Silicon photonics didn’t become a manufacturing reality overnight, and few people have shaped that journey as directly as Dr. Yusheng Bian. Recently named an Optica Fellow for transformative contributions to silicon photonics, Bian has spent almost a decade at GF advancing the innovations that underpin scalable, high‑volume CMOS platforms. Following his recognition as an Optica Fellow at OFC, we sat down with Bian and asked him to reflect on the long arc from early research to real-world deployment and the momentum now carrying the technology forward.

Q: What originally drew you into the field of silicon photonics, well before it became as visible as it is today? 

CMOS technology has been developed for decades, and it underpins much of modern computing with a tremendous impact on everyday life. What initially drew me to silicon photonics was realizing that the same silicon substrate used for electronics could also be used to guide light. 

This light isn’t visible to the human eye. It operates in wavelength ranges like the near‑infrared, but the idea that silicon could propagate light was deeply exciting. Even more compelling was the fact that the same platform could support both electronic and photonic signals. That meant you could integrate the two, enabling not only light propagation but also electrical control of light and conversion between optical and electrical signals. There’s a powerful connection there, rooted in the commonality of a single, well-established technology platform. 

Because silicon already has a mature ecosystem, this convergence creates confidence that the technology can scale. Fundamentally, photonics offers clear advantages over electrons for data movement, including lower loss, higher bandwidth and improved signal integrity. All of that made it feel like a direction that could deliver real value and meaningfully change how systems work. 

Early in my career, it wasn’t certain that silicon photonics would become a mass manufacturable or commercially viable solution. There was real uncertainty. But the potential of the technology, and the passion it inspired, kept me moving forward. Today, seeing this shift take shape and the technology adopted in real‑world applications has been incredibly rewarding. 

Q: What are some things that you’ve really been focusing on in the silicon photonics space? 

I think over the past few years, both GF as a team and I personally have made tremendous progress. We are now seeing AI data centers evolve to a point where traditional copper‑based electronic solutions can no longer keep pace with increasing data demands, especially for data transfer between chips and across data centers. 

Because of that, photonics is now widely recognized as an integral solution for next‑generation data transfer, replacing copper where electronic interconnects can no longer scale. It enables lower power consumption, longer transmission distances and significantly better signal integrity. 

This represents a real paradigm shift, from traditional copper wiring to optical fiber and now toward integrating optical functionality directly into silicon photonics, or photonic engines, that can be placed very close to electronic systems. 

Q: Was this explosion in demand for silicon photonics driven by AI something you anticipated?  

Even a decade ago when we launched our silicon photonics technology around 2016 to 2017, this was not obvious. That was when I joined GF as the first R&D engineer focused on silicon photonics. Back then, we knew silicon photonics was coming, but across the industry there was no clear consensus that it would definitively replace copper or traditional electronic solutions. 

Over the last few years, the AI boom, accelerating data center demand and strong ecosystem momentum have given us much greater confidence. Photonics is now clearly one of the essential solutions needed in data centers to address current bandwidth, power and signal integrity bottlenecks. 

I wouldn’t say this was a big surprise or completely unexpected, but we are seeing strong confidence and real momentum toward photonic solutions across the board. And this extends beyond data centers. We’re also seeing adoption in areas such as photonic computing, quantum computing, sensing and other emerging applications. 

I wouldn’t call photonics a ubiquitous solution yet, but it is becoming very widely adopted as a replacement—or complement—to state-of-the-art electronic technologies. Over the next decade, we expect to see major advancements and extensive deployment across these fields. 

Q: Silicon photonics today seems both established and still evolving. How do you see that balance?  

If you look at what is already in production today and widely adopted by major data center operators—companies like Google, NVIDIA and Meta—it’s mainly pluggable optical engines. These are optical modules shipped in very high volumes, generating billions of dollars in revenue globally, across the U.S., China and other major markets. That technology is already established and well adopted. 

Beyond pluggable optics, there are two additional paradigms emerging. The first is the transition from traditional pluggable optical engines to what we call co‑packaged optics. This is a more compact solution that brings the optical engine much closer to the computing unit, such as the switch ASIC. NVIDIA, for example, is actively driving innovation in this area. This is just the first step toward future data transfer architectures. 

Beyond that, photonics is also expected to play a critical role in quantum computing. GF has strong capabilities in 300mm CMOS integrated platforms that support multiple quantum applications. We partnered with PsiQuantum several years ago, with the goal of delivering industry‑leading quantum computing platforms that meet future requirements. 

Another important area is LiDAR. As we move toward autonomous vehicles, there is a push to replace traditional, bulky LiDAR systems—like the rotating units you see on vehicles today—with much smaller, chip‑scale photonic solutions. Silicon photonics has the potential to enable that transition. There is also significant opportunity in sensing applications, including biomedical sensing. These are all areas where we expect photonics, and silicon photonics in particular, to continue penetrating in the next decade. 

Q: What makes GF particularly well-positioned in silicon photonics? 

GlobalFoundries is well known as an advanced 300mm CMOS foundry. One of the key advantages of combining a CMOS foundry with silicon photonics is the ability to use the same semiconductor technology platform, the same silicon-on-insulator substrate, to guide and propagate light. That convergence is incredibly powerful. 

One of GF’s major strengths is its deep manufacturing expertise and well-developed ecosystem, spanning devices, physics, manufacturing and system‑level integration. This positions GF as a leading foundry capable of driving innovation in silicon photonics, not only at the research level, but also through technology development and into high‑volume production. 

This transition—from research to real-world, mass-produced applications—is something that’s difficult to achieve in academic labs or small research environments. At GF, we’re focused on high‑volume manufacturing to support practical, real‑world products. 

Beyond internal capabilities, GF also works closely with ecosystem partners—OSATs, suppliers, vendors and other collaborators—to continue pushing the boundaries of silicon photonics technology. Any recognition I receive should really be viewed not as a personal achievement, but as a reflection of the entire silicon photonics team and the broader ecosystem we work with. 

We’ve also strengthened our position through recent acquisitions, including AMF and Infinilink’s silicon photonics assets. AMF adds important geographic diversity, particularly in Asia, and gives us a greater manufacturing scale. As a result, GlobalFoundries is widely described by industry sources as the world’s largest pure‑play silicon photonics foundry by revenue. AMF also expands customer coverage across different application areas. For example, in datacom there are different optical bands—O‑band, C‑band—which you can think of simply as different “colors” of light. AMF enables us to serve customers and applications not fully covered by our existing U.S. fabs. 

Additionally, Infinilink’s technology strengthens our IP development capabilities, allowing us to offer not only PDKs but also silicon photonics IP servicing, providing customers with more complete, system level solutions. 

Q: What technical challenges still need to be solved for silicon photonics to move even faster? 

Silicon photonics is still evolving, especially when AI data centers continue to drive this huge demand, and there is continued need to drive higher bandwidths and lower power consumption. From a practical perspective, the photonics team is advancing data rates from 100 gigabits per second to 200 and now pushing toward 400‑per‑lane class. This represents a significant leap, enabling more data per channel at higher speeds and lower power. We believe we are at the forefront of this effort and will continue pushing the boundaries. 

Q: On a personal level, what keeps you excited about your work? 

I’m deeply grateful to be part of this multidisciplinary team. Every day, I work with talented colleagues from different backgrounds, gaining new perspectives and learning continuously. This is a novel technology, full of opportunity, and collaboration is essential to pushing it forward. Persistence has also been critical. Staying curious and remaining resilient in the face of setbacks makes all the difference. In fact, many of the challenges we encounter ultimately become catalysts for innovation. Don’t be afraid of mistakes because progress comes through collaboration across disciplines. 

Q: What does it mean to you to be named an Optica Fellow? 

I was both thrilled and humbled to receive this recognition because it was the first time at GF that we’ve had an optical fellow. Optica is the largest professional society in optics and photonics, not just silicon photonics, so I was very honored. But this honor is not just a personal achievement. It reflects the collective contributions, innovation and collaboration of the entire GF silicon photonics team over the past decade. So, I’m excited about this current moment and I and even more energized by what lies ahead—growing the team, expanding the ecosystem and delivering solutions our customers value as we keep pushing the technology forward. 

Q: Outside of work, what do you enjoy doing?  

I enjoy hiking and I also use AI for some creative work, like composing music and singing. Music is something I’ve been trying to pursue the past couple years because I think that exploration brings a new perspective and helps strike a good balance. Sometimes, through these activities, you can actually learn something new. Working in a field that is drastically different from your main area can help you connect the dots and bring new ideas back into your work. That kind of cross disciplinary thinking is sometimes where true innovation comes from. 

Quantum computing is no longer science fiction confined to the lab. It is a new computing paradigm with the potential to solve problems beyond the reach of classical systems. From simulating complex molecules for new drugs to discovering new materials, quantum computers will transform entire industries. But building quantum computers at scale requires more than brilliant ideas. It demands world-class semiconductor manufacturing expertise. That’s where GlobalFoundries (GF) plays a critical role. As the leading specialty foundry, GF is uniquely positioned to accelerate the quantum revolution, no matter which hardware approach ultimately wins. 

What is quantum computing? 
Quantum computing harnesses the three rules of quantum mechanics: superposition (a qubit can exist in multiple states at once), entanglement (qubits can be linked so the state of one instantly influences another) and interference (which amplifies the probability of correct answers while canceling incorrect ones). 

While a classical bit can exist only as a 0 or a 1, a qubit can exist in a superposition of both states at the same time until it is measured. When many qubits are combined, the system is described by an exponentially large set of possible states, represented mathematically as amplitudes – a property often referred to as quantum parallelism. A sequence of quantum gates transforms those amplitudes, but when the computation is measured, it still produces only a single result. The practical advantage of quantum computing comes from algorithms that harness quantum interference to increase the probabilities of correct answers while suppressing incorrect ones, thereby increasing the likelihood of obtaining a useful result, for example, through amplitude amplification techniques that generalize Grover-style speedups. 

Quantum computers are therefore best viewed as special-purpose accelerators. They outperform classical systems on specific problems such as simulating quantum systems, structured search, sampling, or extracting global properties, while complementing classical computers for everyday workloads. 

A gamechanger for industries 
The impact of quantum computers with sufficiently capable qubits will be profound: 

• Pharmaceuticals and life sciences: Accurate molecular simulations could slash drug discovery timelines from years to months. 

• Finance: Quantum algorithms will revolutionize portfolio optimization, risk modeling, fraud detection and derivative pricing, for faster and more accurate insights in volatile markets. 

• Logistics and supply chain: Real-time optimization of routes, inventory and manufacturing schedules could cut costs dramatically and improve resilience. 

• Materials science and energy: Direct modeling of complex quantum interactions among electrons could accelerate the design of breakthrough materials such as superconductors, advanced batteries, photovoltaics and solid-state electrolytes. 

• AI and machine learning: Quantum-enhanced models promise breakthroughs in pattern recognition and generative AI. 

• Cybersecurity: While quantum computers threaten current encryption (prompting the shift to post-quantum cryptography), they also enable ultra-secure quantum key distribution. 

McKinsey and others estimate the value of quantum computing at tens to hundreds of billions of dollars annually once systems with a few hundred to a thousand logical qubits (the error-corrected units needed for reliable computation) become available. 

The question isn’t if-it’s when and how fast we get there. 

The road to scale: Manufacturing and the modality challenge 
Quantum computing is entering a new phase. The question is no longer whether qubits can work in a lab, but how complete quantum systems will be manufactured and scaled reliably for real-world deployment.  

Today, there is no industry-wide alignment on a single qubit modality. Leading approaches include superconducting circuits, trapped ions, photonic qubits, silicon spin qubits, neutral atoms, topological qubits and others. Each has distinct strengths and trade-offs in terms of fidelity (accuracy of quantum operations), coherence (how long the qubits retain their “quantumness”), scalability (how many qubits can you cram into a system) and operating temperature. This diversity is healthy and drives innovation, but it also means the ultimate winners will be those who can manufacture at volume- reliably and at reasonable cost- regardless of which architecture prevails. 

That is exactly where GF’s semiconductor expertise gives the quantum ecosystem its strongest foundation 

Why GF is the best partner-across any modality 
Given the uncertainty around which qubit modalities will ultimately prevail, the critical challenge for the industry is not proving isolated devices in the lab, but enabling repeatable, high‑yield manufacturing with a clear path to volume production. Rather than betting on a single qubit technology, GF takes a manufacturing‑first approach to quantum computing – building scalable, configurable semiconductor platforms that can support a wide range of quantum architectures as they mature. 

That is where GF’s role in the quantum ecosystem is fundamentally different. 

GF’s strategy is rooted in leveraging existing, qualified semiconductor platforms and extending them, where needed, to meet emerging quantum requirements. This approach dramatically reduces development risk, cost and time compared to building bespoke, one‑off processes for each modality. It also enables quantum teams to anchor their roadmaps to technologies that are already proven in high‑volume manufacturing environments. 

Across modalities, quantum systems increasingly converge on a common set of manufacturing needs: tight process control, materials uniformity, integration of electronics and photonics, ultra‑low‑noise interfaces and advanced packaging to combine heterogeneous components. These requirements align directly with GF’s core strengths as a specialty foundry. 

GF brings together: 

• FD‑SOI technologies such as 22FDX®, which are actively being explored by the quantum community for tightly integrated classical control, readout and system‑on‑chip architectures 

• High‑voltage and RF‑capable platforms, enabling power delivery, signal generation and amplification functions that are increasingly critical as quantum systems scale 

• Advanced heterogeneous integration and packaging, allowing quantum processors, control electronics, photonics and interconnects to be combined into manufacturable system‑level solutions 

• Silicon photonics platforms on 300 mm wafers that form a scalable foundation for photonic quantum systems as well as optical interfaces 

• Importantly, GF enables R&D, prototyping and early‑stage quantum development on the same industrial manufacturing infrastructure used for volume production. This continuity helps quantum system developers avoid costly transitions between research fabs and production fabs – a challenge that has historically limited scalability in emerging technologies. 

As quantum computing progresses from experimentation toward deployable systems, manufacturability will increasingly determine which architectures scale, and which do not. By remaining technology‑agnostic and focused on extensible, reproducible manufacturing platforms, GF provides a stable foundation for innovation across superconducting, photonic, spin‑based, atomic and hybrid quantum approaches. 

In a field defined by architectural uncertainty but united by the need to scale, GF’s role is not to choose winners – but to enable them.

By Tim Nutt

This article was originally published on EDN and is republished here with permission. Read the original article on EDN. 

The rapid adoption of artificial intelligence (AI) across consumer and commercial markets is driving unprecedented investment in high-performance computing and networking. As AI models scale and proliferate across diverse applications, demand for compute power keeps rising. To meet this need, the power consumption of heterogeneous processing units (XPUs) is projected to climb from today’s 1–1.5 kW to more than 5 kW by 2030 [1]. This surge in power requirements is driving demand for denser, more efficient power conversion solutions from grid to core.  

The emerging power distribution architecture enabling AI scale 
Distribution of 415-480 VAC within datacenters causes a patchwork of electrical conversions. AC power needs to be converted to DC power to support battery backup, and back to AC for further distribution.  But as AI systems scale up, this energy loss is too costly to absorb. A key focus area for the industry is high voltage direct current (HVDC) distribution which reduces conduction losses, as well as the number of conversion stages across large clusters. 

The main proposed solutions are either ±400 V (Mt. Diablo) or 800 V (Kyber) DC power delivery.  The first phase of HVDC solutions will still rely on 415-480 VAC distribution with a sidecar power rack, thereby reducing some power conversion losses.  This step has fewer power conversion stages than existing systems and reduces conduction losses by delivering HVDC to the adjacent compute rack.  However, to further eliminate power conversion stages, data centers will distribute HVDC throughout the cluster.  Additional energy savings will be achieved by implementing the 800V DC-DC conversion within the system trays in compute racks, reducing busbar conduction losses. 

This deployment will require a significant step up in density and efficiency. The past few months have seen hyperscalers specifying their general needs [2] of higher rack-power capacity, power efficiency, density and scalability, as well as vendors responding with proposed converter topologies and considerations to meet those needs [3].   

This marks real progress, and it’s already clear that the key performance goals of the solutions are within reach. The benefits of these next-generation power delivery architectures include: 

1. High conversion ratio – Conversion from HVDC distribution to very low XPU core voltage with as few stages as possible requires a large step down ratio (>1000:1).  Solutions based on wide bandgap semiconductors such as gallium nitride (GaN) achieve higher conversion ratios due to higher breakdown voltages and reduced conduction and switching losses compared to silicon-based solutions. 

2. Significant density increase compared to current power supply unit (PSU) designs – The increase in XPU power consumption does not come with a corresponding increase in available volume for power electronics. Computer and network architectures impose a constraint on physical distance, creating a need for more compact power components. Thanks to their excellent switching characteristics, GaN power semiconductors can support higher frequency operation, allowing smaller energy storage components such as capacitors and inductors or transformers. 

3. Extremely high efficiency at scale – The extraordinary growth in datacenter power consumption means that power losses in every stage translate directly to energy costs.  Thus, the conversion ratio and high density must be achieved without sacrificing efficiency. GaN devices offer the best figures of merit – including including lower specific on-resistance, minimal switching charge, and better high-frequency FOM –  which result in the highest efficiency for a given ratio and density. 

How GaN is driving datacenter innovation 
The datacenter market demands not only advanced performance but also exceptional quality and reliability. Increasingly, industry consensus points to Power GaN as the key enabling technology for HVDC solutions in data centers.  

GlobalFoundries is developing GaN platforms to support this transition, including HV (650V) and MV (200V and below) devices. These platforms will offer industry leading figures of merit with the reliability and ruggedness that hyperscalers require to deploy AI at scale. 

Opportunities for Scaling HVDC Architectures
Looking ahead to broad solution deployment, there are several major opportunities that remain, each offering room to drive the next wave of innovation on topology selection and device optimization: 

• Establishing clear safety and isolation requirements: To date, safety and isolation have been discussed only in broad terms, but HVDC architectures will require isolation. Achieving safety and isolation compliance through spacing (creepage and clearance) can come at a significant cost to density, while achieving compliance mechanically via conformal coating or potting can degrade thermal performance – both of which complicate serviceability of systems in the field. Defining the right balance represents a major opportunity for innovation in materials, mechanical structures, and system architecture. 
• Defining EMI/EMC requirements for scaling next generation datacenters: With datacenters requiring strict electromagnetic interference (EMI) and electromagnetic compatibility (EMC) standards, the industry must determine how topologies can meet them. If bulky filter components are required to scale HVDC solutions, this may prevent density targets being met, potentially forcing alternate topology selection. It is crucial that these requirements scale to multi-GW datacenters allowing clusters to interoperate, otherwise compatibility and performance are at risk. 
• Converging on optimal step-down ratios and system-level power conversion strategy: Will the industry converge to a 16x or 64x step-down or, as the HVDC converter moves into the system tray, will system designers optimize the power conversion stages around different voltage levels?  If solutions are customized based on system-level optimization, this will likely lead to a need for regulated HVDC converters as well as unregulated fixed-ratio, with the two types having distinct transient requirements. These tradeoffs will affect overall system design in the future, from rack input to XPU. 

Enabling Scalable, Efficient, and Sustainable Datacenters 
As these solutions evolve and mature, GF will collaborate with our customers to optimize device development, integrate driver and sensor functionality with power devices, and heterogeneously integrate power devices with additional components.   

It is encouraging that along with the activity around converter feasibility, industry participants are also extremely active in pursuing open standards, such as the Open Compute Project’s Power Distribution sub-project, which will provide a roadmap for scalable, interoperable HVDC architectures.  

Adoption of HVDC architectures allows operators and OEMs to convert efficiency gains directly into XPU and network-cluster performance — delivering more usable Floating-point Operations Per Second (FLOPs) from the same energy footprint while reducing energy losses, lowering operational costs, improving rack-level density, and advancing sustainability goals through more efficient power delivery. Meeting these stringent demands at massive scale requires solutions that ensure interoperability and long-term ecosystem value remain top priority. 

注意事项。 

[1] Future AI processors said to consume up to 15,360 watts of power — massive power draw will demand exotic immersion and embedded cooling tech | Tom’s Hardware 

[2]  Asset Share – NVDAM  

[3] Swing Aboard the 800-V Bus: NVIDIA’s AI Power Architecture and the Chips to Drive It | Electronic Design