During the 2024 financial report season, the massive investments made by global tech giants in artificial intelligence (AI) garnered widespread attention and debate. Companies like Meta, Google, Microsoft, and Amazon have announced plans to spend billions of dollars in an effort to maintain their competitive edge in the global AI race. However, with the rise of the Chinese AI startup DeepSeek, these tech giants’ investment strategies and market responses have come under scrutiny. DeepSeek, with an investment of less than $6 million, launched an open-source AI model, R1, which in some aspects outperforms OpenAI’s O1 model, shaking the industry.

The Necessity of Huge Investments

The financial reports of US tech giants typically include large-scale AI investment plans. For instance, Meta plans to invest $60-$65 billion by 2025, Microsoft has committed $80 billion, Google expects to invest $75 billion by 2025, and Amazon plans to increase its capital expenditures to $100 billion. These investments imply that tech companies are increasing their spending on AI infrastructure, including high-performance data centers, specialized chips, and servers, to support AI model training and application.

However, the emergence of DeepSeek has raised questions about the necessity of such enormous investments. DeepSeek managed to develop an AI model with 671 billion parameters using only 2048 GPUs and an investment of less than $6 million, surpassing OpenAI's leading technology in certain tasks. Compared to the massive investments made by the tech giants, DeepSeek's success seems to suggest a more cost-effective and efficient approach to AI development, leading to a reevaluation of the traditional investment model.

Investment Risks and Market Doubts

DeepSeek’s low-cost and low-resource success story has led the tech industry to question whether all AI technologies require massive capital expenditures. The rise of DeepSeek has sparked broad discussions in the market about the investment strategies of American tech giants. Analysts and investors have argued that the over-reliance on huge capital spending may not necessarily yield sufficient returns. Daniel Newman, an analyst at Futurum Group, noted that while these investments help advance AI technology, the current AI market’s profitability model is unclear, and there is a risk of over-investment.

These doubts are not limited to investors; they have also infiltrated the leadership of tech companies. Microsoft and Meta, for example, have insisted that their large-scale investments are aimed at establishing long-term strategic advantages in the AI field. Meta’s CEO Mark Zuckerberg believes that capital expenditures will be a core advantage in future competition. He argued that, despite the attention DeepSeek’s success has garnered, the AI industry is evolving rapidly, and DeepSeek’s rise is merely a temporary disruption. In the long term, investment in infrastructure and capital remains crucial.

In-Depth Analysis of the Rationality of AI Investments

Yann LeCun, Meta’s chief AI scientist, presented a different perspective on this issue. He argued that DeepSeek’s low cost primarily concerns training phase investments, not the substantial expenditures needed for inference infrastructure. LeCun emphasized that to provide AI services to billions of people globally, tech companies must invest heavily in inference, particularly in areas like video understanding, reasoning, and large-scale memory. Comparing DeepSeek’s training costs with the comprehensive capital expenditures of tech companies is a misunderstanding of AI investment.

The investments made by tech giants are not solely focused on training AI models but also on enhancing inference capabilities to meet the rapidly growing demand for AI services worldwide. For example, Microsoft and Google’s cloud platforms have seen exponential growth in the demand for AI services, making the development of inference infrastructure critical. While DeepSeek has made breakthroughs in training, it still cannot replace the long-term investments made by tech companies in inference, data security, and service customization.

Changes in the AI Competitive Landscape

The rise of DeepSeek has undoubtedly shaken the global AI industry, but this does not mean that the AI investments of tech giants are now irrelevant. In fact, DeepSeek’s success more likely proves that AI technology may head toward a more commodified, cost-effective direction. Even so, tech giants still possess significant capital advantages and robust technological infrastructures, allowing them to maintain differentiated competitive advantages through continuous optimization and enhancement of inference capabilities, data security, and personalized services.

In the future, the AI industry will continue to develop more efficient and widespread applications, with technology and capital barriers likely diminishing. AI technology will become more accessible and cost-effective. However, the capital expenditures of tech giants are not only aimed at current competition but also at long-term plans for technological upgrades and global service expansion. Therefore, while DeepSeek’s low-cost technology provides new insights into the market, the ultimate winners in the global AI landscape are likely to be those tech giants who continue to invest in and innovate in areas like efficient inference, technological customization, and global service.

Conclusion

In conclusion, the rise of DeepSeek has offered a new perspective on the tech industry and prompted investors to question the necessity of large-scale AI investments. However, tech giants still insist that these investments are essential for securing strategic advantages in the future AI race. As AI technology continues to evolve and the market matures, finding the right balance between capital investment and technological innovation will be key for tech companies to remain competitive in the future.