Sunday, May 31, 2026

What deserves most attention today is the two-level differentiation trend in the AI industry. SoftBank has spent 75 billion euros to build AI infrastructure in France, and Anthropic is valued at nearly one trillion yuan. These numbers sound scary, but there is a cruel reality behind them: the big model track has become a capital-intensive arms race. The entry barrier is now ridiculously high, and small players are either acquired or transformed into vertical applications. Look at the energy AI agent developed by Sige New Energy. This is a smart way-instead of hard-working and universal models, it is better to deeply cultivate a certain industry and use AI as a tool.

When it comes to vertical applications, the two directions of pet robots and industrial robots are quite interesting. Generation Z buys AI pets to relieve boredom, and Foxconn looks for a robot solution provider to improve the efficiency of its production line. On the surface, it doesn't work well, but in fact they are all good examples of AI implementation. But the problem is that the former can easily be reduced to a toy, while the latter has to face the cruel ROI assessment of the manufacturing industry. That passenger robot's half-year revenue of 20 million yuan sounds good, but you should know that the customization needs of industrial scenarios can drag you to death, and you may end up turning into an outsourcing company.

The most ironic thing is the reaction of the developer community. Some people on V2EX already feel that going to school is a waste of time and can learn programming by chatting directly with AI. This reflects a dangerous trend: junior developers are increasingly relying on AI to generate code, but still miss complex problems. You see OpenAI pushing biodefense tools and NVIDIA engaging in LocateAnything. These high-end players are solving practical problems, while ordinary programmers are still struggling with HP printer drivers. The technological gap is widening, and it is likely that a group of "pseudo-developers" will emerge in the future, who will only tune APIs but not understand the underlying principles. It will be even more difficult for companies to recruit people, because people with five years of experience on their resumes may not even be able to write the basic algorithm by hand.

The scale of energy agents and robots is reshaping the industrial cost structure. The release of Sige New Energy's global AI agent, SigenAgent, is by no means an isolated event-behind it is AI's strategic transition from information processing to physical world control. When Sige combines energy management with the hierarchical logic of autonomous driving, it is essentially building the "nerve center" of the power network, which secretly echoes the construction of a factory with an annual output of tens of millions of Tesla Optimus humanoid robots: the former optimizes energy flow efficiency, and the latter consumes cheap labor. What is even more interesting is that Jingdong's "Nirvana Project" trains workers to repair robots, exposing the collective anxiety of traditional enterprises in the face of AI costs. The tram road maintenance fee dispute is essentially a derivative of this anxiety. When new energy vehicles weigh more than 4 tons but are exempted from road maintenance fees, they are actually using policy subsidies in exchange for AI's window to reshape the transportation industry.

The global computing power competition triggers a new geopolitical defense line. SoftBank's 75 billion euro bet on the French AI computing power cluster is like a thunder. Its location at Dunkirk points directly to the English Channel Data Corridor, which is in sharp contrast to Japan's "Battlefield Cleanup" military exercise in which 120,000 residents were evacuated in the Okinawa Islands. These two events jointly reveal the birth of a new digital frontier: the former competes for economic sovereignty over the computing power radiation circle, and the latter previews the physical defense when the technology supply chain breaks. Pfizer and Cinda Biotech's US$10.5 billion oncology drug cooperation is particularly subtle in this context-in the shadow of tightening investment in biotechnology in the United States, multinational pharmaceutical companies are voting with real money, trying to bypass geopolitics and establish cooperative enclaves at the cellular level. The drama of the SpaceX valuation controversy and the Blue Origin explosion is nothing more than a ripple effect of the space infrastructure race.

The explosion of AI agents has given rise to the "wet ware revolution" and emotional computing has become a new battlefield. The secret of Anthropic's close to a trillion-dollar valuation is hidden in the developer forum's lament that "after chatting with Gemini for four hours, I feel that class is a waste of time." When Claude is used for automated penetration testing (digging holes), and when LFM models are frantically chasing stars on GitHub, it means that AI agents have broken through the technical toy stage and entered the deep area of productivity reconstruction. But what is even more disruptive is the sudden emergence of the pet robot track-Fu Zai, which sells for 369 yuan, has become popular among the silver-haired people, exposing the hard need for emotional compensation of Generation Z. This forms a wonderful chain of cause and effect with the OpenAI biodefense program: when GPT-Rosalind tried to block the spread of pathogens, emotional agents were blocking the mental epidemic of modern people. The essence of Huawei Hongmeng and Jingdong 618's competition for entrance is a key infrastructure battle to fight this spiritual epidemic.

The intersection of three major trends hides the winner and winner in the next three years: the integration of energy and AI will rewrite the industrial cost formula, but the tram road maintenance dispute reveals that institutional lag may become a stumbling block; sovereign-level computing power investment must be synchronized with geo-defense, otherwise Japan's technological conflict previews may turn from exercise to reality; the field of emotional computing may seem like a blue ocean, but when the 369-yuan pet robot begins to collect brainwave-level data, its ethical risks far exceed that of conversational AI. Decision makers are advised to pay attention to one key indicator: the day when the hourly salary of robot maintenance technicians exceeds that of programmers is the real signal that the industrial earthquake has begun.

Today's technological signals present two significant themes: First, the field of artificial intelligence continues to surge, with huge investments and high valuations common, which seems to herald a disruptive future; second, with this fanaticism, some hidden worries and reality The friction also gradually emerges, reminding us to return to rationality.

First of all, AI's "agent" narrative is being fully rolled out. Sige New Energy has released a global AI agent, claiming to bring energy management into the "era of agents"; at the same time, Anthropic's valuation is close to one trillion yuan, and some analysts even dare to bet on its "ten trillion yuan" valuation. The FOMO (fear of missing out) emotion is beyond words. Softbank even spent 75 billion euros to build an AI computing cluster in France, demonstrating its great confidence in AI infrastructure. From a specific application perspective, from AI pet robots targeting the lonely economy of Generation Z, to Tesla's planning of a humanoid robot factory with an annual production capacity of 10 million units, to satellites moving from passive response to autonomous decision-making in "space smart driving", AI is trying to penetrate into all aspects of our lives, even deep space. These all point to a trend: AI is no longer just a tool, it is given greater autonomy and decision-making capabilities, and becomes an "agent" to solve problems. However, in this wave of agents, how big the core technological breakthroughs are and how cost-effective they are often concealed by high valuations and grand narratives. History tells us that excessive expectations are often a precursor to bubbles.

Secondly, under the appearance of AI prosperity, real challenges and ignored costs are emerging. Jingdong Liu Qiangdong's "Brotherhood Theory" and "Nirvana Project" show that even large manufacturers realize the high cost of DAU in the AI era and require fewer but stronger people to improve efficiency. This is in sharp contrast to the logic of the Internet's pursuit of scale expansion in the past. It also suggests that what AI brings is not brainless expansion, but careful calculation of efficiency and cost. What deserves more attention is the discussion of tram road maintenance fees. The weight and damage to roads of new energy vehicles, as well as the lag of the existing tax system, reveal the shift in infrastructure and public service costs brought about by technological changes. As the demand for AI computing power continues to rise, SoftBank has invested 75 billion euros in France to build a 5 gigawatt computing power cluster. Have we fully assessed and prepared to bear the huge energy consumption and environmental costs behind this? Although OpenAI launched a special biodefense program out of good intentions, it also vaguely reveals the other side of AI's double-edged sword: while its powerful capabilities are used for defense, they may also bring new risks and ethical challenges.

To sum up, AI is undoubtedly the core driving force for current technological development, and its penetration and evolution in various fields are eye-catching. But we cannot be carried away by superficial valuation orgies and grand narratives. The real value lies in whether AI can bring sustainable and universally beneficial benefits, rather than just speculation in the capital market. Its high operating costs, impact on existing infrastructure, and potential social and ethical risks are the focus of our long-term observers. These real frictions may be the most real resistance and adjustment driving force in AI development in the next few years.