CoverStory:AIDrivesMarketsasValuationsRaceAheadofEarnings

Stock markets around the world surged to record highs in April, defying geopolitical tensions, volatile energy prices and uncertainty over interest rates, as investors crowded into a small group of artificial intelligence (AI) and semiconductor giants.

The rally has been strikingly narrow. Rather than broad-based corporate strength, gains have been powered by a handful of mega-cap companies, masking weaker momentum across the rest of the market.

In the United States, the technology sector contributed nearly 80% of the S&P 500’s estimated 15.1% year-on-year earnings growth in the first quarter. The market has effectively split in two: the Magnificent Seven — Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia and Tesla — and the remaining 493 stocks.

The appetite of investors for AI appears to be far from its peak. Space exploration firm SpaceX is preparing for what could be the largest initial public offering in history, while large language model (LLM) pioneers OpenAI and Anthropic are planning to go public by the end of the year or early 2027.

China’s markets are moving in a similar direction but with a different underlying rhythm. In the Chinese mainland A-share market, an AI hardware rally has rotated through computing power, optical modules, domestic chips and humanoid robots. In a bid to foster innovation, regulators overhauled listing standards for the ChiNext board in April, opening the door to unprofitable hard-tech firms.

Meanwhile, Hong Kong has become a launchpad for AI startups, pushing fundraising to a five-year high in the first quarter. The enthusiasm has spilled into the primary market, with Chinese venture capital and private equity funding surging past 400 billion yuan ($59 billion) in the first quarter alone.

Yet beneath this coordinated capital rush, a critical debate is emerging: Is this a legitimate re-rating of the technology sector, or an overstretched bubble? And as valuations soar, how can investors distinguish between companies with genuine AI capabilities and those merely riding the hype?

Bubble echoes in an AI market

On April 27, the S&P 500 closed at a record 7,173.91, up nearly 30% from a year earlier, while the Nasdaq Composite touched an intraday all-time high. On the same day, Japan’s Nikkei 225 surged to a record 60,537.36, and South Korea’s KOSPI reached a new peak of 6,615.03.

Upward revisions to the S&P 500’s earnings expectations came almost entirely from the Magnificent Seven technology giants, according to Torsten Slok, chief economist at Apollo Global Management. He said the benchmark has effectively become “an AI index” because of its lack of diversified growth drivers.

Mega-cap technology stocks have become the all-weather trade of the current bull market, with capital crowding into sector leaders regardless of broader macroeconomic conditions. A late-April report by JPMorgan said the S&P 500’s rally had begun well before geopolitical risks became clear, suggesting the surge was driven by confidence in AI fundamentals rather than clarity on war trajectories.

Yet when an asset class is widely viewed as a safe haven capable of both offense and defense, valuations risk becoming detached from reality. The prevailing justification for the AI bet — that massive investments in AI infrastructure will guarantee sustained profit growth for years — is largely untested.

The current frenzy has drawn comparisons with the internet bubble of the late 1990s. The Shiller cyclically adjusted price-to-earnings (CAPE) ratio recently climbed to between 39 and 40 — a level surpassed only once in its 155-year history, just before the dot-com crash in March 2000. A Deutsche Bank survey found 57% of institutional investors consider an “AI valuation collapse” the biggest market risk for 2026.

Bulls argue that today’s technology giants, unlike the cash-burning startups of 2000, boast median net profit margins approaching 30%. Alphabet, Microsoft, Amazon and Meta all reported strong earnings in late April, with Alphabet’s large backlog pointing to durable future revenue. Moreover, modern AI infrastructure is funded primarily through corporate cash flow and equity, limiting risks of contagion to the banking system.

However, such analysis cannot avoid a central question: How much future earnings growth has already been priced in? According to Sequoia Capital, every $1 invested in GPU infrastructure requires about $4 in application-side revenue for the investment chain to be commercially viable. Nvidia’s data-center revenue in 2025 alone implies the market needs to generate roughly $775 billion in application revenue. Yet actual revenue was only $150 billion to $200 billion, leaving a gap of around $600 billion.

Skeptics question how much future growth has already been priced in. Jeremy Grantham, co-founder of GMO who correctly warned of the 2000 and 2008 crashes, recently cautioned that the current AI boom is highly likely to end in a collapse. He said the AI-driven rally could be the market’s final surge before a downturn.

“Do not assume market prices are rational just because no serious institution is telling you to sell,” Grantham said, criticizing institutional managers who follow the herd rather than risk their careers by betting against a bubble.

China’s hard tech frenzy

In China, the AI investment thesis is shaped as much by geopolitics and policy as by profits. In late April, optical chipmaker Yuanjie Semiconductor Technology Co. Ltd. briefly overtook Kweichow Moutai as the most expensive A-share stock, capping a roughly 1,500% surge over the past year.

Investors are increasingly embracing a “Chinese-style” tech valuation framework, arguing that strategic industries with long research and development cycles should not be judged by traditional price-to-earnings metrics. Themes such as domestic substitution for Nvidia, embodied AI and sovereign computing power are driving valuations in their own right.

Wu Yongqiang, an investment strategist at Bank of East Asia, said the intensifying AI race between China and the U.S. has galvanized sentiment. Investors see American AI stocks soaring while Beijing elevates AI to a policy priority, reinforcing expectations of sustained demand and earnings visibility. At the same time, geopolitical risks tied to the Middle East are largely concentrated in energy, making the AI supply chain, which is relatively insulated from oil-price swings, a perceived haven for capital.

An investor checks real-time quotes for Yuanjie Semiconductor Technology Co. Ltd. The stock briefly surpassed Kweichow Moutai during intraday trading to become the most expensive A-share on April 17, 2026

Policy and capital are reinforcing each other. By the end of April, Shanghai’s STAR Market hosted more than 600 hard-tech firms, including 15 with market capitalizations above 100 billion yuan. Technology companies now account for more than 40% of listings on the Shanghai main board.

The IPO pipeline underscores this shift. In the first quarter, 30 companies went public on the A-share market, raising 25.88 billion yuan, up 57% from a year earlier, with hard tech leading issuance. Listing reviews are accelerating as well, with 46 companies scheduled for hearings across the Shanghai, Shenzhen and Beijing exchanges, compared with only eight a year earlier.

Regulators are also loosening the rules. In April, the China Securities Regulatory Commission unveiled reforms to Shenzhen’s ChiNext board, allowing qualified, unprofitable tech firms to list. The new standards set thresholds based on market value, revenue growth and research spending — but drop the requirement for profitability.

Markets have moved ahead of the rules. The ChiNext Index hit an intraday high of 3,785.23 on April 23, its strongest level in nearly 11 years. Funds have piled into optical-communication names such as Zhongji Innolight Co. Ltd. and Eoptolink Technology Inc. Ltd., betting that American AI infrastructure spending will lift Chinese suppliers.

That crowding is beginning to worry some investors. A senior fund manager said that while the AI story is credible, earnings may lag behind expectations, warning that a 30% to 50% correction would not be surprising if overseas demand weakens.

Hong Kong’s IPO rush

Hong Kong is capitalizing on the hard-tech boom, sustaining momentum into 2026 after reclaiming the top spot for global IPO fundraising last year. Developers of AI models, chips and robotics are flocking to the market.

Caixin calculations show that as of April 24, 29 of the 45 companies listed on the Hong Kong Stock Exchange this year were in AI and healthcare, giving the IPO pipeline a distinctly hard-tech tilt. HKEX disclosures indicate 413 companies were waiting to go public as of late March.

Post-listing performance has been strong. From the start of the year through April 27, new listings delivered an average first-day gain of 37% and an average cumulative rise of 71%. Fourteen have more than doubled, including Zhipu AI, MiniMax and Manycore Tech Inc.

The influx is reshaping valuation dynamics. The long-standing premium of A-shares over H-shares has narrowed to a seven-year low, with companies such as Contemporary Amperex Technology Co. Ltd. and Montage Technology Co. Ltd. trading at a premium in Hong Kong. Middle Eastern sovereign wealth funds are also increasing allocations, taking anchor stakes in firms such as MiniMax.

Goldman Sachs said liquidity remains supportive. In an April 19 report, the bank projected Hong Kong equity financing would reach $110 billion in 2026, including $60 billion from IPOs and $50 billion from secondary offerings. It cited strong corporate payouts, increased allocations from global long-term investors and continued southbound inflows from the Chinese mainland.

Zhipu AI Chairman Liu Debing (right) strikes the gong at the Hong Kong Stock Exchange during the company’s listing ceremony on Jan. 8, 2026, in Hong Kong

The surge is exposing regulatory gaps. In March, Hong Kong’s Independent Commission Against Corruption and the Securities and Futures Commission launched a crackdown codenamed “Fuse” on alleged insider trading linked to IPO placements. Market participants say some syndicates are cornering allocations to drive up debut prices before selling into retail demand.

Venture capital piles into AI

The frenzy in public markets is spilling into venture capital. In the first quarter, China’s venture capital and private equity markets logged nearly 3,200 deals, up 33% from a year earlier, with total investment exceeding 400 billion yuan.

Much of the activity is concentrated in hard tech. The electronic information sector led with more than 1,000 deals and 141.5 billion yuan in funding, followed by advanced manufacturing and healthcare. Within that, semiconductors, AI and robotics drew the most interest, according to data from CVSource.

Humanoid robots and embodied AI have emerged as standout themes. Investment in robotics has surged sharply over the past two years, with deal value jumping from 3.5 billion yuan in early 2024 to 46.5 billion yuan in the first quarter of 2026. Capital is flowing across the value chain, from core components to AI systems and downstream applications.

Funding rounds are getting larger. Since the start of the year, more than 20 Chinese robotics firms have raised over 10 billion yuan in total. In April, embodied AI startup TARS raised $455 million in a Pre-A round led by Hillhouse Investment and HongShan Capital Group, one of the largest early-stage financings in the sector.

Exit expectations are also changing behavior. With broader listing channels in Hong Kong and on the ChiNext board, investors are placing greater weight on technical capabilities and product architecture, often ahead of traditional metrics such as cash flow.

Fundraising is accelerating in parallel. Nearly 1,900 new funds were launched in the first quarter, almost doubling from a year earlier, with participation by financial institutions rising sharply. Banks, insurers and securities firms are emerging as major sources of capital, alongside state-backed investors.

Yet, the rush for scale is beginning to distort incentives. Some investors say startups are prioritizing fundraising optics over product viability, raising concerns that parts of the market may be running ahead of sustainable demand.

Separating real AI from hype

As capital floods the sector, a divide is emerging between genuine AI beneficiaries and speculative imitators, said a senior investor. Broadly, winners fall into two camps: hardware firms with tangible earnings tied to AI demand, and software companies with competitive technology but valuations that far exceed current revenue. The rest are companies adopting AI labels with little underlying business exposure.

Hardware makers have delivered some of the biggest — and most volatile — gains. Chipmaker Cambricon swung to profit in 2025 as domestic cloud providers turned to local suppliers amid U.S. export controls, though growth has begun to slow. Optical module leaders Zhongji Innolight and Eoptolink Technology have secured large overseas orders, but early signs of plateauing growth point to rising price competition. Fiber-optic producer Yangtze Optical Fibre and Cable Joint Stock Limited Company has also seen a sharp revaluation, driven by demand from AI data centers.

A booth by TARS is seen at the 2026 Appliance & Electronics World Expo (AWE 2026) on March 13, 2026, in Shanghai

In software, large language model developers such as Zhipu AI and MiniMax are posting rapid revenue growth, with MiniMax generating more than 70% of its sales overseas. Rivals including Moonshot AI and Stepfun are preparing for Hong Kong listings. Yet heavy spending on model development means many remain deeply unprofitable, leaving their path to sustainable earnings unclear.

As the initial euphoria fades, investors are bracing themselves for a shakeout, the investor said. Companies riding the AI narrative without clear commercial traction are likely to be the most vulnerable if sentiment turns.

Qu Yunxu contributed to the story

Contact reporter Han Wei (weihan@caixin.com)