FADU is the Next-Gen Flash SSD Solutions Leader

In 2015, I founded this company driven by a vision to address emerging opportunities and challenges in the semiconductor industry. Before this, I was a partner at Bain & Company, where I extensively covered the technology sector in Asia, focusing on markets like China, Korea, and Japan. My primary clients included SK Hynix, Samsung, LG, and Huawei, and my work concentrated on the semiconductor market. This experience revealed a significant transformation underway in the industry.

FADU HQ building

Around 2013, through my business with Samsung, I immersed myself in understanding market trends by studying numerous analyst reports on the semiconductor sector. At the time, the prevailing belief was that the market was saturated, dominated by giants like Intel and AMD, with minimal room for innovation. Industry consensus pegged growth rates at approximately 2%, with most demand being met by legacy players. Innovation was largely perceived as unnecessary, focusing instead on manufacturing advancements such as new process nodes.

Consolidation was a key trend, with many believing that the only growth opportunity lay in mergers and acquisitions. Semiconductors were primarily used in IT devices like PCs, laptops, and smartphones, with chips such as CPUs and DRAMs fulfilling most demands. However, this perspective overlooked the profound shifts starting to emerge.

By 2014-2015, new applications like AI, data centers, cloud computing, big data, and automotive technologies began driving semiconductor demand. Two major trends reshaped the landscape. First, shift to centralized computing. Computing power increasingly moved from individual edge devices to centralized data centers. Second, IT diffusion across industries. Sectors like automotive and finance began integrating IT technologies, spreading semiconductor applications into homes, cars, and other domains. These changes rendered legacy architectures inadequate. Dominant players, with their standardized CPU architectures, could no longer meet the diverse requirements of data centers, automotive systems, and IoT applications. A new era of chip innovation was on the horizon.

During my time at Bain & Company, I observed a generational gap in the semiconductor talent pool, especially in the U.S. Most key engineers were in their 50s, 60s, or even 70s. This stemmed from decades of neglect, as the industry was considered mature and unworthy of fresh innovation. While electrical engineering departments continued producing graduates focused on manufacturing, computer science departments had shifted their focus to software, following the dominance of companies like Google, Apple, and Facebook. As a result, there was a "lost generation" of semiconductor architects in the U.S.

In contrast, Korea’s semiconductor industry, led by giants like Samsung and SK Hynix, continued to invest heavily in nurturing young talent. These companies funded research projects and supported the development of skilled engineers in computer science and semiconductor architecture. Korea and Taiwan emerged as unique regions with the capability to drive innovation, as other major players like Japan, Europe, and the U.S. had deprioritized semiconductor development, while China was still catching up.

Recognizing these opportunities and challenges, I envisioned leveraging Korea’s young and talented semiconductor engineers to drive innovation on a global scale. By bridging Korean expertise with global market demands, I believed we could address the evolving needs of the industry.

In 2015, Korean semiconductor companies were largely focused on supplying big conglomerates like Samsung and SK Hynix. While this approach ensured stable revenue streams, it limited growth potential. Most Korean companies aimed to be second- or third-tier suppliers, securing 40% of the global market by catering to our customer’s needs. This strategy, while lucrative, did not align with my goal of building a first-tier global semiconductor company.

Instead, I aimed to work directly with end customers, such as major U.S. tech companies like Microsoft, Meta, and Google, which were driving the market. To achieve our vision, we needed to bypass traditional supply chains and establish direct relationships with global leaders.

Over the past decade, we have worked relentlessly to realize this vision. Today, I am proud to see our initial purpose coming to fruition. By focusing on innovation, nurturing talent, and targeting global markets, we have positioned ourselves as a key player in the evolving semiconductor industry. While most Korean companies remain domestically focused, we have expanded beyond these boundaries, paving the way for sustainable growth and global impact.

After conducting interview with other equipment or material makers in the semiconductor industry in Korea, we learned most of them started as suppliers and are now starting their journey abroad. However, many observers in the media criticize Korea’s fabless industry, saying that it is lagging behind.

Over the last 30 years, most Korean suppliers have chosen the easier path of selling their products to Korean conglomerates. Their motivation is straightforward: Korean conglomerates typically do not purchase the most critical chips from Korean companies. Instead, they aim to replace expensive foreign suppliers with cheaper Korean alternatives. This strategy has shaped their approach—conglomerates do not demand cutting-edge technology from Korean suppliers but look for second- or third-tier products to meet commoditized needs.

I chose not to follow this path. Our approach is different. In terms of capability, I believe Korea has a significant advantage. From 2000 to 2015, Korean and Japanese PhDs in semiconductor architecture were exceptionally talented. If we can leverage this talent pool, there is enormous potential. The market now demands entirely new types of semiconductors across various sectors, including automotive, industrial machinery, data centers, and AI. By focusing on these opportunities and breaking free from the domestic "safe zone," Korean companies can truly compete on the global stage.

Right now, there seems to be a ceiling in semiconductor dimensions—we are reaching the three-nanometer, two-nanometer thresholds. Ten years ago, it was feasible to keep reducing the size by building on the same architecture as technology advanced. How do you realistically see the opportunities for Korean SMEs?

That’s no longer the case. In the past, supply and demand grew in parallel. However, between 2010 and 2015, that balance shifted. While advancements in technology continued, the cost reductions became incremental. Today, a single chip isn’t enough to build large-scale models. You need tens of thousands of GPUs for just one model. Technological improvements alone cannot meet this new demand, which highlights the need for architectural innovation.

Around 2015, the semiconductor industry reached a transitional period where new types of demands emerged. This required entirely new chip architectures. Merely relying on manufacturing capabilities became insufficient to address these challenges.

When you introduced this new mindset and architecture, how did the market react? What challenges did you face, and how did you overcome them?

I focused on assembling a capable team skilled in designing acceleration semiconductors, which require new architectures. The SSD controller space stood out to me. It’s a fascinating niche that bridges logic and memory chips. Initially, I felt diving straight into logic chips would be challenging since we were an unknown company. Instead, I saw an opportunity in SSD controllers—logical chips tied to memory. This was an ideal entry point for a Korean company targeting the global market.

In 2014, I identified a team in Seoul National University that had been researching SSD controller architecture for over a decade. I hired nearly the entire lab, including five PhDs and five master's graduates. This became our core engineering team. The timing was perfect—by 2015, the SSD market transitioned from the SATA interface (shared with hard disk drives) to the PCIe NVMe interface, which was designed specifically for flash technology. We focused on introducing a customized NVMe SSD controller architecture.

Incumbent SSD companies struggled with this shift due to legacy technology debt. Companies that had been selling chips based on older architectures couldn’t easily pivot to new designs without overhauling their entire software and hardware stacks. As a newcomer, we had the advantage of designing our architecture from scratch, aligning perfectly with new demands.

This approach opened a unique window of opportunity starting in 2015, which continues today. New demands, such as GPUs for AI models, require innovative architectures. Relying on legacy designs creates limitations, while new architectures tailored to emerging needs offer significant differentiation.

When presenting this approach to investors, I often joked about how Korea’s unique focus on SSD architecture gave us an edge. Around 2015, no other country was as invested in SSD controller research. This was largely because the memory chip market—dominated by Korean companies—wasn’t incentivized to explore this area globally. Two universities stood out in SSD controller research: Sungkyunkwan University (supported by Samsung) and Seoul National University. I recruited talent from Seoul National University, while Samsung likely took graduates from Sungkyunkwan. In my view, Seoul National University produced the best engineers in the field.

This ability to attract top talent was key to our differentiation. Korean universities continue to foster innovative ideas, particularly in semiconductors. As a result, numerous fabless companies have emerged in Korea, striving to introduce groundbreaking products. While some will inevitably fail, I believe that a few will succeed in delivering innovations not possible in other countries, simply because those regions lack the necessary expertise.

You introduced a new approach to collaborating with NAND manufacturers and SSD makers. Could you explain how it works? What advantages does this offer to newcomers?

Around 2016 or 2017, the entire storage industry began to realize that NVMe was the key innovation for the future. Introducing NVMe SSDs became a significant advantage for storage customers. At Dell Technologies World, Broadcom’s CTO highlighted this shift, explaining that SSDs required NVMe and a completely new architecture, vastly different from legacy designs. This posed a significant challenge for incumbent players, as transitioning from legacy to new architecture involved substantial costs and effort. However, for smaller companies starting from scratch, this presented an opportunity. I thought his explanation perfectly described our position and potential.

Could you provide more details about your project work?

We designed an entirely new controller around 2018. Unfortunately, at the time, no one was willing to listen to us. We believed our product was the best in the industry, but we were an unknown startup, which made it difficult to gain trust. Additionally, our product was crucial for data centers, and customers were hesitant to take risks with a new supplier.

Initially, we planned to supply our chip directly to SSD module makers like Samsung or SK Hynix. However, even they showed little interest. Startups inherently represent risks, and most preferred to rely on established suppliers. Realizing this, we pivoted from being solely a fabless chip company to becoming a full module designer. We hired experts in SSD module design to demonstrate that our controller worked flawlessly as part of a complete SSD solution.

This new strategy allowed us to approach end customers directly, such as global big tech companies. We provided them with samples of our SSDs, showcasing our innovative design and inviting them to evaluate the product. By this time, the market had started transitioning from legacy architectures to new ones, creating a window of opportunity.

A global big company tested our SSD and found it to perform exceptionally well. This led them to encourage SSD vendors to adopt our controller. As a result, we began working with Korean NAND Flash Memory manufacturer in 2021, marking the start of our business relationship. That success proved our capabilities and attracted the attention of other potential customers.

By 2022, we began expanding our customer base, securing partnerships with companies like Western Digital. Our approach is twofold: First, we persuade end customers, such as data center operators, by demonstrating how our SSDs improve performance and reduce costs. Once they are convinced, their influence helps us work with SSD manufacturers who supply to these end customers. This strategy has been instrumental in building our business and scaling our operations.

Is your ultimate goal to return to working with the major SSD makers, or do you plan to maintain a presence in both markets?

Ultimately, our focus is on chip design. While we develop modules, it’s not because they generate significant revenue for us. Instead, modules are a strategic tool to demonstrate the capabilities of our chips and convince end customers of their value. Selling modules isn’t our core business. Our goal is to design exceptional chips, validate their performance through module development, and then supply those chips to our partners who manufacture the modules.

This approach has proven effective. It involves finding talented individuals, identifying opportunities to grow in the global market, and bridging the gap between Korea’s technical talent and global demand. I firmly believe that Korea has an abundance of brilliant talent capable of driving innovation in areas requiring cutting-edge solutions. We are continually building new products to meet emerging demands, particularly in data centers and AI.

Our strategy revolves around expanding our portfolio, identifying the right teams, developing innovative products, capturing new demand, and introducing these solutions to the global market. Globally, the semiconductor market is straightforward: the U.S. accounts for about 60% of overall demand, China makes up 20–25%, and the rest of the world contributes 15–20%.

Our approach focuses on proving our products with top-tier U.S. companies first. Success in the U.S. establishes our credibility and makes it much easier to penetrate other markets, including China. If we become a recognized brand in the U.S., it simplifies convincing customers in other regions to adopt our solutions.

Recently, NVIDIA rolled out 13,000 Blackwell GPUs, in which your controller is part of the ecosystem. Do you think this partnership provides the credibility you need to approach a broader range of customers?

Our approach remains consistent—we aim to create the best products, regardless of the specific project or partnership. While this collaboration demonstrates the quality and reliability of our solutions, our focus is always on delivering exceptional performance and value directly to the end customers. It's not just about one project; it's about establishing long-term trust and delivering innovation across all customer relationships.

I understand your focus is on data centers and AI. Do you think this segment of the market will be sufficient for FADU, or do you envision expanding into other areas, such as automotive, and shifting your approach to develop new types of logic and memory connections?

Our primary interest lies in areas that demand new types of innovation—that’s our sweet spot. Competing in markets reliant on legacy products would put us at a disadvantage against large incumbents. To win, we need to identify where innovation is required and leverage our top talent to create products tailored to those demands.

Automotive is an interesting sector, but it lags behind in terms of technological innovation. The automotive industry often adopts trends from more advanced markets, like PCs or data centers, rather than driving innovation itself. For this reason, automotive is not a priority for us. While the application of technology in automotive differs, the underlying technology is often not new. Companies like Analog Devices, TI, and Renesas dominate this space, and their offerings cater to proven, established technologies. Even when automotive companies express interest in "new" products, these are typically adaptations of existing technologies rather than groundbreaking innovations.

In September, President Biden mentioned banning Chinese software in the development of chips for the automotive market. Do you think such regulations could disrupt the market and supply chain? Could this present an opportunity for FADU?

I don’t see this as a major issue for the automotive industry. As mentioned, the automotive sector requires new chips, but these chips don’t rely on groundbreaking technology. Instead, they often use proven technologies from other sectors, with some level of customization for automotive applications. Chinese companies already possess the capabilities to produce these chips effectively, so I don’t anticipate a significant disruption.

So, I believe we will remain focused on data centers. As I mentioned, the automotive market is better suited for incumbents with established expertise. For FADU, the data center market aligns better with our strengths and strategy.

You became the first unicorn in the fabless industry in Korea, which was a significant milestone. However, you’ve also faced criticism from the media regarding your results. Can you explain what happened?

Let me provide some context. As I mentioned earlier, we struggled to generate revenue until 2020. We founded the company in 2015, and our initial belief was that after building our first chip in 2018, we’d see strong sales within the first year. However, despite having a chip ready, we faced over two years of challenges and didn’t record meaningful revenue until 2022.

From 2015 to 2021, we essentially had zero revenue while investing heavily in chip development. However, thanks to partnerships with global big tech companies, as well as domestic and international memory manufacturing companies, and a space company, we generated 56 billion KRW (around $40 million) in 2022. This was a significant milestone and validated our efforts, garnering attention from the market. As 2023 began, many companies approached us, requesting supply.

Based on customer forecasts at the start of the year, our business plan projected revenue of $300 million—six times our 2022 revenue. Excited by this growth, we began preparations for an IPO. However, by the end of Q1, the market crashed. In April 2023, SSD exports—mainly from Samsung and SK Hynix—plummeted by 80% year-on-year, reflecting the collapse of the global market.

Interestingly, in June, Micron’s CEO reported during their shareholder meeting that they believed the worst was behind them and expected a rebound in H2 2023. We shared this expectation, but the reality was harsher—market conditions worsened. Ultimately, we lost many opportunities.

I understand the criticism we’ve received. However, I want to emphasize two points. First, we have always been transparent about our challenges. Second, we failed to fully grasp the magnitude of market fluctuations, largely due to our lack of experience with such severe downturns. Additionally, after our success in 2022, our team was overly optimistic and relied heavily on customer forecasts for 2023.

How do you see the semiconductor industry rebounding in 2025, and how can FADU position itself to leverage this?

Predicting the semiconductor industry, particularly for memory chips, is notoriously difficult, even six months ahead. Let me share some insights into the market dynamics. 2023 was a particularly tough year. Between 2015 and 2020, the semiconductor industry was expected to grow at just 2% annually. However, in 2020, the market suddenly grew by 30%. This surge, driven by the COVID-19 pandemic, was unprecedented. Initially, many companies reduced capacity in 2020, anticipating lower demand. But in 2021, demand surged as consumers bought PCs, smartphones, and other digital devices for remote work and online activities. Companies then overinvested in capacity, expecting the digitalization trend to persist.

By 2022-2023, demand waned, leaving the market oversupplied. If not for the pandemic’s unpredictable impact, supply and demand might have aligned with forecasts, avoiding such a severe downturn. Going forward, while fluctuations are inevitable, I don’t foresee the kind of extreme supply-demand gap we saw in 2023 repeating itself.

However, there’s another significant factor—China. The Chinese government recognizes the strategic importance of developing its semiconductor industry, particularly given US restrictions. They’ve heavily subsidized domestic chipmakers like CXMT, which now accounts for over 10% of the global memory market. These companies don’t need to prioritize profitability due to government support, which disrupts the global supply-demand balance.

In 2024, the memory market will likely remain challenging due to oversupply from China. However, Chinese manufacturers primarily focus on low-end products for consumer goods. High-end segments, such as data centers, where FADU operates, will be less affected. While the broader semiconductor market may struggle, we believe our focus on data centers and high-end applications will allow us to perform well despite these challenges.

What will be the key drivers for your company next year?

From 2022 to 2026, we anticipated exponential revenue growth, but the two-year delay disrupted our plans. Despite this, we have successfully retained and expanded our global customer base. The lack of sales last year wasn’t due to product issues but rather the sluggish market. For example, Meta anticipated an undersupply of SSDs in 2022 and overstocked inventories, resulting in zero SSD purchases in 2023.

Additionally, 2023 saw a significant boom in AI. Companies prioritized GPU purchases over other components, leaving products like SSDs with minimal demand. However, this trend has shifted, and companies are once again purchasing SSDs. Our customer relationships remain strong.

2024 FMS (Future of Memory and Storage) exhibition

The customer testing process for SSDs is rigorous, often requiring six months to a year. Testing for many of our products has already begun, which positions us for substantial revenue growth starting next year. The numbers will speak for themselves, and we are confident in our trajectory.

That said, there’s a slight chance of a market downturn similar to 2023, though this seems unlikely. Assuming that scenario does not repeat, I am confident everything will progress smoothly, and we will achieve our goals.

Have you developed a plan to diversify your revenue sources, such as by expanding your client base or product portfolio, to ensure a stable revenue stream that is less susceptible to market fluctuations?

Our primary focus is on leading customers—particularly the major data centers. If these big players adopt our products, the second- and third-tier customers will naturally follow. Securing a foothold with the largest data centers is our top priority, and we’ve already made significant progress in that regard. The key challenge now is to wait for this to translate into real revenue generation, which we expect to happen next year. You’ll begin to see tangible results then.

So, next year, you believe that not only the Big Five but also the second- and third-tier customers will follow suit? Since they don’t operate at the same scale, how do you customize your solutions to meet their needs?

Yes, that’s correct, and that’s where our partners, the module makers, come in. For example, if Western Digital sells its solutions incorporating our controllers to a company like Meta, the second- and third-tier suppliers will naturally approach us to source similar solutions.

What are your plans for future development?

We’re closely monitoring emerging needs for innovative chips in data centers and AI. Our primary challenge is identifying and securing the right talent in Korea and globally. While we are confident in our expertise in chip development, we also need strong software talent to complement our hardware capabilities.

One interesting initiative we’ve undertaken is tapping into the global talent pool. For instance, I’ve been fascinated by how to identify top-tier talent. I’ve often referred to the International Olympiad in Informatics (IOI), which showcases the best young software engineers from around the world. To appear in the IOI Hall of Fame, a student must have won at least two medals, including a gold, demonstrating consistent excellence on the global stage.

We’re actively expanding our reach to capture global talent. Currently, we have offices in the U.S. and China are exploring further expansions to strengthen our talent acquisition strategy.

How would you describe working at FADU? Could you share a message for potential software talents about the benefits of working at FADU?

With the rise of AI and data centers, people are beginning to realize that software cannot stand alone. Chips are the fundamental building blocks. Only with the right chips can you develop software on top of them. Combining chip technology and software is more important than ever. At FADU, we are at the forefront of developing some of the most crucial chips in the industry. If you have the right software capabilities, you can truly thrive with our chip technology. This is the message we share with our software engineers: to build a system, you need the right chip, and that's where we come in. We are not focused on application software engineers; we are looking for system software engineers. These professionals need to collaborate with a leading chip company, and that's the environment we provide—it's the perfect playground for developing the next big thing.

Are there any specific messages you would like our readers at Newsweek to take away from your interview or this report?

I still believe in the importance of globalization, even though the US seems to be moving away from it. To me, this approach doesn’t make sense. The US cannot handle everything on its own; cooperation is essential. Specifically, Korea has a strong attitude, particularly in the semiconductor market, and we've proven that this mindset works well globally. Despite concerns over Korea's situation, I firmly believe that no one in the world can build a system without Korea's involvement. We're demonstrating this not only in the memory chip market but also in the logic chip market. In fact, we continue to prove that Korea can build the best chips for driving innovation in the industry. That’s the message we want to share—we will continue to prove ourselves by introducing innovative products and leveraging Korea's capabilities. Even in the future, when it comes to fabless companies, Korea cannot be overlooked. We are poised to cover a significant portion of the value chain and become a global leader in the fabless industry.

For more information visit: https://www.fadu.io/en/home-re/