AI Commercialization: How Companies are Cashing in on the AI Revolution

It has taken decades for AI to become the investment honeypot it now is. The concept was first proposed in a meeting in Dartmouth College in 1956. Now well into the 21st century, a wave of computing breakthroughs and big data storage advances have accelerated the development of AI to the cusp of mainstream commercial application.

Success in vertical supply chain integration and automation showed us the power of artificial intelligence in data-driven efficiency. Customer services are another example. AI can handle various customer inquiries at the same time. Public safety departments are using AI to facially identify suspects. In manufacturing, Tesla’s factories exemplify the role of AI, as robots handle battery assembly and welding. Computer vision monitors production in real time, detecting and correcting defects as they occur. Predictive maintenance prevents equipment failures, dramatically reducing downtime.

Capital markets are optimistic, filled with investors betting on AI’s transformative potential.

• In 2024, Databricks, a US data lake platform, raised $10 billion.
• OpenAI completed a $6.6 billion financing in October 2024, with a valuation of $157 billion.
• In China’s AIGC industry, data shows 84 financing events in the third quarter of 2024, totalling over 10.5 billion RMB, mostly in the 200–500 million RMB range, averaging around 260 million RMB per deal.

But investment is not just in the efficiency premium. In an industry ecology in which humans and AI applications fight out their differences, investment avenues have expanded to ease the process.

Take manufacturing for example. As AI and human production lines need different lighting conditions and conveyor belt speeds, companies have to balance their human staff with models that are gaining versatility and adaptability via self-generation. AI is getting better at adapting to the needs of different industrial scenarios but needs unified industry standards and facilitation in bridging data islands. For this, industry needs open technology to promote data standardization and data sharing, and more in-depth cooperation between equipment manufacturers, software developers and end users.

Between late 2022 and mid-2024, over 78,000 AI-related companies were registered in China according to Titanium Media. High valuations of companies like Zhipu AI, Moonshot, and Baichuan Intelligence—each exceeding 20 billion RMB—highlight the concentration of resources. Although many have since suspended operations, this resource concentration has brought money into the game. The top firms are investing more in R&D and being rewarded by market growth. Conversely, smaller and medium-sized enterprises face difficulties in accessing capital, innovation, and certification. This risks reducing technological diversity and slowing progress overall.

A “Sword of Damocles” hanging over AI companies: data

For early AI firms, balancing data needs with privacy risks has been tricky. Data is essential for training algorithms, and obtaining it legally is costly. Some startups have resorted to questionable means—hidden clauses, data crawlers, or grey transactions—to amass data.

According to Cyberhaven, in 2024, employee uploads of sensitive data to AI tools surged by 485%, with more than 2 million such uploads per 100,000 employees. This exposes companies to enormous privacy risks.

The industry faces a dilemma: technological breakthroughs demand vast data, but ethical governance and user protection lag behind. For example, DeepMind’s processing of 1.6 million medical records without explicit consent led to legal scrutiny, while the Italian Data Protection Authority flagged ChatGPT’s data collection practices as illegal.

Achieving a balance between innovation and privacy requires a joint effort: establishing a co-governance network involving technologists, legal experts, and the public. This ensures a dynamic equilibrium—fostering innovation without compromising fundamental rights.

The future is already here

AI technology enters the market as a tool to increase efficiency, and achieves commercial breakthroughs through automated processes and data analysis, but this is at a high cost.

According to data from the China Industrial Internet Research Institute, in 2024, leading manufacturers such as Byte Volcano Engine, Alibaba Cloud, and Baidu Cloud took the initiative to break the cost dilemma and set off a large-model price war, with price cuts generally reaching more than 90%.

The seemingly brutal price game is actually a strategic move to seize market shares in the AI ecology: stimulating market demand through price cuts and trading in short-term concessions for long-term market status, with the aim to accelerate the implementation of large language models and to cultivate user habits.

The open source ecosystem of DeepSeek-R1 has demonstrated another effective approach, through “algorithmic innovation with limited computing power,” while cutting costs to just a small fraction of Open AI. DeepSeek attracts global developers to innovate based on its technology through open source, making its technical path an industry benchmark. At the same time, by lowering the technical threshold, it can also stimulate long-tail innovation.

The revelation brought to the industry is that in the context of the rapid development of AI, the efficiency revolution is not achieved overnight, but a protracted battle of continuous investment and innovation. Enterprises need to constantly explore how to maximize the use of technology with limited resources, while finding the intersection of technology and business needs.

However, the operation of AI systems depends on massive data, and the data involves many ethical issues in the process of collection, storage, use and sharing, such as data privacy protection, data bias, algorithm discrimination, etc. Only when data ethics are fully guaranteed can AI technology truly realize its due social value and commercial potential.

Finally, while using AI to improve efficiency, AI users should also be wary of AI illusions, that is, AI tools would generate seemingly reasonable but actually false or misleading information, that ordinary users often find difficult to distinguish from authentic information. This has the potential to cause a massive crisis of trust in AI applications in this commercial teething phase.

This article was originally published in Chinese on The Economic Observer.