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

AI trading in 2026 is spreading from chips all the way to power systems. Data centers require gas turbines, transformers, transmission equipment, and a stable power grid. Traditional industrial assets, once shunned by the capital markets for many years, have suddenly taken center stage in the tech wave. In the second quarter of this year, GE Vernova's orders reached $24.2 billion, an 88% year-over-year increase; backlog orders rose to $176 billion, with electrification orders related to data centers exceeding $5 billion in the first half of the year, more than double the total for 2025. Meanwhile, GE Aerospace's second-quarter revenue was $13.3 billion, up 21% year-over-year, with orders increasing 17% to $16.5 billion, and free cash flow growing 43% to $3 billion. Both companies raised their full-year guidance simultaneously. It's hard to imagine that these popular assets were once packed into the same vast empire, which in 2008 needed government credit, Buffett, and capital markets to simultaneously bail it out. GE's turnaround is also quite special. It did not restore the original company intact but spent more than a decade selling assets and repaying debts, ultimately splitting itself into three companies. A name that has dominated American business history for over a century earned the qualification for renewed growth by ending its old era. From light bulbs to the world's largest market capitalization, GE once represented America itself. In 1892, Edison’s companies merged with Thomson-Houston to form General Electric. For more than a century thereafter, GE...

On August 18, the US stock AI sector collectively cooled down, with Micron dropping about 7% in a single day, becoming one of the most prominent companies in the semiconductor sell-off. This decline is not hard to understand. Since the beginning of 2026, Micron's stock price had once surged about threefold, with the market already pricing in AI memory shortages, product price increases, and high profit margins. The 30-year US Treasury yield also rose to its highest level since 2007, prompting investors to reconsider a question: can the money earned by the AI industry cover the increasingly expensive capital costs? However, if we look back three years, Wall Street was worried about something else: how long Micron would continue to lose money. In fiscal year 2023, the company's revenue fell from $30.76 billion to $15.54 billion, nearly halving; gross margin dropped from 45% to negative 9%, with a full-year net loss of $5.83 billion. By the third quarter of fiscal year 2026, Micron's single-quarter revenue had reached $41.46 billion, with GAAP net profit of $28.24 billion, both setting records. A company once regarded as a "cyclical stock selling commodity chips" has, in just three years, become one of the most profitable segments in AI infrastructure. This turnaround may seem sudden, but behind it lies a survival history spanning more than forty years. When Intel left the memory business, Micron chose to stay. Micron was founded in 1978, starting in Boise, Idaho, USA. The early DRAM market was dominated by American companies, but by the 1980s, Japanese manufacturers quickly took market share with advantages in yield, scale, and price.