By Kimberley Kao and Sherry Qin

Asian stock markets were locked in a tug-of-war in the first half of the year caught between Middle East tensions and the artificial-intelligence boom that laid bare a chasm between AI haves and have-nots.

But cracks are appearing in the AI trade, whiplashing tech-heavy indexes as doubts about the profitability and sustainability of the AI buildout creep in.

South Korea, Taiwan and Japan--clear beneficiaries of the world's appetite for the components that power AI-- were Asia's top three performing markets during the first half, shrugging of fears that an oil shock due to the Iran war could derail energy-importing economies.

Market "performance has increasingly depended on whether a market sits close to the AI supply chain," said Charu Chanana, chief investment strategist at Saxo.

Those AI winners now face a reality check as investors start demanding proof that the huge sums poured into AI will yield actual profits and productivity gains.

Though other factors like corporate-governance reforms have also buoyed indexes, the staggering amount investors have poured into AI and tech means the buzziest stocks are heavily owned and vulnerable to profit-taking.

Over-concentration in Asia's AI superstars is causing concern that investors are putting all their eggs into one basket.

"While fundamentals remain supportive, the sheer concentration of investor exposure increases the risk of sharper swings," said Angela Cheng, head of research at CGS International.

Only a handful of stocks are driving the bulk of gains in South Korea and Taiwan: Memory makers Samsung Electronics and SK Hynix, alongside contract chip manufacturer TSMC. They are the only Asian companies in to the $1 trillion dollar club, all with close ties to AI chip juggernaut Nvidia.

Slumps in shares of Samsung and SK Hynix, which together make up around half of the Kospi index's total weighting, have triggered multiple trading halts from authorities to curb volatility.

"A sharp move in either name drags the whole index with it before the other roughly nine hundred listed companies get a say," said Zavier Wong at eToro.

AI and chip-related companies have experienced similar swings in Japan.

Tech conglomerate SoftBank--a proxy for the AI trade-- Advantest and Kioxia, to name a few, have oscillated sharply as market sentiment flits between euphoria over the technology's long-term promise and pessimism over monetization and sky-high valuations.

Even with volatility rising, the three indexes have posted multiple record highs. The Kospi more than doubled its value over the first half, while the Nikkei rose 39% and the Taiex jumped 59%.

It remains to be seen if they sustain that in the second half.

Lorraine Tan, director of equity research for Asia at Morningstar, sees the outlook as challenging and probably still volatile. As the tech sector is largely fully valued, investors could go to other sectors with more reasonable upside, such as healthcare, Tan said.

The AI divide is visible in China's onshore markets too. While AI-enabers like optical module makers Zhongji Innolight and Eoptolink Technology sent the ChiNext Index 36% higher, the Shanghai Composite Index grew a mere 3.2%.

Last year, Hong Kong had its best year since 2017. That hasn't been the case so far in 2026. Despite a torrent of listings, many by tech and AI firms, Hong Kong's Hang Seng Tech Index shed 19% in the first six months.

Drawn by the record-setting rallies elsewhere, investors have pulled money out of heavyweight Chinese internet stocks listed in the city. A push factor has also come from concerns over AI investments and China's subdued consumption.

Hong Kong's market lacks exposure to AI infrastructure, DBS Group Research analysts said in a note. Although valuations are cheap, DBS expects global investors to remain focused on infrastructure and hardware elsewhere.

Despite the unevenness, analysts still see plenty of catch-up opportunities in the region, where AI development is still in its infancy.

The first round of AI-led rallies centered around the infrastructure names. The next leg could be about finding "AI efficiency winners," the companies and markets that can help reduce AI costs or broaden adoption, Chanana said.

Write to Kimberley Kao at kimberley.kao@wsj.com and Sherry Qin at sherry.qin@wsj.com