Small Modular Nuclear Power: AI Competing for Electricity, Nuclear Power Begins Factory Production In August 2026, TerraPower, founded by Bill Gates, signed a cooperation agreement with South Korea's SK Innovation to jointly participate in small modular reactor projects in the United States and overseas. South Korean companies aim to provide key equipment, engineering construction, and operational capabilities, transferring supply chain experience accumulated from shipbuilding, refining, and large machinery industries into the new generation nuclear power sector. A few months ago, the U.S. Department of Energy selected eight companies to provide over $94 million in funding to help address licensing, supply chain, and site preparation issues for small modular reactors. Another federal fund of up to $800 million was allocated to TVA and Holtec to advance the first projects in Tennessee and Michigan. The U.S. Department of Energy announced the TVA and Holtec projects in May 2026. The revival of nuclear power has been advocated for many years. The force bringing it back into capital focus now comes from a very modern scenario: AI data centers are lining up to compete for electricity. Why AI is putting nuclear power back on the table Over the past decade, the global power system has mainly expanded around wind, solar, and natural gas. Renewable energy costs have rapidly declined, and construction cycles are shorter than nuclear power, but data centers require continuous year-round operation. Model training cannot be paused midway due to cloudy days, no wind, or grid congestion, so companies need stable, low-carbon power sources. Traditional large nuclear power plants can provide stable baseload power, but the problem lies in the engineering scale. Each power plant is too large to...

Around the year 2000 in the United States, spending weekends browsing bookstores was a very formal leisure activity. People would walk into the warmly lit Borders inside the mall, first flipping through a few novels in the new books section, then going to the music section to listen to CDs. The bookstore smelled of coffee, had wooden bookshelves, and sofas where you could sit for hours. Students came here to find research materials, parents brought their children to book clubs, and people just off work would casually buy a magazine. At that time, large chain bookstores seemed to have everything the internet couldn't replicate: space, atmosphere, inventory, professional staff, and the serendipitous feeling of encountering books. Ten years later, the same bookshelves were labeled with clearance discounts, the coffee machines were removed, and even the tables and chairs became items for sale. Readers did not disappear, and the global book market did not stop operating. What disappeared was the era when consumers had to drive to large bookstores to find a book. 1. It turned bookstores into cultural department stores Borders was founded in 1971 in Ann Arbor, Michigan, by brothers Tom and Louis Borders. They were early adopters of computer systems to track inventory and regional sales data. While many independent bookstores still relied on the owner's experience to stock books, Borders was already able to analyze reading preferences in different cities and then decide which books each store should carry. In the 1990s, suburbanization in the U.S., shopping malls, and large retail businesses expanded simultaneously. Malls needed anchor stores that could attract customers long-term, and consumers were accustomed to completing shopping, dining, and entertainment in one trip by car. Borders seized this cycle. It expanded the traditional bookstore

1973, Memphis Airport, USA. Fourteen purple and white small cargo planes were parked beside the runway, with 389 employees waiting for packages to be delivered to the sorting center. FedEx received only 186 shipments on its first night but had to fly to 25 U.S. cities. This system was expensive from day one: planes, fuel, pilots, airports, trucks, and sorting centers all had to be prepared in advance, but revenue would only appear after customers developed the habit of "overnight delivery." A few months later, the first oil crisis broke out, causing aviation fuel costs to soar. FedEx continued to lose money, financing was nearly cut off, and founder Fred Smith was even unable to pay the next week's fuel bill. Today's FedEx connects more than 220 countries and regions. In fiscal year 2026, revenue reached $94.7 billion, and the value of goods handled annually impacts global manufacturing, consumption, e-commerce, and supply chains. However, when this logistics empire was just starting, there was only $5,000 left in the account. A logistics concept that received low marks from a professor In 1965, Fred Smith was still a Yale University student. In his thesis, he proposed a new transportation model: consolidating goods from different cities to a central hub, sorting them overnight, and then flying them to their destinations. This "hub-and-spoke network" is very similar to a bank clearing center. Planes do not need to fly directly from every city to all destinations; they only need to fly to Memphis uniformly and then redistribute from Memphis. The more cities, the greater the network value, and the marginal efficiency of adding new routes also increases. At that time, U.S. freight mainly relied on passenger plane belly

On August 19, 2026, Unitree Technology was listed on the Shanghai STAR Market, with its stock price surging over 600% on the first day of trading. Behind this capital frenzy, what the market is truly betting on is not robots dancing, boxing, or doing backflips. Investors are expecting a brand-new type of labor commodity: robots that can enter factories already built by humans, use human tools, and undertake tasks such as handling, assembly, quality inspection, and hazardous operations without the need to remodel the entire production line. Unitree delivered over 5,500 humanoid robots in 2025, with revenue increasing from about ¥159 million in 2023 to approximately ¥1.7 billion in 2025. The company is raising about ¥6.1 billion this time, investing the funds into robot models, body development, new products, and manufacturing bases. Unitree's listing and operational data finally provide the robotics industry with a sample that can be directly valued by the capital market. Why do factories need "human-like" robots? Industrial robots have existed for decades. From automotive welding to wafer handling, robotic arms have long been part of modern manufacturing. However, traditional robotic arms are usually fixed within fences, each responsible for a few repetitive actions. When production lines change, companies often need to reprogram, install new fixtures, or even remodel the factory. The commercial logic of humanoid robots comes from another direction: factories, warehouses, and tools worldwide are originally designed according to human height, arm length, and movement patterns. If a robot has two hands, a vision system, and a range of motion close to that of humans, it has the opportunity to directly enter existing environments. What companies purchase is no longer just a single

In 1999, most people who first entered an internet cafe didn't know where to start surfing the web. The computer desktop only had a browser, and the dial-up network emitted a harsh connection sound. After the page opened, many people's first entered website was Yahoo. News, email, chat rooms, finance, sports, shopping, and website categories were all neatly arranged on the homepage. At that time, the internet was like a newly opened city without a map. Yahoo stood at the entrance, hanging signposts for everyone. More than twenty years later, people still search for information, read news, send and receive emails every day, and the online advertising market has grown countless times compared to back then. Yahoo's website still exists, and Yahoo Finance even retains a fairly stable user base, but the power to decide how global internet traffic is distributed has long since shifted to search engines, social platforms, mobile systems, and app stores. What Yahoo lost was not just a website, but the position of the "internet homepage." 1. It organized the chaotic internet into a directory. In 1994, Stanford University graduate students Jerry Yang and David Filo created a website directory, initially just categorizing and organizing their favorite web pages. The early internet was limited in scale, and search technology was immature, so manual screening and classification better matched user habits. Yahoo quickly expanded from a directory into a comprehensive portal. Users came to Yahoo to find websites and also read news, check the weather, send and receive emails, and discuss stocks. The more services added, the longer users stayed; the greater the traffic, the more advertisers were willing to pay. This model

In Seattle in 2001, Amazon's warehouses were piled high with unsold goods, layoffs began in the offices, and the capital markets were discussing an even harsher question: how much longer could this company survive? Before the internet bubble burst, as long as the name included ".com", losses could be explained away as investing in the future. After the bubble burst, investors suddenly only recognized cash, debt, and profits. From its peak in 1999 to its trough, Amazon's stock fell by about 95%, and the capital markets that had supported its expansion quickly shut down. In the first quarter of 2001, the company held about $643 million in cash and marketable securities, but long-term debt reached $2.119 billion, and shareholders' equity was already negative. The GAAP net loss for the quarter was $234 million, and the market began to worry that it would run out of cash before its bonds matured. Amazon's Q1 2001 performance⁠ 25 years later, Amazon's market value surpassed $3 trillion for the first time in August 2026. The same company operates the world's largest e-commerce and logistics network while selling cloud computing, chips, and model services to AI enterprises. Its survival to this day is due to transforming its "growth story" into a cash-generating business at the coldest time for capital. When the internet bubble burst, scale almost became a burden. In the late 1990s, the U.S. was experiencing low inflation, a strong dollar, and a wave of tech investment, with global funds flowing into Nasdaq. Amazon took advantage of the financing window to expand categories, build warehouses, acquire companies, and issued a large amount of convertible bonds. The logic at the time