China's persistent deflation constrains market broadening

Divergence from broader macro trends 

At the recent Central Economic Work Conference, top officials reaffirmed their commitment to the so-called “anti-involution” campaign, aimed at stamping out destructive price wars that have eroded profit margins across sectors such as electric vehicles and food delivery. However, progress has been limited, as government concerns over potential job losses and slower economic growth have restrained aggressive policy actions.

Despite these macroeconomic headwinds, China’s equity markets achieved double-digit returns in 2025. This divergence from broader macroeconomic trends can be attributed to strong performance in sectors like information technology, driven by breakthroughs in artificial intelligence (AI); biotechnology; and industries benefiting from anti-involution initiatives.

Additionally, improved liquidity has supported the market’s re-rating, as savings have flowed back into equities, attracted by dividend yields that are more appealing relative to deposit rates. Meanwhile, fixed income returns have declined, volatility has increased, and with the property market remaining weak, investors are seeking alternative investment avenues.

Going into 2026, weak private sector confidence, subdued consumer sentiment, and supply/demand imbalances are increasingly challenging factors for reflation, and ultimately for corporate earnings. Reviving domestic demand is essential for sustained long-term growth, but redirecting China toward higher levels of consumption will take time. For now, policy remains focused on investment-led and trade-driven growth, emphasising the development of a modern industrial system and technological self-sufficiency. As such, investor focus should stay on areas supported by policy and technological innovation. 

* The GDP deflator is an economic metric used to measure the general level of price changes (inflation or deflation) for all domestically-produced final goods and services in an economy. It is calculated by dividing nominal GDP (current prices) by real GDP (inflation-adjusted/constant prices) and multiplying by 100.

It’s all about AI infrastructure, for now

Global stock markets enjoyed a positive start to 2026, continuing their upward trend from the previous year, primarily fuelled by growing momentum in AI development. Market expectations suggest AI-related spending will surpass 2025 levels as more companies adopt AI technologies. Some estimates project that hyperscalers, the large-scale cloud and data centre providers, will spend approximately US $600bn in 2026, up from around $470bn in 2025.

Notably, the recent rally has been driven less by the companies developing the core AI technologies themselves and more by “picks and shovels” — the suppliers of the essential infrastructure and tools supporting AI growth. These include semiconductor manufacturers, hardware providers, as well as providers of cloud computing platforms, all of which are expected to benefit from the increasing capital expenditure associated with AI development.

Investor attention is currently focused on companies that supply the computational infrastructure needed to build AI ecosystems. Conversely, companies involved in AI software and accelerators have seen more subdued re-ratings, as their applications are still in early commercialisation stages and face higher growth uncertainties.

The ongoing buildout of this complex infrastructure is likely to encounter several bottlenecks, including limitations in computing power capacity, data centre availability, and the need for low-cost, reliable energy sources.

At this stage of development, it is the computing power capacity that is benefiting from a severe supply/demand imbalance, driving prices upward. For example, research provider TrendForce estimates that average Dynamic Random Access Memory (DRAM) prices increased by between 50% and 55% in the fourth quarter of 2025, with orders for 2026 already exceeding available capacity. 

DRAM is a crucial component, because it acts as the high-speed “working memory” of a computer, providing rapid access to the data needed for active processing.

In addition, AI data centres consume enormous amounts of power, and the new DRAM models that are being developed reduce energy consumption during data movement by 40% to 70%, which is vital for scaling AI “superfactories” within existing power grids.

The hardware shortage is expected to continue until at least 2027, according to some estimates, as it takes two to three years to construct new compute manufacturing plants. This prolonged supply constraint is likely to benefit semiconductor companies more than other technology sectors, not just in the US, but globally; notably, with many of the supply chains still residing in Asia.