AI Technology Application and Investment Opportunities | Linear Capital's Perspective

In the current era of rapid AI development, we need to pay attention not only to technological breakthroughs themselves, but also to how these technologies can be applied in practice to create real commercial value. I'd like to share my perspectives on understanding AI in this new era, the current state of applications, the advantages of open-source large models, China-US comparisons, and investment opportunities.

The Dual Breakthrough in Computing Power and Cost

The most distinctive feature of the new AI era is the dual breakthrough in computing power and cost. The advancement in computing power has been primarily led by the US, while the dramatic reduction in cost represents China's important contribution in the open-source large model wave.

Previously, the narrative around AI development focused mainly on pushing the limits of capability, with less attention paid to cost. However, the emergence of open-source large models has fundamentally changed this logic. It has driven costs down by 90%, not merely 30–40%, enabling AI applications to truly move from the realm of the elite to the masses.

While the US still leads in foundational AI innovation, China has tremendous opportunities in application deployment and cost optimization. Chinese companies' characteristics of being fast, efficient, and low-cost will be fully leveraged in this wave of AI applications.

From "Why Use It" to "Why Can't We Use It"

Corporate attitudes toward AI are undergoing a fundamental shift. Last year, most companies were still discussing "why should we use AI." This year, the focus has shifted to "why can't we get it to work." This change in mindset is significant — it shows that companies have recognized AI's value, yet still face challenges in practical implementation.

The main challenges include:

• Computing bottlenecks: High-end GPUs are difficult to obtain and prohibitively expensive. Although open-source models have reduced costs by 90%, building proprietary large model servers remains costly. This has also pushed companies to consider domestic computing solutions, fueling the growth of the domestic computing market.
• Data quality issues: Many traditional companies have accumulated massive amounts of data, but the quality varies widely. Data cleaning and organization are critical — simply purchasing hardware doesn't solve the application problem.
• Technical barriers: Different AI applications have different technical thresholds. Training custom models has the highest barrier, while building knowledge bases integrated with large models is relatively more accessible. The latter allows companies to combine internal data with large model capabilities, enabling AI to answer questions based on internal company circumstances.

Open-Source Large Models: The Perfect Balance of Performance and Cost

The emergence of open-source large models brings several key advantages:

• Performance-cost balance: Performance reaches 90–95% of top-tier closed-source models, but at only 5–10% of the cost. This cost-effectiveness makes AI applications affordable for more companies.
• Open-source transparency: A fully open strategy — including model weights, source code, and technical papers — allows companies to build their own models, ensuring data security without concerns about data leakage. Open-source exists at multiple levels: open data, open model weights, open code, and open papers. The more complete the open-source approach, the more it drives technological development.
• Engineering optimization: While there isn't much raw innovation, optimizing existing concepts to their extreme and integrating them effectively produces results far beyond expectations. This engineering capability happens to be a strength of Chinese companies.
• Distribution strategy: Careful design of release timing and methods can rapidly gain global attention for a company or product. By contrast, some large companies' open-source strategies aren't thorough enough — they only release parameters without open-sourcing code or detailed papers, so others learn the "form" but not the "spirit."

China-US AI Capability Comparison: Each Has Its Strengths

China and the US have taken different paths in AI development:

1. US: Focuses on breakthroughs in frontier innovation, sparing no expense to advance foundational capabilities and pursue exploratory leaps in capability.

2. China: Concentrates on cost optimization and cost-effectiveness, dedicated to making AI capabilities accessible to the masses.

In specific capabilities:

• The US leads in computing power and algorithms
• China has advantages in data processing and hardware supply chains
• In talent, the US has more top-tier researchers, while China has stronger engineering and productization talent

Notably, more high-end talent has begun considering returning to China to start businesses this year — a positive signal.

AI Applications: From Advisor to Executor

The efficiency gains from AI applications often far exceed expectations. For example, in information gathering, traditional methods might take two days, while using an AI Agent (tested with manus.im) can complete the same work in 10 minutes — a remarkable efficiency improvement.

More importantly, AI is transforming from a purely advisory role to that of an executor. Previous large language models were primarily "language in, language out" — like an advisor offering suggestions, but unable to execute specific tasks. Now AI can break complex tasks into multiple steps, calling upon different models or tools to complete each step, ultimately delivering a complete result.

This capability shifts AI from a mere conversational assistant to a genuine task executor, greatly expanding its application scenarios.

Robotics: The Next Investment Hotspot

In the large model space, the competitive landscape has already stabilized, with limited opportunities for new entrants. But the robotics field still holds enormous potential, divisible into three categories:

• Mobile robots: Focused on spatial movement, such as robotic dogs. The technology is relatively mature, with practical deployment expected to begin this year.
• Interactive robots: Concerned with robots' interaction with and manipulation of the physical world. This area has higher technical difficulty, requiring processing of multimodal data including vision, text, and action — with training complexity far exceeding that of language models. Practical deployment in industrial settings is expected within 2–3 years.
• Household robots: Targeting home and elder care scenarios. These robots remain at the demonstration stage, with practical deployment likely requiring 3–5 years, and more general-purpose products possibly needing 5+ years.

The technical challenge in robotics is that it must respect the laws of the physical world, with higher tolerance requirements. The "hallucination" problem in language models could have serious consequences in robotics, demanding higher standards for safety and accuracy.

AI-Era Entrepreneurs: New Traits and Opportunities

The AI era has changed what it demands of entrepreneurs:

• Technical understanding: While AI tools have lowered the barrier for understanding technology, entrepreneurs still need basic technical literacy. Fortunately, AI tools can now help parse technical papers, reducing the learning curve.
• Product refinement capability: The ability to translate technology into quality products is crucial. China has a cohort of "craftsmen" who are obsessively dedicated to product excellence — a valuable talent pool.
• Distribution capability: In an era of information overload, how to tell a compelling product story and design effective distribution paths matters equally.
• Fundraising capability: AI projects typically require substantial capital, so entrepreneurs need fundraising skills.
• Youth: The main force in current AI entrepreneurship consists of those born in the 1990s to 2000s, unburdened by traditional experience and more open to new thinking. Looking back at the founders of Hangzhou's "Six Little Dragons," most were born in the 1980s to 1990s when they started their companies. To invest in entrepreneurs who may shine 7–10 years from now, one should focus on those born in the 1990s to 2000s.

Investment Opportunity Summary

Opportunities in early-stage tech investment are concentrated in:

• Domestic computing: Domestic GPUs and computing solutions are expected to achieve dual breakthroughs in cost-effectiveness and performance within a few years.
• Inference technology and models: AI is shifting from training-centric to inference-centric, an area rich with opportunities.
• Robotics applications: Particularly in industrial, household, and special scenarios (such as elder care).
• AI + Science (bio, materials, chemical, etc.)
• Smart hardware: Truly intelligent hardware products, not merely conceptual ones.
• Autonomous driving: A new wave of technology is accelerating the deployment of autonomous driving.

Overall, China has unique advantages in AI application deployment and cost optimization, and is well-positioned to achieve greater success in productization and democratization. For investors, the key to capturing this wave of AI will be focusing on projects that can deeply integrate AI technology with real-world scenarios and deliver significant efficiency improvements.

About Linear Capital

Linear Capital is an early-stage investment institution focused on "frontier technology + industry," covering frontier technologies such as data intelligence, digital new infrastructure, next-generation robotics technology, and new technological transformations in traditional fields (such as biomedicine, materials, energy, etc.), applied across vertical industries to substantially improve industrial efficiency, empower them to solve pain points, and complete industrial upgrading — achieving excess returns through substantial increases in industrial value. It currently manages ten funds with total AUM of approximately $2 billion.

Our investment stage focuses primarily on leading angel to Series A rounds, with individual investments ranging from $1 million to $10 million (or RMB equivalent).

To date, we have invested in over 120 early-stage teams including Horizon Robotics, Kujiale, SensorsData, Tezign, Rokid, Guandata, Agile Robots, and others. The combined valuation of Linear Capital's portfolio companies is approximately $20 billion. In the near term, Linear Capital is working to become the premier "Data Intelligence Technology Fund," and in the long term, to gradually build itself into the most influential "Frontier Technology Application Fund."