Just after the Lunar New Year, China’s dark-horse AI model DeepSeek—touted as enabling “low-cost model training”—briefly sent a chill through the market. That scare later proved overblown.
By the fourth quarter, however, OpenAI founder Sam Altman and Nvidia CEO Jensen Huang（黃仁勳） jointly unveiled a series of eye-popping AI data-center expansion plans, igniting a new round of the global AI arms race.
As a result, expectations that AI investment would cool sharply in 2026 have been upended. Market-research firms and investment banks are now steadily revising upward their capital-expenditure forecasts for the four major U.S. cloud providers.
The question commanding the most attention is whether OpenAI will go public in 2026.
The company recently completed a major restructuring, transitioning from a nonprofit to a corporate-style structure that more closely resembles a traditional for-profit company—widely seen as a preparatory step toward an IPO.
Reuters has reported that an application could be filed as early as the second half of 2026.
“He simply doesn’t have the money,” said Jonah Cheng (程正樺), a well-known tech analyst and founder of J&J Investment Ltd（騰旭）.
OpenAI’s annual revenue stands at just over $10 billion, yet it plans to spend more than $1 trillion over the coming years building data centers. “It has to exhaust every possible funding channel,” Cheng said. “An IPO is inevitable.”
OpenAI’s latest funding round is expected to push its valuation to $750 billion. Once listed, its market capitalization would almost certainly exceed $1 trillion, making it one of the largest IPOs in history.
Still, Altman struck an ambiguous tone last week when asked about an IPO during a podcast interview. He said he had “zero interest” in running a public company—but added, “Would I be excited about OpenAI going public? In some ways, yes.” He also openly acknowledged that OpenAI needs enormous amounts of capital.
To me, this suggests that Altman has made peace with reality: OpenAI has little choice but to go public. Even so, he remains reluctant, well aware that life as a listed company would bring constant scrutiny from capital markets and sharply limit the freedom he once enjoyed.
Trend One: An Epic Wave of “Hectocorn” IPOs
By comparison, the IPO timeline for OpenAI’s main rival, Anthropic, appears clearer. Multiple media reports indicate that the company has already retained legal counsel and is highly likely to file for an IPO next year.
Anthropic is currently valued at roughly $300 billion, making it one of only six “hectocorns” worldwide—private companies valued at more than $100 billion, or 100 times the threshold of a unicorn.
The last tech company to grow this large before going public was Alibaba, which listed in New York in 2014 at a valuation of about $160 billion, then the largest tech IPO in history.
Next year, the market may witness something unprecedented: three hectocorns entering the public markets in close succession. In addition to OpenAI and Anthropic, there is SpaceX—a two-decade-old giant. Analysts expect SpaceX’s valuation to soar to as much as $1.5 trillion after listing, surpassing even TSMC.
In mid-December, Elon Musk, in his trademark unconventional fashion, all but “confirmed” the IPO. Responding to an article by prominent space journalist Eric Berger, Musk simply commented: “Everything he says is correct.”
That article reported that SpaceX plans to go public and raise $300 billion, with Musk aiming to leverage its tens of thousands of Starlink satellites in low-Earth orbit to build AI data centers in space—powered directly by solar energy far stronger than what is available on Earth, thereby addressing America’s looming AI-driven power shortage.
Between 2026 and 2027, these three hectocorns are expected to unleash an epic wave of “super IPOs” to fund the AI arms race—arguably the single most important event in the tech industry next year.
U.S. media have joked that the side effect will be the creation of countless new AI millionaires in the San Francisco Bay Area, sending local housing prices soaring—suggesting that anyone planning to buy property there should do so sooner rather than later.
For Taiwan’s stock market, however, the IPO frenzy of these mega-clients may not be entirely good news.
The sheer scale of capital absorption by these offerings could crowd out global liquidity. “The capital markets will be drained dry, and the speculative frenzy will end,” said one venture-capital executive with a semiconductor background.
“Shortage” to “Price Surge”
Another critical theme for next year is how the memory-price surge that began in the second half of this year will ripple through the broader industry.
If the defining feature of Taiwan’s AI supply chain in 2025 was “shortage”—tight 3-nanometer capacity at TSMC, shortages of low-CTE fiberglass cloth and high-end copper-clad laminates constraining Nvidia GPUs and Google TPUs, and sending related stocks soaring—then 2026’s keyword will be “price.”
Since the second half of this year, capacity crowd-out from AI-specific HBM (high-bandwidth memory) has created severe supply gaps, triggering astonishing price spikes in the spot market.
For DRAM, mainstream DDR4 prices are up fourfold year-on-year, while older DDR3 prices have surged more than tenfold.
As this trend carries into next year, soaring memory procurement costs will make price hikes for PCs and smartphones unavoidable. Acer and Asus have recently confirmed they will “moderately reflect costs” in pricing.
Foreign media have also reported that even Apple—the buyer with the greatest scale and bargaining power—can no longer fully absorb the increases. Prices for next year’s iPhone 18 lineup are expected to rise by at least 5%.
As a result, the smartphone market, which only just showed signs of recovery this year, may slip back into contraction in 2026. Counterpoint estimates a full-year decline of 2.1%, with smaller brands—lacking pricing power—suffering the most, down an estimated 3.5%. These brands are typically key customers of MediaTek.
The biggest winners will be South Korea’s two memory giants, whose profits could even surpass TSMC’s—an outcome almost unthinkable just a few years ago.
According to forecasts from Samsung Securities and Mirae Asset Securities, Samsung Electronics and SK Hynix could post operating profits of 93.6 trillion won and 91.1 trillion won, respectively, in 2026—roughly $65 billion and $63 billion.
That already exceeds most estimates for TSMC’s 2025 operating profit and could even surpass TSMC’s 2026 results if memory prices surge further.
It is worth remembering that at the peak of the previous memory upcycle in 2021, Samsung’s operating profit was twice that of TSMC. That was only four years ago.
China’s “15th Five-Year Plan”
Another key development to watch next year will be China’s annual “Two Sessions” in March, where Beijing is set to unveil its 15th Five-Year Plan, covering the period from 2026 to 2030.
My greatest interest naturally lies in the AI component. In the policy recommendations released last October, the term “artificial intelligence” appeared eight times, explicitly calling for the full implementation of the “AI+” initiative.
The document states that AI should lead a transformation in scientific-research paradigms, deepen integration with industrial development, seize commanding heights in AI applications, and comprehensively empower all sectors of the economy.
“AI+” clearly echoes the “Internet+” strategy of the 13th Five-Year Plan. That initiative—using the internet to drive traditional industries—was enormously successful.
ByteDance and Pinduoduo（ 拼多多） both rose to global prominence during that period, and China’s mobile-payment penetration has reached 86%, the highest in the world.
Policy documents show that China plans to advance AI in three stages over the next decade. By 2027, AI is to be widely applied across six areas: scientific research, industrial processes, consumer goods, healthcare and education, digital government, and technology exports.
By 2030, AI is expected to be as ubiquitous as electricity or the internet, becoming a major engine of economic growth. By 2035, China aims to build a “smart society,” with AI reshaping culture and human interaction much as the internet once did.
During the second year of the 13th Five-Year Plan, in 2017, China’s internet industry experienced a period of “wild growth” that captured global attention.
China-style innovations such as bike-sharing expanded aggressively overseas, even landing in Taiwan and causing a stir. Will Taiwan see a similar wave of Chinese AI applications during the 15th Five-Year period?
One AI executive who has lived in China for many years told me that the earlier “wild growth” stemmed largely from another pillar of the 13th Five-Year Plan: “mass entrepreneurship and mass innovation.”
Government encouragement unleashed a flood of hot money. “Early-stage funding was easy,” he said. “With just a pitch deck and a demo, you could raise a round.”
It also fostered a “traffic-first mihttps://cwnewsroom.substack.com/p/2026-tech-megatrends-ipo-shockwaves-ai-inflation-samsung-profit-surge-pressuring-tsmc?ndset”—burn cash to scale users first, figure out monetization later.
That atmosphere has now shifted dramatically. Amid China’s economic slowdown and the U.S.–China tech war, AI startups are no longer willing to pursue the kind of reckless investment and explosive growth seen in the past. “The first question now is: how do I survive on my own?” he said.
Hot money is instead flowing into “chokepoint” sectors shaped by the tech conflict, such as AI chips and semiconductor equipment. The recent IPOs of domestic GPU players Moore Threads and MetaX—two of China’s so-called “GPU Four Tigers,” both commanding market capitalizations exceeding RMB 200 billion—are a case in point.
Liang-rong Chen
cwnewsroom.substack.com
