Nvidia and SK Group announce $500 billion strategic AI infrastructure expansion partnership spanning AI factories and next-generation memory.

The annual World Artificial Intelligence Conference along Shanghai's Huangpu River has just concluded in midsummer. Autonomous driving, AI agents, embodied AI, computing chips, industrial intelligence—new technologies continue to emerge, and enthusiasm for the AI supply chain remains unabated.

Yet standing in stark contrast to the fervor on the industrial side is the capital markets' swift depression of the "pause button." On July 24, storage leader Deming Technology (001309.SZ) dropped another 8.07%. From its peak of 980 yuan per share, the stock has plummeted to 399.89 yuan, a decline of 59.19%.

On July 15, 2026, a screenshot circulated widely through investment circles. It showed a paying member in economist Ren Zeping's VIP group complaining of a fully leveraged position in Deming and Jiangbo Dragon (301308.SZ) that triggered a forced liquidation, with losses exceeding ten million yuan.

Since July, global tech stock pullbacks have intensified. In A-shares, as of July 24, the Sci-Tech 50 index has fallen 20.75% from its peak, with numerous tech stocks retreating markedly. Cambricon has fallen 24.38%, Corerain Technologies 26.14%, and Hisense Newbrand 23.08% from their highs—semiconductor, storage, and chip indices have seen declines exceeding 30%.

Globally, the Philadelphia Semiconductor Index and South Korea's storage "dual leaders" have suffered sharp corrections. SK Hynix, for example, plummeted another 8.34% on July 24, falling from 2,987,000 won per share at its peak to 1,759,000 won, a drawdown of 41.11%.

After a year of frenzied pursuit, whether American tech giants or China's AI supply chain—from computing power and optical modules to servers, large language models, and application software—numerous sectors have begun experiencing severe volatility. The era when "anything labeled AI would rise" is fading. The market is beginning to recalculate the true value of enterprises.

Some economists warn that the market's sharp correction is largely due to excessively concentrated trading, particularly when certain investors have added leverage to these sectors, creating overcrowded positions.

Despite continued intensity in AI technology investment enthusiasm, the secondary market has begun revaluing AI investments. Capital markets' first wave of fervor over AI is gradually ending; a second round of pricing around AI's value is quietly beginning.

After OpenAI ignited the generative AI revolution, global tech giants—Microsoft, Google, Meta, Amazon—launched an unprecedented capital expenditure race. GPU supply fell short of demand, HBM storage prices kept climbing, and the entire supply chain of servers, liquid cooling, optical modules, and PCBs benefited broadly.

The domestic market mirrored this pattern. From Cambricon and Corerain to storage leaders and Deming, then robotics, autonomous driving, and large language model applications—nearly anything AI-related has attracted investor capital.

Many companies' price-to-earnings ratios have repeatedly broken historical records. Even enterprises without stable profitability have seen market values grow hundreds of billions on the strength of an AI narrative alone.

Many venture capital professionals reported that in the first half of 2026, numerous AI startups saw valuations marked up on-site. An increasingly broad consensus took hold: AI represents the greatest industrial opportunity of the next decade.

Yet this very consensus created overcrowded trades. The Sci-Tech 50 has cumulatively retreated 20.75% from its peak; the Philadelphia Semiconductor Index faces continued adjustment, South Korea's storage leader SK Hynix has retreated over 40% from its peak, and core supply chains—AI servers, storage, computing—all show clear capital outflows.

A public fund manager observed that this adjustment stems from two factors: overconcentrated tech trading and capital markets beginning to recalculate enterprise value.

Over the past year, the market has been willing to pay premiums for growth over the next decade in advance. As long as enterprises possess an AI concept, regardless of whether earnings have fully materialized or business models remain exploratory, capital has been willing to assign high valuations—because investors were buying imaginary space. Today, this valuation logic is quietly shifting. The market is re-examining enterprises' genuine competitive strength, paying greater attention to whether orders continue growing, profits can materialize, cash flow improves, and technological advantages can translate into commercial value. Investor focus has shifted from "Do they have AI?" to "How much money has AI made?"

Stanley Druckenmiller, regarded as a legendary investor, bought heavily into NVIDIA during the generative AI wave's early days and reaped handsome returns from rapid appreciation. But as NVIDIA's valuation climbed steadily, he chose to gradually reduce his position, publicly stating that his selling didn't reflect dimmed AI confidence, but rather a conviction that markets have already front-loaded years of optimistic expectations. He stated plainly: artificial intelligence will be a revolutionary technology that changes the world, but even the greatest companies require reasonable prices.

Domestic investors have demonstrated similar thinking. Facing the AI wave, a prominent venture capital figure said he has begun re-examining AI infrastructure companies like NVIDIA and, after deeper understanding of industry logic, continuously adjusts his positions. He noted that AI's potential changes may exceed many people's imagination, but truly worthy of investment are not all AI-concept companies, but rather those possessing technological moats, clear business models, and capacity to continuously generate free cash flow.

This unmistakably transmits one signal: AI industry's long-term trend remains unchanged, but capital markets have entered a new pricing phase. Previously, the competition was over who possessed AI concepts; going forward, it will be about who can deliver AI value.

In 1999, internet bubble swept the globe. Simply adding ".com" to a company name could send stock prices soaring. Investors believed the internet would change the world—a conviction later proven entirely correct.

After the bubble burst in 2000, the Nasdaq index fell nearly eighty percent. Countless star companies collapsed and investors suffered massive losses. Many concluded the internet era had ended.

Yet the companies truly destined to change the world—Amazon, Google, Meta—almost all emerged or flourished after the bubble burst. Capital eliminated not the internet, but internet companies lacking competitive advantage.

From 2020 to 2021, solar and electric vehicles surged ahead, with leaders like CATL and Longi Green Energy seeing valuations continuously climb, while numerous cross-sector companies announced entry into the new energy industry.

Subsequently, the industry experienced years of deep adjustment. Looking back today, true survivors remain those with technological moats, scale advantages, and sustained R&D capacity. Companies relying on concept speculation have mostly faded from market view.

Every industrial revolution passes through three clear cyclical phases: technological breakthrough ignites valuation, capital fervor inflates bubbles, and finally value regression and consolidation of leaders.

AI is no exception. Today's correction is not the end of the AI story, but a necessary passage from "believe everything" to "believe what."

Economist Ma Guangyuan believes that during AI revolution, many companies deemed greatest today will likely cease to exist within a decade. This is the defining characteristic of every technological revolution. The most glorious at the start aren't necessarily the ones that endure; instead, many of the truly great enterprises emerge slowly after bubbles burst. The internet revolution exemplifies this perfectly. Many star companies at the peak of the internet bubble have disappeared today; yet truly value-creating firms—Apple, Microsoft, Amazon—entered their sustained growth phase only as the industry matured, business models solidified, and profits materialized. "In every technological revolution," Ma states, "the real big money-makers are not necessarily the fastest risers in the first wave, but those who ultimately convert technology into productivity and technique into profit."

From the primary market, AI and tech-concept investment remain the absolute "main battlefield."

From the secondary market, as margin balances have declined rapidly, UBS Securities China equity strategist Meng Lei believes A-share deleveraging may be nearing completion. As of July 20, 2026, total A-share margin balances stood at 2.70 trillion yuan, falling 312.4 billion yuan from the historical peak of 3.01 trillion on June 25.

By sector, the heavily-watched big tech bloc's margin balance declined from 1.1 trillion on June 26 to 944.9 billion on July 20, a sharp drop of 152.7 billion. Regarding leverage concerns, the margin balance ratios for big tech, ChiNext, and Sci-Tech boards relative to free-float market cap align closely with overall market levels.

Furthermore, signals from the World AI Conference prove more optimistic than the capital markets. Global tech giants continue increasing AI investment. Microsoft persistently expands global AI data centers; Meta further raises artificial intelligence capital expenditure budgets; Google continuously strengthens Gemini ecosystem; domestic internet platforms similarly advance large language model commercialization. In other words, the industry hasn't cooled.

What has truly cooled is capital markets' sentiment. Previously, the market granted high valuations to all AI companies; henceforth, high valuations belong only to those truly creating value.

This signals AI investment logic has undergone profound change. Previously, the competition was who tells better stories; henceforth, who delivers results. Previously, we watched concepts; henceforth, orders. Previously, expectations; henceforth, profits. Previously, fundraising capacity; henceforth, free cash flow.

Capital markets increasingly understand that what truly determines an enterprise's long-term value has never been the trend, but sustained capacity to create value.

Indeed, no great industrial revolution unfolds smoothly. Railways, electricity, automobiles, the internet, smartphones, new energy—all experienced severe adjustments and made countless investors swing between euphoria and fear. Yet history ultimately proved that what endures cycles is never sentiment, but the industry itself. Today's AI faces this very moment.

JPMorgan Chase judges recent A-share AI-themed board corrections as fundamentally leverage purges, not fundamental deterioration signals. China's AI ecosystem's long-term investment logic remains intact.

JPMorgan Chase China equity strategist Zhang Xiaoning stated on July 15 that he rejects the "imminent bubble burst" market narrative, supporting this view with three grounds: healthy balance sheets, continuously improving large language model capabilities, and AI hardware supply bottlenecks remaining difficult to resolve near-term. On liquidity, IT sector margin trading volume has fallen from roughly 12% medium-term peak to 8%-9%, indicating the highest-leverage segments have largely exited forced liquidation, with deleveraging essentially complete.

Fervor will fade, bubbles will burst, but technologies truly destined to change the world often begin realizing their value precisely after bubbles burst. The market is searching for authentic AI.