Among today's technological signals, the most worthy topics are "Industrialization of AI" and "Vulnerability and Reconstruction of Infrastructure"-behind these two clues, there is a larger trend: technology is moving from "showmanship in the laboratory" to "base of social infrastructure", and every step in this transformation process is redefining who is the winner and who will be eliminated.
Let's talk about the industrialization of AI first. The release of GLM-5.3-Flash and Qwen3.8-Flash-Next may seem to be just model iterations, but in fact it is a watershed for AI to move from a "performance race" to a "cost and efficiency race." Smart Spectrum compresses the active parameter to 18 billion, but the price is only one-tenth of that of the previous generation. What does this mean? This means that AI is no longer a game for a few players, but has begun to become an infrastructure like electricity. You can compare the development of the mobile Internet: in the 2010s, the popularization of 4G gave rise to supercomputers in everyone's pocket, giving birth to platforms such as Douyin and Panduo; in the late 2020s, AI's "Flash" series was doing the same thing-Lower the computing power threshold low enough so that every small and medium-sized enterprise and every entrepreneur can afford it. The logic behind this is that when the marginal cost of technology approaches zero, the main battlefield of innovation will shift from "who has a stronger model" to "who can embed the model into specific scenarios faster." For example, AI short films such as "You Don't Have Time to Live Anyway" are popular not because the picture quality is amazing, but because they capture the pain point of "emotion"-when AI sets the production threshold to zero, it becomes an insight into human nature. This also explains why bean buns can crush Kimi in 380 million months: In the era of AI industrialization, scale effects and scene adaptation capabilities are more important than "high-cold technical tonality."
But the other side of AI industrialization is the extreme reliance on infrastructure. The mudslide disaster at the Jilong Port in Xizang exposed a cruel reality: our digital society is built on an extremely fragile physical foundation. A single mudslide can interrupt communications, power, and roads, and lose 265 people-this reminds us that no matter how powerful AI is, it still needs a "visible hand" to ensure its operation. Interestingly, this signal is in sharp contrast to the cooperation between Damao Technology and Fujian Intelligent Computing Ark: the former is the fragility of infrastructure, while the latter is the reconstruction of infrastructure. Damao Technology's "full-stack computing and power collaborative service" is essentially an "AI power grid"-through computing and power collaboration, computing power can be dispatchable and operable like electricity. The logic behind this is that when AI becomes infrastructure, its demand for energy and networks will be the same as the industrial revolution's demand for coal, forcing the entire society to redesign the energy and communications distribution system. Qualcomm Ma Dejia mentioned that "6G is an AI native system" is actually an extension of this trend: the future network will no longer be a simple "connection", but will provide underlying support for real-time computing and distributed collaboration of AI. This means that whoever can take the lead in the "AI grid" and "AI network" will be able to take the initiative in the next decade.
Where these two threads meet is a deeper question: When AI becomes infrastructure, who pays for its externalities? Bill Gates mentioned "robot tax" and "human exclusive post", in fact, is discussing the reconstruction of social contract after AI industrialization. When AI makes some jobs disappear, society needs new mechanisms to distribute productivity dividends. Disaster relief in Xizang exposes another externality: When infrastructure is fragile, who will bear the risks? The 10 million yuan donation from the Xiaomi Charity Foundation may seem to be corporate social responsibility, but in fact it is also a kind of "risk hedging"-in the AI era, the survival of enterprises is increasingly dependent on a stable social environment, and the stability of the social environment is increasingly dependent on the resilience of infrastructure. This has formed a closed loop: AI industrialization requires stronger infrastructure, and the fragility of infrastructure forces society to rethink the issues of "who will build" and "who will maintain it."
How will this trend evolve six months to one year later? AI's "Flash" series will further reduce costs and allow "AI-native" startups to emerge in more vertical scenarios, just like the mobile Internet entrepreneurial wave of the 2010s. But at the same time, the vulnerability of infrastructure will become a bigger bottleneck-you will see more investment in "AI power grids", more attempts to "sink computing power", and there may even be a policy like "electricity inclusive" that forces computing power resources to be tilted towards underdeveloped areas. The biggest variable lies in the reconstruction of social contracts. When AI makes some people "useless", how to avoid social tearing? Gates '"robot tax" may be just the beginning, and there may be more radical policies in the future, such as "computing quotas" and "AI licenses" to balance efficiency and fairness. In this process, the biggest winners will be players who can harness AI industrialization and manage infrastructure risks-such as technology giants with both algorithmic capabilities and power and network resources-while the biggest losers will be AI labs that only show off their skills and local governments that ignore infrastructure vulnerabilities.