The sharp reversal in AI-linked technology stocks has rattled investors, but Sahil Mahtani (pictured), director of the Investment Institute at Ninety One, believes the sell-off says more about market positioning and leverage than the long-term prospects for artificial intelligence.
Morgan Stanley’s long-short index of technology momentum fell 40% in less than a month, its worst stretch in the 27-year history of the index. Goldman Sachs’ high-beta momentum pair was about a third below its recent high.
The broader market barely moved over the same period, with the S&P 500 slipping less than 1%.
“We’ve witnessed one of the sharpest momentum reversals on record, on the reversal of many retail trades earlier this year. This has happened even as the broader index has been quiescent,” Mahtani said.
The sell-off has been amplified by the unwinding of leveraged positions, particularly among retail investors in East Asia.
The proliferation of single-stock exchange traded funds offering as much as five times daily leverage left some markets vulnerable to sharp reversals. More than 10 such vehicles were trading on SK Hynix alone, while assets in leveraged ETFs have contracted substantially in recent weeks.
Retail investors had been important drivers of some of the strongest momentum trades over the past year. As leverage has unwound, selling has intensified.
Mahtani expects corrections of this scale to run their course only once excess leverage has been cleared from the system. Some of the strongest retail momentum occurred in early July, but many of the stocks that had been investor favourites have since fallen sharply.
SpaceX is trading below its IPO price, while rare earth and quantum computing stocks have given back previous gains. Gold and silver also failed to rally despite softer-than-expected inflation data.
Open models change the AI equation
The market correction comes as the AI investment story enters another phase, with open-weight models gaining ground against closed-weight frontier models.
On Arena, an independent AI benchmark, Chinese models have overtaken the leading US models for the first time.
Moonshot’s Kimi-K3 is now the leading open-weight model, allowing companies to run and adapt it on their own infrastructure. Closed-weight models such as ChatGPT and Claude are accessed through their developers.
“The emergence of open-weight models transforms where value is likely to occur in the AI ecosystem, shuffling winners and losers,” Mahtani said.
At the World AI Conference in Shanghai, Chinese President Xi Jinping said China intends to make advanced models inexpensive and widely available.
If Chinese open-weight models can compete with or outperform US closed-weight models at lower cost, the economics underpinning the current scale of AI investment could change.
Mahtani points to China’s experience in industries such as solar, shipbuilding, and electric vehicles, where state backing has contributed to greater competition and pressure on margins.
The pressure could extend to US AI model developers. If companies such as Anthropic and OpenAI cannot build commercially viable businesses, the pace of future investment in AI could slow.
Open-weight models still require substantial computing capacity. Moonshot reportedly gated new sign-ups after demand exceeded available computing capacity.
Lower-cost AI could increase demand for processors, electricity, and data centres, while shifting value towards other parts of the technology supply chain.
A more competitive model layer could redistribute returns rather than eliminate them, with semiconductors, infrastructure, power, and data centres potentially benefiting as value becomes less concentrated among model developers.
What could derail the trade?
Mahtani identifies economic growth and monetary policy among the risks to the current AI-driven market narrative.
The July Bank of America survey found that 54% of fund managers expected a “no landing” for the global economy over the next 12 months, while only 2% expected a hard landing.
US real GDP growth is expected to reach 2.3% in 2026 and 2.2% in 2027, compared with global growth expectations of 3% and 3.4% respectively.
Markets are priced for resilience, leaving them vulnerable to a downside surprise in economic growth.
A more hawkish Federal Reserve is another potential risk. Earlier in the month, Fed governors Christopher Waller and Kevin Warsh had signalled a more hawkish stance, although subsequent inflation data provided little support for a significant inflation resurgence. Both CPI and PPI came in below expectations.
Tariffs, border enforcement, and the renegotiation of the United States-Mexico-Canada Agreement could add to inflationary pressure in coming months. Mahtani notes that the Trump administration has shown a willingness to adjust policy when economic conditions require it.
That leaves geopolitics as another variable for investors to watch.
A renewed escalation in the Middle East could have implications for energy markets. An oil shock linked to the Strait of Hormuz could give the Federal Reserve a reason to tighten monetary policy while economic growth was slowing.
From concentration to competition
Mahtani sees the recent correction as a change in the composition of the AI opportunity rather than the end of the trade.
“The AI trade isn’t over, but it is evolving. Investors now need to understand where value is being created across the AI ecosystem and be open to the fact that some parts of the ecosystem are going to become increasingly commodified.”
If powerful AI models become cheaper and more widely available, value may become less concentrated among a small group of model developers. The beneficiaries could instead be found across the infrastructure required to run AI at scale, from semiconductors and data centres to power and other technology inputs.
The question for investors is shifting from how much AI will grow to where the value created by that growth will accrue.
