Chat with this content: Summary, Analysis, News...
Share on Reddit
Alchip Technologies: The Silent Architect of the AI ASIC Revolution
I. Introduction & Episode Roadmap
Picture this moment: On Wednesday, Alchip Technologies reported a swing into profit last year after reporting net income of NT$308 million (US$10.58 million), compared with a net loss of NT$211 million a year earlier. The year was 2018, and Taiwan's chip design houses were riding the cryptocurrency boom that had sent Bitcoin soaring 1,500%. Yet few outside the semiconductor industry had heard of this company that would, just six years later, have a market capitalization of 303.73 billion TWD—roughly $10 billion USD—making it Taiwan's most valuable IC design house by market cap.
The paradox that defines Alchip Technologies revealed itself starkly on August 13, 2024. The company reported second-quarter revenue down 29.4% year-on-year, its steepest decline in recent memory. In a sector where AI peers were posting growth rates of 50% or even 100%, the result should have been catastrophic. Yet Morgan Stanley published a bullish note that same day, raising its price target from NT$4,088 to NT$4,588. The stock jumped 8% the following day.
How does a company hemorrhage revenue while analysts cheer? How did a Taiwanese chip design firm founded by Silicon Valley basketball buddies become essential to Amazon's AI ambitions? And why, in an era when NVIDIA commands headlines and trillion-dollar valuations, does this $10 billion company remain virtually unknown outside semiconductor circles?
The answer lies in understanding what Alchip really is: not a chip company in the traditional sense, but the silent architect behind the custom silicon revolution that's reshaping artificial intelligence. While NVIDIA sells the same GPUs to everyone, Alchip helps tech giants build their own custom AI chips—a shift that could fundamentally alter the balance of power in the semiconductor industry.
II. Silicon Valley Origins & The PlayStation Connection
The basketball court at a Silicon Valley gym in 2002 seemed an unlikely birthplace for a semiconductor empire. Alchip was founded in 2003 by a group of Silicon Valley veterans that followed a similar path of working for semiconductor companies then moving to the EDA/IP/ASIC ecosystem. In fact, I used to play basketball with the Alchip co-founder/chairman during that time and can tell you he is a fierce competitor.
That fierce competitor was Kinying Kwan, who along with Johnny Shen had just watched their previous company disappear into corporate consolidation. In April 2002 Cadence acquired Simplex Solutions, an ASIC design services company. Alchip was founded six months later by Kinying Kwan and other former Simplex employees as a fabless ASIC supplier. But the team's history went back even further. Alchip's co-founders, Chairman of the Board Kinying Kwan and President and CEO Johnny Shen, both came from Silicon Valley company Altius, specializing in chip design services. Their claim to fame was securing the backend design for Sony's PlayStation 2 chip, which accounted for 80 percent of Sony's revenue and shrunk the chip's size by 30 percent.
The PlayStation 2 connection wasn't just a resume line—it shaped everything about Alchip's approach. Simplex Solutions had designed the graphics ASIC for Sony's PlayStation 2 game console. Sony became an important customer of Alchip as well. In 2006 80% of Alchip's revenues were from Japan, and most of that was from Sony. This wasn't the glamorous work of designing cutting-edge architectures or breakthrough algorithms. It was the grueling, detail-oriented work of physical implementation—taking someone else's grand vision and actually making it manufacturable.
The founders made a counterintuitive bet in 2003. While the mobile revolution was beginning and most ASIC companies chased small, power-efficient chips for phones, Since its establishment, Alchip has focused on ASICs for network communications and high-performance computing, targeting advanced processes. However, before the rise of Bitcoin mining and AI in 2018, the market's mainstream was dominated by small-scale ASICs for mobile devices, and Alchip's revenue remained stagnant at around NT$4 billion.
At the time, a senior figure in the IC design industry reminded Johnny Shen that a chip priced above US$10 would be difficult to sell. "But at that time, the large ASICs we were designing cost a few hundred dollars each," recalls Shen. The industry wisdom said they were crazy. They persisted anyway.
III. The Quiet Years: Building Technical Excellence (2003-2017)
For fourteen years, Alchip operated in the shadows of the semiconductor industry, perfecting a craft that few understood and even fewer valued. The company went public on Taiwan's Emergent Market in 2010, then graduated to the main Taiwan Stock Exchange in 2014, but remained largely unknown outside specialist circles.
What were they doing during these quiet years? Building something that would prove invaluable: deep, almost mystical expertise in advanced packaging and complex chip implementation. The company was founded by semiconductor veterans from Silicon Valley and Japan in 2003 and provides faster time-to-market and cost-effective solutions for SoC design at mainstream and advanced, including 7nm processes.
The numbers tell a story of relentless technical advancement. By 2020, the company could claim something remarkable: Some our more important technology milestones include completing 16nm and 12nm AI and automotive devices over the past 2-years and we will soon announce the successful tape-oust of sub-10nm designs. But even more impressive was their batting average. In an industry where first-time silicon success rates hover around 30%, Alchip achieved something close to perfection.
Their secret weapon wasn't any single technology but rather their relationship with TSMC, the world's dominant contract chip manufacturer. Over the past decade, Alchip has received investments from several tech heavyweights, including Global Future Group, Investar, AcerVC, Cisco Systems, C2Capital, and notably Taiwan Semiconductor Manufacturing Company (TSMC), the biggest contract chipmaker in the world, owns a 20% stake in the firm. As of 2014, over 85% of Alchip projects are outsourced to TSMC.
One project from this era stands out as a harbinger of what was to come. In 2009, IBM joined forces with the University of Tokyo to introduce the world's fastest supercomputer using TSMC's 90nm process. It was later revealed by the media that Alchip was responsible for the design and successful production of the chip. "After that, as long as supercomputing chips were needed, everyone in Japan, China, and even the US came to us," recalls Johnny Shen.
The supercomputer project taught Alchip something crucial: how to handle massive chips with extreme power requirements and complex packaging needs. These weren't the elegant, minimalist designs that won awards at conferences. These were industrial-strength implementations that had to work perfectly the first time, every time.
IV. The Bitcoin Inflection Point (2017-2018)
The cryptocurrency boom of 2017-2018 changed everything for Alchip, though not in the way most people think. Bitcoin's 1,500 percent rally last year led to a robust demand for the vast computing power used to verify transactions of the cryptocurrency, which required application-specific integrated circuits (ASICs) that are designed just for bitcoin mining.
Some mining companies last year outsourced part of their ASIC orders to Alchip and Global Unichip, Sanford C Bernstein & Co said in November. The work itself wasn't particularly glamorous—Bitcoin mining ASICs are essentially the semiconductor equivalent of sledgehammers, designed to do one thing with maximum efficiency. But they required exactly the skills Alchip had been building: advanced process implementation, high-power design, and the ability to move from concept to production at breakneck speed.
Subsequently, bitcoin vendors also sought out Alchip. The financial impact was immediate and dramatic. After years of stagnant revenue around NT$4 billion, the company suddenly found itself swimming in orders. The 2018 swing to profitability that opened this story was just the beginning.
But the real value of the Bitcoin boom wasn't the revenue—it was the capabilities Alchip developed. Bitcoin mining chips pushed the boundaries of what was possible in terms of power consumption and thermal management. A mining ASIC could consume 3,000 watts or more, generating heat that could literally melt conventional packaging. Alchip learned to handle these extreme conditions, developing expertise in advanced packaging technologies that would soon prove invaluable.
The company also learned something about market dynamics. The Bitcoin miners were willing to pay premium prices for faster time-to-market. Every day of delay meant millions in lost mining revenue. This taught Alchip to optimize not just for technical excellence but for speed—a combination that would make them irresistible to hyperscalers racing to build AI infrastructure.
And top Chinese AI companies SenseTime and Cambricon also produced chips with design assistance from Alchip. The company was building a reputation as the go-to partner for anyone who needed advanced chips built quickly and reliably.
V. The AI Revolution & Hyperscaler Wars (2019-2023)
The transition from Bitcoin to AI wasn't a pivot—it was an evolution. The same capabilities that made Alchip invaluable to crypto miners made them essential to cloud giants building custom AI silicon. Fiscal year 2023 revenue, ending December 31, reached a record $978.4 million, a 112.5% increase over FY 2022 revenue of $460.5 million. Operating income for FY 2023 achieved a record $120.7 million, an increase of 55.7% over 2022 operating income of $77.5 million. FY 2023 net income reached $106.6 million a 73.3% increase over FY 2022 net income of $61.5 million.
The transformation was staggering. Alchip specializes in the design services of large-scale AI high-performance computing chips and is one of the major users of TSMC's CoWoS production line. It competes with such chip giants as Nvidia, AMD, and Broadcom for production capacity. Last year, Alchip's contribution to high-performance computing solutions accounted for over 80 percent of its revenue, of which advanced processes below 7nm accounted for 68 percent. This year, AI-related businesses are expected to account for 60 percent of Alchip's revenue, doubling over last year.
But the real story was landing Amazon Web Services as an anchor customer. The relationship transformed both companies. It turns out Amazon is working on just that, partnering with ASIC design company Alchip to build custom chips for training and inference with the self-explanatory names Trainium and Inferentia.
The numbers tell only part of the story. On a geographic basis, revenue from North America experienced a huge year-on-year jump. North America accounted for 63% of all 2023 revenue, up significantly from its contribution of 39% in 2022. What those numbers don't capture is the strategic importance of the work. Amazon wasn't just buying chips—they were buying independence from NVIDIA's monopoly.
President and CEO Johnny Shen, in his remarks, identified strong AI chip shipments to the company's largest North American customer as the major driver of 2023 revenue growth. He also pointed out that 70 to 80% of 2023 total revenue was production sales, with NRE (Non-Recurring Engineering) revenue accounting 20 to 30% of the 2023 total.
The AWS relationship revealed something fundamental about Alchip's business model. Unlike traditional chip companies that design products and sell them to multiple customers, Alchip was essentially a mercenary—but a mercenary with PhD-level expertise and billion-dollar fabrication relationships. They didn't own the designs they implemented, but they owned something potentially more valuable: the knowledge of how to turn those designs into reality.
By 2024, the company's dependence on AWS had become both its greatest strength and most obvious vulnerability. According to company filings, Customer A—widely understood to be AWS—accounted for 60.16% of revenue, or NT$31.27 billion. This concentration would typically terrify investors, but in Alchip's case, it represented something different: a partnership so deep and strategic that disentangling it would be almost impossible.
VI. The Advanced Packaging Game-Changer: CoWoS
If there's one technology that explains Alchip's rise from obscurity to indispensability, it's CoWoS—Chip on Wafer on Substrate. According to Alchip Technologies' President and CEO, Johnny Shen, "Packaging is the new 'Moore's Law' for powerful high-performance computing challenges. Understanding and applying the technology is critical to meeting the demand for more functionality and greater performance in a smaller physical footprint.
To understand why CoWoS matters, imagine trying to build a skyscraper. Traditional chip packaging is like building single-story houses and connecting them with roads. CoWoS is like stacking floors vertically with high-speed elevators between them. TSMC's Chip on Wafer on Substrate with Silicon Interposer (CoWoS-S) provides best-in-class package technology for ultra-high performance computing applications, such as artificial intelligence (AI) and supercomputing. Wafer-level system integration provides high-density interconnects and deep trench capacitors over a large silicon interposer area to accommodate various functional top die, including logic chiplets, with high-bandwidth memory (HBM) cubes stacked over it.
The technology isn't just complex—it's capacity-constrained. NVIDIA dominates the capacity demand, accounting for a significant 63% of the total demand, indicating its leadership position in adopting CoWoS technology. Broadcom follows at 13%, marking it as the second-largest contributor to CoWoS demand, albeit far behind NVIDIA. AMD and Marvell each share 8%, tying for the third spot, suggesting comparable interest in this technology from both companies. Other contributors, including AWS + Alchip (3%), Intel (2%), Xilinx (1%), and Others (3%), account for much smaller shares.
That 3% allocation for AWS and Alchip might seem small, but it represents something crucial: guaranteed access to the most coveted manufacturing capacity in the semiconductor industry. Alchip engineers also detailed the company's record-setting advanced package production, including pioneering CoWoS designs on advanced nodes, power levels exceeding 1000W, more than 3.3x reticle interposers, and larger than 70x80mm² packages.
The company's CoWoS expertise didn't develop overnight. Our leading edge CoWoS service platform is scoped to cover from system planning, interposer design, test, qualification and to production." This end-to-end capability is rare. Most companies might excel at one part of the process, but Alchip had mastered the entire chain from design through production.
By 2024, the CoWoS shortage had become the semiconductor industry's biggest bottleneck. As things stand, TSMC is just barely meeting the current demand for this packaging method – never mind future demand – which is why last year the company announced plans to more than double CoWoS capacity by the end of 2024. Having secured capacity years in advance, Alchip found itself in an enviable position—able to deliver what money alone couldn't buy.
VII. The Trainium Saga & Battle with Marvell
The story of Amazon's Trainium chips reads like a corporate thriller, complete with betrayal, redemption, and billions of dollars at stake. The first-generation Trainium was Alchip's masterpiece, a chip that proved AWS could compete with NVIDIA in AI training. Alchip is reportedly designing the new 3-nanometer chip, called Trainium 3, for Amazon Web Services. Alchip customizes chips for customers and also serves as a coordinator for customers to produce chips at foundry service providers such as Taiwan Semiconductor Manufacturing Co.
Then came the shock. The second-generation Trainium, expected to cement Alchip's position, went to Marvell instead. The loss wasn't just financial—it was existential. With 60% of revenue from a single customer, losing AWS would have been catastrophic.
In 2022, AWS released its Trainium1 and Inferentia2 chips. The Trainium1 chip and Inferentia2 chips are nearly the same, except that the Inferentia2 chip only has two Neuronlink-v2 interconnect ports vs the Trainium1's four ports. ... Trainium1/Inferentia2 has been underwhelming for GenAI training due to uncompetitive scale-up and scale-out networking, with many software bugs also detracting from customers' workloads. As such, Trainium1/Inferentia2 has instead been used for training non-complex non-GenAI internal Amazon workloads such as credit card fraud detection AI models as well as for inferencing for Anthropic and Amazon internal workloads. Ironically, Trainium1 is better for GenAI inferencing than training. Internally, Amazon has also been using Inferentia2 for inferencing such as on Prime Day 2024 with over 80k Inferentia2/Trainium1 chips being used to power an ML based assistant for Amazon helping Amazon.com prime members.
The second-generation loss revealed uncomfortable truths about the ASIC business. Technical excellence wasn't enough. Marvell brought something Alchip couldn't match: American headquarters, existing relationships with other hyperscalers, and perhaps most importantly, no dependency on TSMC's contested Taiwan facilities.
But the story didn't end there. Industry sources suggest Alchip has already secured both the third and fourth-generation Trainium projects. With 3-nanometer chips ready to be shipped, Alchip has started development of 2-nanometer chips, Shen said. Building on its success in offering 3-nanometer chip design services, the company is confident about providing 2-nanometer chips to the same customers, he told investors earlier this month.
The comeback reveals something about competitive dynamics in custom silicon. Switching ASIC partners isn't like changing suppliers for memory chips or standard components. The knowledge accumulated over years of collaboration—understanding of design quirks, optimization techniques, manufacturing relationships—creates switching costs that go beyond mere economics.
VIII. The NVIDIA Partnership Surprise
May 18, 2025, marked a watershed moment that few saw coming. MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys and Cadence are among the first to adopt NVLink Fusion, enabling custom silicon scale-up to meet the requirements of demanding workloads for model training and agentic AI inference. NVIDIA today unveiled NVIDIA NVLink Fusion™ — new silicon that lets industries build semi-custom AI infrastructure with the vast ecosystem of partners building with NVIDIA NVLink™, the world's most advanced and widely adopted computing fabric.
The announcement seemed paradoxical. Why would NVIDIA, the undisputed king of AI chips, help others build custom alternatives? "A tectonic shift is underway: for the first time in decades, data centers must be fundamentally rearchitected — AI is being fused into every computing platform," said Jensen Huang, founder and CEO of Nvidia. "NVLink Fusion opens Nvidia's AI platform and rich ecosystem for partners to build specialized AI infrastructures."
For Alchip, inclusion in this exclusive club represented validation at the highest level. "Alchip is supporting adoption of NVLink Fusion by broadening its availability through a design and manufacturing ecosystem, encompassing advanced processes and proven packaging and supported by the ASIC industry's most flexible engagement," said Johnny Shen, CEO of Alchip. "It's our contribution to ensuring that the next generation of AI models can be trained and deployed efficiently to meet the demands of tomorrow's intelligent applications."
The technical implications were profound. Alchip's silicon-proven 2.5D and 3D design flow, part of the NVLink Fusion ecosystem, optimizes essential flows such as power delivery, Die-to-Die electrical interconnect, and system-wide thermal characterization. The new 3DIC design flow's power delivery module includes power integrity, power grid design (with through-silicon via distribution), and power integrity simulation and sign-off capabilities. Die-to-Die electrical interconnect capabilities address low clock skew across dies, process variation immunity, noise immunity, data transmission across different power domains, and inter-die setup/hold timing margin. The design flow also covers thermal characterization to enhance power density, perform 3D non-uniform power mapping, mitigate 3D thermal crosstalk effects, and model package and system cooling solutions.
But beneath the technical specifications lay a more strategic reality. Technologies like Nvidia's NVLink typically use multiple protocols, including physical layer protocol (PHY), data link layer protocol, transport protocol, memory coherence protocol, atomic operation protocol, and NUMA protocol. While companies like Alchip can access everything related to hardware, the software layer responsible for initiating and configuring links is still controlled by Nvidia.
NVIDIA wasn't opening its kingdom—it was creating carefully controlled duchies. By allowing companies like Alchip to build NVLink-compatible chips, NVIDIA ensures that even custom silicon remains within its ecosystem. It's a brilliant strategy: embrace the inevitable move to custom chips while maintaining architectural control.
IX. The Business Model & Competitive Dynamics
Understanding Alchip's business model requires throwing out everything you know about traditional semiconductor companies. They don't design chips. They don't own fabs. They don't even own most of the IP they implement. So what exactly do they do?
"When hyperscalers are looking for ASIC providers in the market, they don't have too many choices," said Daniel Wang, Alchip's chief financial officer, at its first quarter investor conference in early May. "Broadcom, Marvell, Alchip, MediaTek, GUC [Global Unichip], probably you can only have these five [that] can provide the most leading-edge process node and the large-scale design ASIC to date."
The competitive landscape reveals why Alchip's position is both precarious and powerful. Broadcom dominates with 55-60% market share, offering full-service ASIC solutions to the biggest players. Marvell claims 13-15%, leveraging its networking heritage and American headquarters. Alchip, despite its smaller size, has carved out a unique niche through technical excellence and TSMC relationships.
The gross margin story tells everything about Alchip's strategy. While NVIDIA enjoys gross margins above 70% and Broadcom maintains margins around 65%, Alchip operates at 15-20%. This isn't a weakness—it's a feature. Our design platform also offers a knowledge-based design flow for different applications in a manageable QoR range at each design steps to achieve superior power, performance and area witin a controllable design cycle for large scale designs. This translates into faster time-to-market and ensure one-pass silicon success.
The pricing philosophy reflects a fundamental truth about the ASIC business: the value isn't in the margin on the chip, it's in enabling the customer to compete. When AWS saves hundreds of millions by using custom chips instead of NVIDIA GPUs, paying Alchip's fees is a rounding error.
One Taiwanese IC design executive added, "TSMC usually doesn't engage directly with customers. Another company usually has to act as an intermediary to manage the process." Taiwanese companies, with their ability to handle everything from chip design to manufacturing and packaging domestically, are highly competitive and ideally positioned for that role in this AI chip boom, the executive said.
The geographic advantage can't be overstated. Being in Taiwan, literally next door to TSMC, means Alchip engineers can be in the fab within hours if problems arise. They speak the same language—not just Mandarin, but the specialized dialect of semiconductor manufacturing that takes years to master.
X. Current State & Future Trajectory
As October 2025 draws to a close, Alchip stands at an inflection point. Fiscal year 2024 revenue, ending December 31, reached a record $1.62 billion, a 65.4% increase over FY 2023 revenue of $978.4 million. Operating income for FY 2024 achieved a record $202.3 million, an increase of 67.6% over 2023 operating income of $120.7 million. FY 2024 net income reached $200.8 million, an 88.4% increase over FY 2023 net income of $106.6 million. Earnings per share for the FY 2024 achieved a record NTD81.3, exceeding last year's previous record of NTD45.5.
The technology roadmap points to continued leadership. "This is a must-win [project]," Shen said yesterday. The company isn't just working on current generation technology—they're already deep into 2nm development, staying ahead of the curve that most competitors are still trying to reach.
On a process node basis, advanced technologies, defined as devices manufactured at 7nm-and-below technologies, accounted for 96% of 2024 revenue, up from 89% in 2023. Devices manufactured on 16nm and 12nm technologies contributed 2% to 2024 revenue, while 28nm-and-above technologies accounted for the remaining 2%.
The company's evolution beyond AWS dependency is perhaps the most crucial development. Last year, Amazon.com Inc subscribed to 224,537 shares, or 0.28 percent, of Alchip for NT$535 million through a private placement. While the stake is small, it represents something larger: a deepening of the partnership from vendor-customer to strategic alliance.
The expansion into new geographies tells another story. Opens Vietnam Center, Expands Japan ResourcesAlchip Technologies, the High-Performance ASIC leader, is expanding its global engineering footprint with the launch of new design centers in Vietnam and the expansion of its Japan Design Center. These strategic investments underscore Alchip's commitment to scaling its high-performance computing (HPC) design capabilities and meeting growing demand at leading-edge nodes, including 3nm.
But perhaps the most telling indicator of Alchip's future is its board composition. While, 4 new independent directors were added to the chip company's 7-member board including Jerry Tzou, Director in Advanced Packaging Business Development of TSMC. Having TSMC representation at the board level isn't just about governance—it's about strategic alignment at the highest levels.
XI. Playbook: Lessons for Founders & Investors
The Alchip story offers contrarian lessons for anyone building or investing in deep tech companies. First, the power of technical depth over breadth. While competitors tried to be everything to everyone, Alchip focused relentlessly on one thing: making the most complex chips manufacturable. They didn't try to design revolutionary architectures or create new categories. They became the world's best at implementation.
Second, the art of riding technology waves without drowning in them. PlayStation 2 to Bitcoin to AI—each transition could have killed the company. Instead, each wave left them stronger, with new capabilities that made them indispensable for the next wave. The key wasn't predicting the waves but building capabilities that transcended specific applications.
Third, the value of being boring infrastructure. Alchip will never have NVIDIA's brand recognition or Apple's consumer cachet. But being invisible has advantages. When you're infrastructure, you're essential. When you're essential, you have pricing power that transcends traditional margins.
Fourth, the Taiwan semiconductor ecosystem advantage can't be replicated. It's not just about proximity to TSMC—it's about decades of accumulated knowledge, relationships, and trust. Over the past decade, Alchip has received investments from several tech heavyweights, including Global Future Group, Investar, AcerVC, Cisco Systems, C2Capital, and notably Taiwan Semiconductor Manufacturing Company (TSMC), the biggest contract chipmaker in the world, owns a 20% stake in the firm.
Fifth, managing customer concentration risk by becoming indispensable. Traditional wisdom says having 60% of revenue from one customer is dangerous. But when switching costs are measured in years and billions of dollars, concentration becomes a moat, not a risk.
Finally, the power of aligning with secular trends rather than fighting them. Alchip didn't try to compete with NVIDIA—they enabled NVIDIA's customers to build alternatives. They didn't fight the move to custom silicon—they became the arms dealer for the custom silicon revolution.
XII. Bear vs. Bull Case
Bear Case:
The bear thesis writes itself: extreme customer concentration coupled with geopolitical risk equals potential catastrophe. If AWS decides to bring ASIC design in-house or switch permanently to Marvell, Alchip loses 60% of revenue overnight. The company's attempts at diversification have been modest at best, with other customers representing fragments compared to the AWS whale.
The competitive landscape is intensifying. Broadcom and Marvell aren't standing still, and both have advantages Alchip can't match—U.S. headquarters, broader IP portfolios, and relationships with multiple hyperscalers. As custom ASICs become more strategic, customers might prefer working with larger, more diversified partners.
The geopolitical risk looms largest. In April 2021 the US Government blacklisted seven Chinese supercomputing companies due to alleged involvement in supplying equipment to the PLA, Chinese military–industrial complex, and WMD programs. In response Alchip and TSMC suspended new orders from Chinese supercomputing company Tianjin Phytium Information Technology. Phytium accounted for 39% of Alchip's revenue. If U.S.-China tensions escalate, Alchip could find itself caught in the crossfire again.
The technology risk is real too. As chiplets and advanced packaging become commoditized, Alchip's competitive advantage could erode. TSMC is expanding CoWoS capacity aggressively, potentially reducing the scarcity value of Alchip's secured allocations.
Bull Case:
The bull thesis rests on one unassailable fact: the world needs more custom AI chips, and very few companies can actually build them. President and CEO Johnny Shen credited strong 7nm AI chip shipments to North America for strong 2024 revenue growth. He also cited continuous improvement in operation efficiency and improved non-operating income, primarily higher interest income, for 2024 net income growth exceeding 2024 revenue growth.
The AWS relationship, rather than being a weakness, represents an insurmountable moat. The knowledge accumulated over years of collaboration—understanding AWS's design philosophy, optimization for their specific workloads, integration with their infrastructure—can't be replicated quickly. The fact that Alchip apparently won back Trainium 3 and 4 after losing Trainium 2 proves this point.
The NVIDIA partnership opens entirely new markets. Being one of the chosen few for NVLink Fusion positions Alchip at the center of the semi-custom revolution. Every company that wants to build custom chips compatible with NVIDIA's ecosystem becomes a potential customer.
The secular trend toward custom silicon is unstoppable. As AI workloads become more specialized, general-purpose GPUs become less efficient. McKinsey predicts that by 2030, the majority of AI workloads will run on ASICs. Alchip is positioned to capture a significant share of this transition.
The valuation remains compelling despite the recent run-up. At $10 billion market cap on $1.6 billion revenue growing 65% annually, Alchip trades at a fraction of the multiples assigned to AI semiconductor darlings. If the company can maintain growth while gradually diversifying its customer base, the stock could easily double from current levels.
XIII. Epilogue: The Power of Being Essential
The basketball court where Kinying Kwan once battled colleagues has given way to a different kind of competition—one where the stakes are measured in billions and the game is about enabling others' ambitions rather than showcasing your own. From shrinking PlayStation 2 chips by 30% to enabling Amazon to challenge NVIDIA's AI dominance, Alchip's journey illustrates a profound truth about technology markets: sometimes the most valuable position isn't at the center of attention but in the shadows, making everyone else's dreams possible.
The company that started with refugee engineers from a failed acquisition has become something far more significant than its founders likely imagined. They're not just designing chips—they're enabling the democratization of AI infrastructure. Every custom chip that reduces dependence on NVIDIA, every hyperscaler that gains silicon independence, every breakthrough in AI training efficiency—behind many of these advances stands Alchip, silent and essential.
The future of computing increasingly depends on custom silicon. As AI models grow from billions to trillions of parameters, as inference moves from cloud to edge, as new architectures emerge that we can't yet imagine, the need for companies that can translate vision into silicon will only grow. In this future, Alchip's invisibility isn't a bug—it's the ultimate feature.
Taiwan's role in global semiconductor dominance often focuses on TSMC's manufacturing prowess. But Alchip represents something equally important: the design expertise and implementation knowledge that turns cutting-edge process technology into working products. Without companies like Alchip, TSMC's advanced nodes would be like Formula 1 racetracks with no one who knows how to build the cars.
As we stand on the precipice of an AI-transformed world, the Alchip story reminds us that revolutions require more than visionaries and breakthrough technologies. They require the patient builders, the careful implementers, the companies willing to do the hard, unglamorous work of turning dreams into silicon. In the end, being essential beats being famous. And in the semiconductor industry's great game, Alchip has quietly positioned itself as a piece without which the board cannot function.
The stock market, in its occasional wisdom, has begun to recognize this value. But the real judgment will come from history. When we look back at how artificial intelligence reshaped computing, how custom silicon broke the GPU monopoly, how the balance of technological power shifted in the 2020s, we'll find Alchip's fingerprints everywhere—invisible to most, essential to all.
StockTwits
Spotify
Apple Podcasts
Amazon Music
Audible
YouTube
Castbox
Pocket Casts