Is China's AI investment boom a bubble?
Our answer is: "Parts of it are."
During the dot-com crash from 1999 to 2001, Amazon's revenue nearly doubled, while its share price fell by 94%. Yet from its 2001 low to today, its stock has delivered extraordinary long-term returns.
This illustrates one of our core convictions about investing through cycles: The price was a bubble. The industry was not. A market correction can coexist with a growing industry, and long-term value can continue to compound beyond short-term market volatility.
At HyT Capital's Private Salon in Singapore, Yelin Sun, Founding Partner of HyT Capital, opened his keynote with this question. He shared our perspectives on China's deep-tech investment cycles, the next wave of AI opportunities, and how to identify and capture structural alpha.
As AI capabilities advance rapidly, which critical barriers to industrial adoption are being cleared? And can China's distinctive industrial base turn this wave of technological progress into compounding advantages and long-term value creation?
01 When Technology Outpaces Adoption
Technology and adoption drive value creation. Capital and policy shape entry, financing and exit.
Distinguishing short-term market hype from long-term industrial opportunities requires understanding how four cycles—technology, industry, capital and policy—interact and move at different speeds.
Technology cycles typically unfold over 5–10 years, while industry cycles often take 10–20 years or longer to mature. Capital markets can experience multiple swings over much shorter periods, while policy cycles follow longer-term trends with periodic fluctuations.
These four cycles do not move in sync. The real challenge is to understand their interplay and identify when technological capabilities can translate into tangible industrial value.
Technology builds capability. Barriers decide when it scales.
The path and pace of value creation differ significantly across sectors. A common mistake is to assume that products in the physical world can scale at the same speed as those in the digital world.
ChatGPT reached 100 million users in just two months. But hardware and physical-world applications face more gates to scale. Unlike software, they must overcome real-world deployment barriers, including reliability, cost, production and delivery at scale.
Yet clearing a technological bottleneck does not automatically translate into commercial success.
At HyT Capital, we believe that a genuine industrial opportunity emerges when three elements come together: advancing technology capabilities, clearing barriers to deployment, and creating a step-change in customer value.
A company that addresses one or two of these elements can be a good company. But exceptional companies—those with the potential to become category leaders—bring all three together: advancing technology capabilities, clearing deployment barriers, and creating genuine 10x customer value.
02 Where RSI Meets China's Industrial Depth
The journey from technological advancement to industrial adoption can be long. But could an emerging force accelerate that process?
We believe Recursive Self-Improvement (RSI) is an important emerging variable that could accelerate the next wave of technological and industrial progress.
Strictly speaking, RSI refers to AI systems improving their own capabilities. In a broader sense, AI-assisted R&D, rapid experimentation and simulation could create accelerating improvement loops, likely delivering results in specific tasks first.
What happens when AI goes beyond helping people complete tasks and begins accelerating R&D, experimentation, simulation and iterative improvement itself?
The potential extends beyond AI coding and model R&D in the digital world. It could also accelerate progress in the physical world—from chip design and electronic design automation (EDA) to robotics development through simulation and real-world testing, and scientific discovery in materials, biology and chemistry.
This is where China's structural AI advantage comes into play.
RSI and broader AI-driven improvement loops rely on more than computing power. They also require well-defined tasks, rapid experiments or simulation, and real-world feedback where available. While models and algorithms influence the potential pace of AI advancement, real-world industrial environments provide critical conditions for validating and continuously improving those capabilities.
China has built a vast, highly integrated advanced manufacturing ecosystem, spanning electric vehicles and batteries, solar PV, semiconductor packaging and testing, and industrial robotics. These existing production lines, supply chains and real-world applications are not merely potential markets for AI. They could also provide the environments in which AI systems are tested, receive feedback and continuously improve.
Underpinning this industrial depth is China's deep bench of AI researchers and engineers. Large-scale simulation, parallel computing, engineering optimization and real-world deployment all depend on substantial engineering talent and execution capabilities.
As technological advances move into real-world applications and generate feedback for the next round of improvement, AI capability development, engineering validation and industrial adoption could form a mutually reinforcing cycle.
China's structural AI advantage lies not only in individual technological breakthroughs, but also in the compounding interplay between technology, engineering talent and its vast industrial ecosystem.
03 Identifying What Drives Scale
Spotting a technology trend is only the first step toward capturing an industrial opportunity.
As AI continues to expand the boundaries of what is possible, the more important questions are which critical bottlenecks are being cleared, and when those breakthroughs can translate into customer value at scale.
Embodied AI
The Door Has Just Opened
Before robots get smarter, they must last longer.
Embodied AI illustrates a divergence across cycles: capital-market expectations are rising, technology has yet to converge, and industrial adoption has only just begun. Despite growing capabilities, reliability, cost, thermal management, power efficiency and wiring wear remain critical challenges to deployment at scale.
This perspective is reflected in HyT Capital's portfolio positioning across embodied AI.
On the industrial adoption side, we focus on critical building blocks for robotics at scale: Hillstream (wireless motion control), Huake Cold Core (high-density cooling), LDRobot (cost-effective perception), and InnoScience (GaN power devices).
On the capability advancement side, we have backed Hillbot (world models for robotic reasoning) and pursued dedicated AI inference chips for robotics. We have also invested in task-first embodied AI teams, including MORPHI, AnyVerse, KunlunXing and XiaoYu, building capabilities through specific commercial tasks.
From critical components to the robot brain and commercial tasks, progress requires these capabilities to mature and work together. Scaling robotics is not just about what robots can do, but whether they can reliably and effectively perform tasks that customers actually need.
Optical Interconnects & Inference Silicon
Ahead of the Curve
Optical interconnects and inference silicon illustrate another type of investment opportunity: identifying critical technology components in an evolving industry and positioning before demand fully materializes.
Based on our understanding of these critical technology components and the industry's evolution, we built positions in Joywell, T-OnE and SILITH years ahead of the broader surge in demand.
A similar approach applies to AI inference silicon. As AI applications expand, growing inference and agentic workloads are driving demand for new computing architectures.
From our early investment in TheoEnd (agentic CPU) to LuminNX (inference AI chip), and more recently opportunities in dataflow architectures and specialized robot-brain chips, we continue to identify critical technologies as inference workloads evolve.
XR
Timing Matters
XR offers another lesson: timing matters.
In 2021, the metaverse narrative ran ahead of the devices and content needed to support it. The vision was compelling, but hardware limitations—including display size, weight, brightness and cost—along with an underdeveloped content ecosystem, constrained the everyday user experience.
By 2026, however, advances in display technology and AI video generation are beginning to clear bottlenecks on both the hardware and content sides, potentially changing the conditions for commercialization.
HyT Capital backed StarkSemi, which is working toward brighter, lighter and lower-cost displays for AR glasses, as well as Sand.ai, which focuses on AI video generation. We are also evaluating product teams capable of integrating these advances to deliver 10x customer value.
The same technology can face fundamentally different commercialization conditions at different stages of the industry cycle.
Across embodied AI, optical interconnects, inference silicon and XR, the pace of industrial adoption varies. Yet our investment assessment consistently comes back to three questions:
How far have technological capabilities advanced? When can the critical barriers to deployment be cleared? And when will meaningful customer value begin to materialize?
04 From Identifying Opportunities to Capturing Value
Seeing a trend is not the same as capturing its value.
As technology and industry continue to evolve, the real challenge is not only to anticipate where the next opportunities will emerge, but also to identify the founders capable of turning technological breakthroughs into scalable businesses.
At HyT Capital, our investment methodology is grounded in continuously tracking technological inflection points, industry maturity and the potential for customer value creation.
We first identify technology inflection points, assess the maturity of the industry and the deployment barriers that remain, and evaluate the potential to create 10x customer value. Together, these elements determine the Critical Success Factors (CSFs) for a sector. Once we identify those factors, we look for the founders best positioned to execute against them. We call this Founder Market Fit.
This means looking beyond technological excellence alone. We seek founders who understand the real constraints of industrial adoption, can navigate engineering and commercialization challenges, and ultimately translate technological capabilities into sustainable customer value.
Some valuations may be ahead of reality, while many customer needs and wants remain unmet. We invest where technology and the right founders can meet those needs at scale.
See & Capture.
SEE how technology, customer value and industrial adoption interact, and identify the opportunities emerging from that interplay.
CAPTURE those opportunities by backing exceptional founders who can overcome the barriers between technological breakthroughs and industrial adoption, turning them into scalable businesses.
Across market cycles, the value worth capturing is not the premium created by short-term market sentiment, but the enduring value generated when technological progress meets real industrial demand.
Understanding how value is created, backing founders who can turn it into scalable businesses, and capturing long-term opportunities where technology meets industry—this is what HyT Capital remains committed to.