AI startups are starting to "tackle the tough nuts".

Amid the waves of global competition and industrial upgrading, the artificial intelligence industry is no longer just a race of algorithms and models. Instead, it has become the core efficiency engine of technological transformation through deep integration with the real economy.

Against this backdrop, the "AI Gathering · Leading the New Era" 2025 Artificial Intelligence Innovation Roadshow Conference, hosted by BOE Technology Services and co - organized by Digital Friends Club, LIMO Zero - Second, ZODIAC Place, and 36Kr, successfully concluded in Beijing. To meet the requirements of the project for empowering the high - quality development of the park's science and technology association, under the guidance of the Innovation Service Center of the Beijing Association for Science and Technology and the Chaoyang District Association for Science and Technology in Beijing, this event brought together innovative enterprises, investment institutions, and industry experts from fields such as AI, industrial intelligence, and new materials, creating a high - value industrial docking platform.

As the host, a representative of BOE Technology Services said in the opening speech: "As a digital economy industry alliance initiated by BOE Technology Services, the Digital Friends Club is not just a platform. We prefer to define it as a community and hope to become industry partners in the digital transformation process."

In addition, each of the co - organizers of this roadshow conference has its own characteristics: The Digital Friends Club is committed to building an open and win - win industrial exchange and cooperation platform; LIMO Zero - Second has profound industrial incubation and cultivation resources; ZODIAC Place provides an ideal space and vertical industrial service platform for innovative enterprises; 36Kr provides all - round support for the event with its powerful media resources and innovation ecosystem. This lineup of co - organizers ensures a comprehensive improvement in the event's technological depth, resource docking, and communication influence.

This innovation platform, initiated by industry leaders and jointly built by multiple institutions, is accelerating the construction of an innovation ecosystem based on hard technology, connecting technology, capital, and the market.

The Cornerstone of Building the Industrial Ecosystem

In the first half of the roadshow, the participating projects all belong to one of the most important tracks in the technology field - new materials.

Lu Bin, the co - founder of Gallium Future, pointed out in his speech that the cost of substrates is more than 40 times that of silicon carbide, and there is a shortage of epitaxy. The key technologies and equipment are under embargo. What Gallium Future is doing is to abandon the mainstream technical route and adopt the "heteroepitaxy" technology to grow gallium oxide films on mature silicon carbide substrates. "Gallium oxide has high voltage resistance and low loss, making it an ideal material for the next - generation ultra - high - voltage power devices."

In the most traditional material R & D field, New Research Intelligent Materials promotes the transformation of the material R & D paradigm with "AI + materials". By using the material R & D intelligent agent, it breaks through the technical bottlenecks in the entire chain from laboratory R & D to large - scale mass production of materials. Nan Kai, its co - founder, proposed at the event site that through the L3 intelligent agent, New Research Intelligent Materials can increase the efficiency of enterprise material R & D by 70%, which can also solve industry problems such as over - reliance on manual labor, long cycles, and knowledge isolation in the laboratory stage.

Zhao Huibo, a director of Multidimensional Capital, also gave some reasonable advice. The cycle of new materials is very long, and it is best to first break through a potential large - scale track.

When AI Deeply Integrates with Scenarios

Currently, the clearest trend is that AI is changing from a "technical tool" to an integrated "ecosystem component". When the technical foundation is strong enough, AI can be better and more deeply integrated with technical scenarios.

As the first commercially landed industrial embodied intelligent system in China, Hyperion has achieved mass production. Liu Quanshan, the CFO of Hyperion, proposed at the roadshow site that the market in the industrial field has always existed. Once the technology makes a breakthrough, the market will be immediately released.

In the software and model layer, the industry is more concerned about the issue of "trustworthiness". Gao Li, the vice - president of Color Intelligence Technology, pointed out that there is a lack of verification of the cornerstone of "charter knowledge" in general large models. "Color Intelligence Technology has developed a trustworthy knowledge model and has obtained double filings from the National Internet Information Office, solving the problems of inaccuracy and hallucination in enterprise intelligent agents."

After solving the trustworthiness problem, the next question is how AI can become an enterprise's "business partner". The answer from Yue Dian Technology is to deeply integrate enterprise private data with AI technology, build a trustworthy enterprise - level Data + AI development platform, and train an Ontology - base reasoning model with enterprise private data to achieve business process reengineering and efficiency multiplication.

Exploring the Future of Artificial Intelligence with Hardware - Software Integration

Finally, when AI technology tries to reach a broader consumer market, all entrepreneurs must answer a question: How can they achieve the "daring leap" from technology to products?

As a hardware - software integrated "AI shop assistant" designed for mom - and - pop stores, Yunjinwei integrates functions such as security, business analysis, and intelligent product selection on a computing base station. Wang Wenyi, the CEO of Yunjinwei, said that high - frequency computing and long - term memory technologies will bring a value revolution to smart retail.

Investors also shared their observations from the perspective of market rules. Gao Zedong, an investment director of Delian Capital, analyzed in a round - table discussion: "In essence, users will calculate the incremental value or core selling points provided by enterprises and how much more they need to pay. The arrival of AI will not completely subvert all hardware in a short period."

In the final round - table session of the event, Gao Zedong also said that multi - modal large models provide new supply - side changes in aspects such as perception, planning, interaction, generation, and execution. At the same time, the computing power at the edge is also developing rapidly, accelerating the application of models at the edge. The continuous maturity and combination of model capabilities, along with the upgrade of edge computing power, continuously give birth to new hardware. "In the AI era, hardware will liberate people more, allowing them to focus on their physical and mental health, return to real social relationships, and discover the beauty in life."

This also indicates that perhaps the next wave of AI application explosions will occur in countless ordinary scenarios around us. Under this practical trend, AI projects that can integrate into the industrial ecosystem and quickly verify the business closed - loop are more likely to survive the cycle and become real industrial transformers.

This article is from the WeChat official account "36Kr". 36Kr published it with authorization.