In July 2026, Visa launched a new system called the Visa Stablecoin Platform. Banks, fintech companies, and payment service providers can use the same platform to mint, redeem, custody, and transfer stablecoins, while connecting bank accounts, wallets, and internal approval processes. The most interesting aspect of this development is the change in Visa's role. In the past, it was responsible for connecting banks, consumers, and merchants; today, it is beginning to integrate blockchain directly into its settlement system. As of March 2026, Visa's stablecoin settlement business has an annualized scale of about $7 billion, with a 50% growth in a single quarter. The pilot currently supports nine blockchains, allowing participants to continue settlements on weekends and holidays without waiting for traditional banks to reopen. Visa's stablecoin settlement announcement and the Visa Stablecoin Platform show that stablecoins are gradually evolving from trading media within the crypto community to backend currency for the global payment system. From "on-chain dollars" to payment infrastructure Stablecoins initially grew rapidly mainly due to crypto trading demand. The crypto market operates 24/7, and traders need a relatively stable-priced dollar substitute that can move across different platforms at any time. USDT and USDC thus became settlement tools among exchanges, on-chain protocols, and market makers. This infrastructure was later discovered by cross-border trade, freelancers, overseas enterprises, and emerging market users. Traditional cross-border payments often require going through exchange

In October 2014, when Lisa Su took over as AMD's CEO, the company had little room for error. The PC market was entering a downturn, Intel firmly controlled the high-end processors, and Nvidia was continuously expanding in the graphics card sector. AMD's product performance lagged behind, gross margins declined, and the market even began to speculate whether it would be acquired or broken up. That year, AMD's revenue was about $5.51 billion, with a net loss of $403 million, year-end cash and marketable securities of about $1.04 billion, but debt as high as $2.21 billion. By 2015, revenue further shrank to $3.99 billion, and losses in the computing and graphics business expanded to $502 million. At that time, AMD was like a marathon runner carrying debt, with two competitors ahead who had more cash and greater technological leads. Breathing room gained through game consoles AMD was able to survive the most dangerous phase, thanks first to Sony and Microsoft. During the PC processor slowdown, AMD secured semi-custom chip orders for the PlayStation 4 and Xbox One. Although the gross margin on game console chips was not impressive, it provided stable large-scale shipments and cash flow, allowing the company to continue funding R&D. This was a typical survival deal: first secure cash flow, then bet limited resources on products that determine the future. After Lisa Su took office, AMD did not chase all markets simultaneously. AMD narrowed its focus, concentrating R&D on high-performance computing, server processors, and the new generation Zen architecture.

In July 2026, Meta announced a deal that truly represents this era. It plans to build a 1GW data center campus in El Paso, Texas, USA, with an estimated total development cost of about $14 billion. Besides Meta, the investors include BlackRock's Global Infrastructure Partners and HPS Investment Partners. After the deal is completed, the fund managed by BlackRock will hold 80% equity in the joint venture, while Meta will retain only 20%. Once the campus is built, Meta will use it through a lease arrangement for up to 20 years. The entire project also includes approximately $12.5 billion in debt financing. Simply put, the tech giant is responsible for defining computing power needs, designing the campus, and becoming a long-term tenant; asset management companies provide equity capital; banks, private credit, and bond investors continue to leverage. AI data centers are beginning to have a financing structure similar to airports, power plants, and highways. Official Meta announcement This may be one of the most noteworthy changes in the AI industry currently. When tech companies' cash runs out In recent years, the most attention in AI investment has been on GPUs. Whoever can buy more NVIDIA chips may have stronger model training capabilities. However, as AI competition expands from thousands to millions of GPUs, the bill starts to include more items typically seen in traditional industrial projects: land, power grids, substations, natural gas pipelines, energy storage, cooling systems, fiber optics, and long-term power contracts

In 1975, Kodak engineer Steven Sasson created a machine as bulky as a toaster. It had no film, used an electronic photosensitive element to capture images, stored photos on magnetic tape, and had a resolution of only about 0.01 megapixels. After taking a photo, it took more than twenty seconds to wait before connecting to a dedicated device to display it. This machine was later regarded as the world's first digital camera. At that time, Kodak was at the pinnacle of the global photography industry. It owned film, photo paper, chemical materials, cameras, developing equipment, and a worldwide sales network. Digital cameras seemed distant, expensive, and their image quality was far inferior to film. No one expected that the experimental product made by Kodak engineers would dismantle the entire film empire decades later. What it sold was never just a roll of film. In 1888, George Eastman launched the Kodak camera for the mass market. Consumers only needed to press the shutter and then send the camera back to the company for processing; Kodak was responsible for developing, printing, and reloading the film. Photography gradually transformed from a skill of a few professionals into a way for ordinary families to record life. The expansion of the global middle class in the 20th century created an extremely stable growth curve for Kodak. Weddings, travel, graduations, birthdays, the birth of children—every important moment required film, and each roll of film brought demand for developing and photo paper. Cameras were more like low-cost entry points; film, photo paper, and developing services were the long-term profit sources. This model is somewhat similar to later printer cartridges and coffee capsules: first get the device into the home, then continuously profit from each use. Kodak controlled the

Ford in 2026 is doing something very similar to what it did 120 years ago: moving mature manufacturing capabilities into a newly emerging large market. This time, the target is no longer limited to automobiles. In May this year, Ford officially launched Ford Energy, planning to invest about $2 billion to convert existing battery capacity in Kentucky into an energy storage system factory, with deliveries starting at the end of 2027 and a target annual capacity of at least 20 GWh. Customers include power companies, data centers, and large industrial enterprises. Subsequently, it signed a five-year framework agreement with EDF Power Solutions, under which the latter can purchase up to 4 GWh annually, with a potential total of 20 GWh over five years. AI data centers compete for electricity, and the global grid needs to absorb more wind and solar power; energy storage has transformed from a new energy accessory to a fundamental infrastructure for the digital economy. Battery factories, once burdened by electric vehicle investments, now have an outlet toward AI capital expenditures. Ford's revenue in Q2 this year was $48.3 billion, with adjusted EBIT of $2.5 billion, a 17% year-over-year increase; adjusted free cash flow was $2.1 billion, and total liquidity exceeded $43 billion. The company also raised its full-year 2026 adjusted EBIT guidance to $10 billion to $11 billion. These figures look like the normal financial report of a long-established industrial company. Rewind to 2008, Ford was uncertain even whether it could keep the blue oval trademark. The king of assembly lines was first trapped by its own scale. 1908

Over the past two years, the whole world has been scrambling for GPUs. Tech companies are competing for chips, sovereign wealth funds are investing in data centers, and governments are incorporating AI into their industrial strategies. The capital market once believed that as long as you had more high-performance chips, you could secure a ticket to the AI era. As data centers are being built one after another, a new bottleneck has emerged: chips can be obtained, but electricity supply is not guaranteed. Training models requires electricity, inference requires electricity, and server cooling also requires electricity. How fast data centers can continue to be built now depends on local grid capacity, substation equipment, power generation structure, land, water resources, and administrative permits. The AI competition is expanding from a chip war into a global race for energy and infrastructure. A data center is becoming an industrial city. Traditional internet data centers mainly handle search, e-commerce, video, and cloud storage. With the emergence of generative AI, computing density has significantly increased, requiring a large number of GPUs to run continuously for long periods, and cooling systems must expand accordingly. The International Energy Agency estimates that global data center electricity consumption will reach about 945 TWh by 2030, exceeding Japan's current annual electricity consumption. From 2024 to 2030, data center electricity demand is expected to grow about 15% annually, a rate more than four times that of other electricity sectors. Among this, AI-driven accelerated server electricity consumption is expected to grow about 30% annually. The United States and China are expected to contribute nearly 80% of the global increase in data center electricity consumption. By 2030, data centers may account for nearly half of the growth in U.S. electricity demand. U.S. Department of Energy

In the 1990s, many American families had a fixed weekend routine: driving to the nearby Blockbuster, discussing what movie to watch tonight in front of a whole wall of videotapes, then taking home several boxes of movies. That blue and yellow sign was once as ubiquitous in cities as McDonald's. What it seemed to sell was movie rentals, but what it truly occupied was the gateway to home entertainment. By the end of 2005, Blockbuster had over 9,000 stores in the United States and 24 other countries. Its vast physical network, film inventory, and brand recognition made it the undisputed leader in the global video rental industry. Just five years later, Blockbuster filed for bankruptcy protection. How could an industry leader with global stores, tens of millions of customers, and a strong brand exit the market in such a short time? Blockbuster benefited from the boom in home entertainment. Founded in 1985, at that time, cable TV was just beginning to spread in the U.S., home entertainment demand outside of cinemas was rapidly growing, and VCRs were gradually entering middle-class households. Watching movies was still limited by time and space. TV programs aired according to schedules, cinemas had fixed showtimes, and consumers who wanted to decide when to watch had almost no choice but videotapes. Blockbuster transformed scattered small video stores into standardized chain businesses. With uniform storefronts, centralized purchasing, computerized inventory management, and opening stores in the growing suburbs. This model hit multiple trends in the U.S. economy at the time: suburban population expansion, mature car culture, growth in commercial real estate, and increased household spending power. The more Blockbuster stores there were