The first-quarter earnings reports from seven major U.S. tech giants have silenced analysts who had been warning of an “AI bubble.” In just 90 days, these seven companies collectively posted net profits of $230 billion—a staggering figure that once seemed unimaginable.
To put that in perspective, consider the combined earnings of China’s most profitable internet and banking firms: Alibaba, Tencent, ByteDance, JD.com, Xiaomi, plus the six largest state-owned banks. Working a full year, their total profits barely reach $2 trillion. But these seven U.S. companies achieved what China’s top players do in months within just three months.
Among them, two stand out. First, Google posted a quarterly net profit of $62.5 billion, up nearly 80% year-over-year. That’s about $425 billion in three months—more than Tencent’s net profit over two years. Google’s aging search engine, now powered by AI Overview, has been revitalized, and its cloud business is soaring thanks to AI.
Then there’s Jensen Huang’s NVIDIA. The chip giant reported first-quarter revenue of $68 billion and net profit of $43 billion, with an astonishing profit margin. For the full fiscal 2026, NVIDIA guided net profit to $120 billion—over $800 billion in Chinese yuan. That’s an average of $22 billion added to its coffers every day. Such earnings velocity would have been unthinkable for a chip company a decade ago.
As profits surged, so did the stock price. NVIDIA’s market cap hit $5.2 trillion, making it the world’s most valuable company. At 35 trillion RMB, that’s equivalent to 15 Alibabas or 9 Tencent. Over the past year and a half, its market cap swelled by $2 trillion—the size of Apple’s entire market value. In business history, few companies have grown at this pace.
Why does NVIDIA dominate so completely? Many compare it to selling shovels during a gold rush, but that’s an understatement. It’s more like collecting a tax on the very process. Anyone wanting to build a large AI model must use NVIDIA’s H100 or H200 chips, priced at nearly $300,000 each. Even at that price, customers must wait in line. Silicon Valley giants now beg for access; without NVIDIA’s chips, even the best models can’t run.
NVIDIA’s moat is deep. It relies on TSMC’s custom 4nm process, with limited capacity that no one else can access. The H100 requires HBM high-bandwidth memory, produced by only three companies: SK Hynix (70% market share), Samsung (20-30%), and Micron (a sliver). Thus, NVIDIA locks up the entire supply chain’s scarce resources.
This dynamic has also supercharged South Korea’s stock market. NVIDIA’s demand for HBM pushed SK Hynix’s revenue toward 500 billion RMB this year, with record net profit. Samsung’s quarterly operating profit exceeded 260 billion RMB, matching its entire annual profit from previous years. These two stocks account for nearly half of the KOSPI’s weight, driving the index up nearly 80% in just over four months, propelling South Korea to the world’s seventh-largest stock market.
Behind NVIDIA’s success lies a battle of fundamental physics. Each H100 requires six HBM memory chips, and every stage—chip manufacturing, advanced packaging, lithography precision—pushes the limits of materials science. TSMC engineers dream about yield rates. This is no longer just business; it’s a championship round of national industrial capability.
The ultimate cost? Electricity. Thousands of server racks run 24/7. Microsoft and Google are signing long-term contracts with nuclear plants; Amazon is building data centers next to reactors. Wall Street’s sharpest investors are quietly shifting capital into energy and power stocks. Even nuclear fusion, once science fiction, is now hot. The AI arms race may ultimately be decided not by algorithms but by power plants.
Looking back at China: Huawei’s Ascend is struggling to catch up, domestic HBM development is advancing, but we must admit the gap. Being squeezed at the throat is bitter, but this mountain must be climbed. NVIDIA’s story shows that whoever controls the key nodes at the top of the industrial chain makes the rules. China rose on manufacturing; in the new era of computing power, it must fight a tough battle again.
There are no shortcuts. Whether large AI models will transform the world like electricity or the internet remains uncertain. Optimists call it the start of an industrial revolution; skeptics see a game of hot potato. But one thing is already clear: the road builders are collecting tolls with hands full. When the dust settles, whether the gold is real or not, Jensen Huang’s vault will be overflowing. That’s the true business landscape of 2026.