your influence in the future of open infrastructure.

Act Now
Kate Hendle kate@opencompute.org
Dirk Van Slyke dirkv@opencompute.org

OCP Taiwan warmly invites you to join this landmark summit and co-create the next chapter of open computing for the AI era. Register today to take part in Asia-Pacific’s most influential open hardware event — and help shape the future.

Overview of Key OCP Events in 2025

2025 OCP Global Summit

• Date: October 13–16, 2025

• Location: San Jose, California, USA

• Theme: "Leading the Future of AI"

• Format: A four-day premier global summit on open computing technologies, featuring keynote speeches, co-located project workshops, technical sessions, an expo hall (including the Innovation Village), and networking events.

2025 OCP EMEA Summit

• Date: April 29–30, 2025

• Location: The Convention Centre Dublin, Ireland

• Highlights:

  • The largest regional OCP event in Europe, with over 1,141 participants.

  • Focus areas include data center sustainability, energy efficiency, AI, data protection, power and cooling, and advancements in quantum and optical communications.

  • Includes the Future Technologies Symposium, exploring forward-looking technologies such as liquid cooling, quantum systems, and AI computing architectures.

2025 OCP Canada Tech Day

OCP Canada Tech Day will bring together Canadian technical and business leadership, along with the OCP Community, to discuss and share progress and ideas important to advancing data center and edge technologies for AI. OCP Canada Tech Day is intended to showcase the latest innovations and technical advancements from around the world. Featuring talks from experts who are researching and designing the next generation of data center and edge computing equipment in the morning, followed by two afternoon breakout sessions covering (1) responsibly deploying next generation open AI/HPC clusters and (2) open ecosystems for mobile network hardware platforms.

• Date: June 12, 2025

• Location: Montreal Convention Centre, Canada

• Key Insight:
  Despite a broad range of topics covered, silicon photonics emerged as a focal point for the industry. As a critical enabler for overcoming the limitations of electronic interconnects and driving high-bandwidth, low-power computing, silicon photonics is rapidly becoming central to the evolution of AI, big data, and next-generation data center architectures.

One company gaining increasing attention in this space is Ranovus, a global leader in optical engines and Co-Packaged Optics (CPO) solutions. Their innovations highlight the accelerating convergence of photonics and advanced packaging.

In future reports, we will provide an in-depth analysis of Ranovus' recent announcements at SEMICON Taiwan 2024, and explore how its technology may integrate with System on Wafer (SoW) architectures to form a foundational element of next-generation AI computing platforms.

Ranovus: Pioneering a Photonics-Driven Future for Data Center Interconnects

Founded in Ottawa, Canada, Ranovus specializes in the development of high-bandwidth, low-power optical engines and interconnect solutions. The company’s core technologies include its proprietary ODIN® photonic engine platform, integrated silicon photonics, on-chip amplifiers, laser arrays, and Wavelength Division Multiplexing (WDM) technologies. What sets Ranovus apart is its ability to integrate both electronic and photonic components within a single package, enabling a true Electronic-Photonic Integrated Circuit (EPIC) architecture.

The ODIN platform supports the development of 800G, 1.6T, and even 3.2T optical modules, incorporating power-optimized designs that achieve energy efficiency below 1pJ/bit—significantly outperforming traditional pluggable optics. Emphasizing an architecture-driven photonic integration approach, Ranovus provides flexible and customizable interconnect configurations tailored to AI chip packaging needs, including co-packaged optics (CPO), near-packaged, and chiplet-to-chiplet optical solutions.

Potential Integration with Advanced Computing Platforms

1.AMD: Photonic Partner for GPU/NPU Platforms

Ranovus has repeatedly demonstrated the integration of its Optical I/O platform with AMD’s Instinct GPU systems. Using its ODIN photonic engine, Ranovus can deliver over 3.2 Tbps of multi-channel optical output to AMD's HPC platforms—effectively resolving the electrical interconnect bottlenecks in high-density GPU configurations. For future GPU chiplet architectures, Ranovus’ photonic engines are designed to integrate directly into multi-chip module (MCM) or SoIC systems.

2.MediaTek: Optical Foundation for ASIC and Edge AI

MediaTek has been actively advancing its ASIC and edge AI strategies, particularly in areas such as AI chip interconnects and low-power packaging integration. Ranovus’ EPIC platform and photonic engines offer strong integration potential with MediaTek’s proprietary SerDes IP and advanced packaging technologies (e.g., FOPLP), positioning Ranovus as a promising optical interconnect partner for Taiwan’s future system-level AI solutions.

3. Cerebras: Natural Interconnect Extension for Wafer-Scale AI

Cerebras' Wafer-Scale Engine (WSE), which utilizes an entire wafer as a single computational unit, presents enormous external interconnect demands. Ranovus’ compact multi-channel ODIN engine supports chip-to-chip optical links, providing a low-latency, high-bandwidth, and power-efficient communications backbone for wafer-scale platforms. Especially in multi-wafer data transport or System-on-Wafer (SoW) architectures, Ranovus could play a pivotal role.

Ranovus and the Future of SoW Architectures

System-on-Wafer (SoW) represents one of the ultimate forms of future AI system architecture—integrating multiple compute and memory modules directly on a single wafer, interconnected through interposers, optical interfaces, and power delivery networks to form a unified system. Ranovus’ optical engines are poised to serve as the core enabler of high-speed node-to-node communication within SoW, reducing routing complexity while achieving unprecedented energy efficiency and modular scalability.

Not Just Optics—An Architectural Forerunner

Ranovus’ value lies not only in building ultra-high-speed photonic engines, but in its architectural foresight. It delivers customizable solutions tailored to compute platform needs—from pluggable optics, to co-packaged optics, and even wafer-integrated interconnects. As the industry transitions from external pluggable modules to in-package interconnects, Ranovus is clearly emerging as a key enabler and leader in this paradigm shift.

Looking ahead, we will continue to track Ranovus’ technology showcases and customer collaborations at key events such as Semicon Taiwan, OFC, and Supercomputing, and analyze how its innovations are shaping the next-generation ecosystem for AI interconnects and advanced packaging.

EXFO: A Key Measurement Partner in Silicon Photonics

EXFO is a leading company specializing in optical communication and photonic component testing. In recent years, the company has actively expanded into the Silicon Photonics domain by offering advanced test and measurement solutions tailored to the high-speed, multi-channel nature of silicon photonic chips.

EXFO provides a modular optical testing platform that supports a wide range of functionalities—from electro-optical characterization, waveguide loss, and optical modulation efficiency, to bit error rate (BER) and eye diagram analysis. These systems feature high sensitivity, low noise, and automated control, making them ideal for both wafer-level production testing and laboratory validation of silicon photonic devices.

By accelerating both design cycles and mass production, EXFO’s solutions improve test consistency and yield, establishing the company as a critical technology partner in Photonic Integrated Circuit (PIC) verification workflows.

---

Alphawave Semi: A Core Driver of High-Speed Connectivity and Electro-Optical Integration

As the demand for high-speed, low-latency data transmission continues to surge—fueled by AI computing, cloud data centers, and 5G infrastructure—connectivity architecture has become a battleground in semiconductor design. Based in Canada, Alphawave Semi is an advanced chip design company specializing in high-bandwidth connectivity solutions. Leveraging deep expertise in SerDes IP, chiplet architectures, electro-optical co-packaging, and custom SoCs, Alphawave has rapidly emerged as an essential enabler of data-driven infrastructure.

From High-Speed IP to Optical Chips: Full-Stack Connectivity Solutions

Founded in 2017 by industry veteran Tony Pialis, Alphawave Semi has built its reputation on licensable high-performance connectivity IP, covering standards such as PAM4/NRZ SerDes, PCIe, CXL, Ethernet MAC, and UCIe. These IP solutions are widely deployed across AI accelerators, networking chips, storage controllers, and ASICs—establishing a reliable, high-speed backbone for modern systems.

Beyond IP licensing, Alphawave actively develops custom SoC designs and chiplet-based modular platforms, offering end-to-end solutions from design and validation to packaging integration. Their designs are compatible with TSMC’s advanced nodes (3nm–16nm) and packaging platforms such as CoWoS and InFO-oS, making them highly adaptable for applications in AI/HPC, switching, and low-power computing.

Advancing Silicon Photonics and CPO for the AI Era

To address the performance and energy limits of electrical signaling in next-gen AI data centers, Alphawave has entered the domain of electro-optical integration, launching chip and module solutions for Co-Packaged Optics (CPO) and Silicon Photonics applications. These products enable high-speed, low-power, co-packaged optical interconnects, breaking through the physical limits of copper interconnects and offering ideal solutions for large-scale AI clusters and distributed architectures.

In 2022, Alphawave strengthened its capabilities by acquiring OpenFive, the SoC division of RISC-V company SiFive, boosting its SoC design and optical integration capabilities. Since then, Alphawave has launched multiple multi-chip system platforms that support HBM, high-speed SerDes, and integrated controllers, further reinforcing its position in the AI and cloud computing markets.

Global Market Presence and Strategic Collaborations

Alphawave Semi serves more than 85 SoC customers and 40+ IP licensees, with over 350+ successful tape-outs and shipments exceeding 150 million chips globally across North America, Asia-Pacific, and Europe. In June 2025, Qualcomm announced a $2.4 billion acquisition of Alphawave Semi, highlighting the company’s strategic value in shaping the future of data center and interconnect technologies.

Alphawave also maintains strategic partnerships with Siemens, TSMC, and Synopsys for process validation, EDA tool integration, and advanced packaging test flows—building a comprehensive, high-performance connectivity platform.

A Silicon Platform Powering the Future of Connectivity

Amidst the rapid scaling of AI models and growing computational intensity in the cloud, Alphawave Semi stands out with its end-to-end stack of connectivity solutions—from IP to SoC, from electrical to optical. Whether it's ultra-low-power SerDes supporting PCIe 6.0 or co-packaged silicon photonic modules for AI accelerators, Alphawave is redefining the boundaries of performance, energy efficiency, and scalability in data center interconnects—delivering truly forward-looking innovations to the industry.

---

Lightmatter: Pioneering the Photonic Computing Era

As AI models scale exponentially and data center bandwidth demands surge, the limitations of traditional electronic architectures—particularly in power consumption, bandwidth, and latency—are becoming increasingly apparent. Lightmatter, an innovative company based in California, is redefining the core architecture of AI data centers by leveraging photons as the fundamental unit for both computing and interconnect. The company’s mission is to build a high-performance, energy-efficient world enabled by electro-photonic integration.

Vision: From Electronic to Photonic Logic

Born from MIT’s in-depth research on the feasibility of photonic computing, Lightmatter focuses its strategy on three foundational technologies:

• Envise: a photonic compute accelerator

• Passage: a photonic interconnect engine

• Idiom: a software abstraction layer enabling photonic hardware adoption

Together, they form a complete photonic computing stack, from hardware to system-level solutions.

Envise: The AI Inference Engine That Computes with Light

Envise is a hybrid electro-photonic AI accelerator that offloads compute-intensive tasks—such as matrix multiplications—to photonic chips, while using traditional electronics for control and non-linear operations. This architecture delivers both high performance and exceptional energy efficiency. Lightmatter has successfully demonstrated ResNet and BERT models on Envise, achieving the same accuracy as 32-bit digital systems—without requiring quantization. This milestone affirms the commercial viability and precision of photonic accelerators.

Passage: Breaking Interconnect Bottlenecks with Silicon Photonics

One of the most pressing challenges in data centers and AI clusters is the bandwidth and latency of inter-node interconnects. Lightmatter’s Passage products, built on silicon photonics, deliver hundreds of Tbps per module using WDM (Wavelength Division Multiplexing) and up to 256 optical fibers, enabling ultra-dense I/O. The company utilizes 3D packaging to vertically integrate photonic interconnects directly with GPU/NPU modules, minimizing latency and energy consumption. This architecture is particularly well-suited for AI mega-clusters, LLM training, and real-time data processing.

Idiom: Enabling Seamless AI Model Portability to Photonic Hardware

To lower adoption barriers, Lightmatter developed Idiom, a software platform that supports mainstream AI frameworks and offers automated model optimization and deployment tools. Idiom abstracts away hardware complexities, allowing AI models originally built for electronic systems to be executed seamlessly on the hybrid photonic-electronic architecture.

Commercialization and Strategic Partnerships

Lightmatter has partnered with GlobalFoundries to manufacture its photonic interconnect chips using mature silicon photonics processes. Packaging and 3D assembly are handled by Amkor. According to company announcements, the Passage M1000 series will enter commercial deployment in 2025, providing plug-and-play photonic interfaces for hyperscale data centers and AI networks.

Lightmatter is also actively involved in standardization bodies such as the Optical Internetworking Forum (OIF), Ultra Ethernet Consortium (UEC), and UALink, contributing to the development of open specifications for photonic interconnect in next-gen cluster architectures.

Financial Strength and Market Potential

In 2024, Lightmatter raised $400 million in Series D funding, backed by strategic investors including GV (Google Ventures), Microsoft, and HPE Pathfinder. The company’s valuation has reached $4.4 billion, and it is reportedly considering an IPO to support its expansion plans. Analysts believe that if Passage and Envise gain adoption among hyperscalers, Lightmatter is poised to become a leading force in photonic computing.

The Next Leap in AI Infrastructure

Photonic technology is no longer a laboratory prototype—it is rapidly becoming a commercial reality. Through fully productized hardware and software solutions, Lightmatter is bringing photonic computing and interconnect into mainstream AI infrastructure. In the coming years, as generative AI continues to strain existing architectures, high-bandwidth, low-latency, energy-efficient photonic solutions are expected to become standard in data centers—and Lightmatter stands at the forefront of this transformation.