The AI trade is shifting from chips to Chinese platforms

Perhaps most interesting question in the AI trade is no longer whether the theme is real. It is where the market has already paid too much for it, and where the next layer of value may be hiding.

The difference matters because client portfolios are increasingly exposed to AI through the most obvious beneficiaries: chipmakers, data-centre infrastructure, hyperscalers and the small group of companies already rewarded for the boom.

Caroline Cai, chief executive officer at Pzena Investment Management, is making a different point. Her argument is not that the first wave of AI winners is broken. It is that some of those investment cases have matured, while other, less loved parts of the market may now offer better asymmetry.

In particular, she is looking at Chinese digital platform companies such as Alibaba and Tencent, where the market is still cautious, but the embedded AI opportunity may be underpriced.

From hardware winners to platform laggards

The rotation is important. Cai funded part of the move by trimming Samsung and TSMC. Both have been central to the AI hardware and semiconductor supply chain. In Samsung’s case, she said the thesis played out sooner than expected as memory chip prices soared on AI demand.

A good business riding a powerful trend can still become a poor investment. Valuation discipline means asking not only whether demand is strong, but whether the share price already reflects it.

Cai’s pivot is therefore less about abandoning AI hardware and more about looking for the next mispriced expression of the same theme. Chinese technology stocks have not enjoyed the same broad investor enthusiasm.

Tencent and Alibaba have faced concerns around competition in their platform businesses, while Chinese technology shares listed in Hong Kong remain almost 25 per cent below their 2025 peak, according to the source material.

For value investors, that scepticism is precisely the point. Cai said Chinese digital platform companies “look inexpensive” and offer more upsides given their potential to transform everyday life.

Parts of Chinese technology still trade at a discount. Yet the same companies are investing in AI applications across commerce, payments, cloud, advertising, gaming and consumer services.

Paying less for the productivity option

“You’re not paying a lot for that possibility that AI does bend the curve on productivity.”

– Caroline Cai, chief executive officer, Pzena Investment Management

That is a different proposition from paying peak multiples for companies already treated as AI winners. It is a call option on productivity, but without the same valuation burden. That distinction is useful in client conversations.

AI exposure does not need to mean buying the most expensive names in the most crowded part of the market. It can also mean owning companies where the market is not yet giving full credit for the possible earnings uplift.

This is particularly relevant in the early stages of a technology cycle, when it is still difficult to know which companies will ultimately capture the economics. Cai believes this kind of strategy is best placed to pay off before the eventual winners and losers become clear. This uncertainty can be uncomfortable, but it is also where valuation matters most.

The capital spending backdrop also supports a more nuanced view. US technology leaders are committing enormous sums to data centres and computing power. Four of the largest are forecasting combined capital expenditure of around US$650 billion in 2026, mainly on new infrastructure and related equipment.

By contrast, leading Chinese internet firms including Alibaba, Tencent, Baidu, JD.com and Meituan are expected to spend more than US$240 billion by 2030, while collectively holding US$224 billion in cash reserves.

The comparison is not simply about who spends more. It is about the return on that spending. A lower capital intensity model may require a lower hurdle to create shareholder value, particularly if AI can be embedded into existing platforms with large user bases.

The monetisation question

Cai’s argument ultimately comes back to monetisation. The AI capex race in developed markets is often framed around model quality, cloud infrastructure and data-centre scale. Chinese platforms may instead offer a more application-led path, using AI inside services that already reach hundreds of millions of users.

“When you look at the quality of the model that’s coming out and the focus on applications, it might be a more interesting way of monetizing AI than what we see in the developed world,” Cai said.

The next phase of AI investing may not belong to whoever builds the largest model or spends the most on infrastructure. It may belong to whoever embeds AI into daily life, improves productivity and converts existing ecosystems into higher-value platforms.

None of this removes the risks. Chinese platform companies still face competition, regulatory uncertainty, geopolitical sensitivity and investor scepticism. There is also no certainty that AI applications will bend the productivity curve quickly enough to justify even discounted valuations. But that is precisely why the opportunity sits in the value bucket rather than the momentum bucket.

The practical message is that AI is becoming a rotation story as much as a growth story. The first wave rewarded hardware and infrastructure. The next may reward companies where the market has not fully priced the productivity option.

Cai’s move from Samsung and TSMC toward Alibaba and Tencent says nothing about confidence in AI. It says everything about the discipline of not overpaying for it.