Kenanga Research has on Friday downgraded Malaysia’s technology sector rating to neutral from overweight as external risks have intensified, particularly from global memory tightness, supply-chain fragility and persistent cost pressure across the value chain.
“All in, external risks to the sector — particularly global memory tightness — have intensified versus three months ago,” the research house said in a note.
Firstly, the US technology sector is likely to remain volatile in the near term. This is because of firmer oil prices and inflation concerns continue to weigh on risk appetite and valuation multiples, especially for high-growth names, despite a still-supportive earnings and capex backdrop for artificial intelligence (AI)-linked semiconductor, custom silicon and data-center infrastructure players.
“More importantly, any disappointment in AI adoption or returns could quickly spill over to tightly linked counterparties, amplifying downside risks and increasing the likelihood of valuation de-rating,” said Kenanga.
Secondly, supply-chain fragility remains a concern, particularly for critical materials and logistics, including packaging inputs such as substrates and leadframes, where geopolitical chokepoints could disrupt both front-end fabs and backend operations; recent Middle East tensions have also underscored the risk of knock-on effects through higher energy and freight volatility.
Thirdly, broad-based cost inflation and margin pressure persist across the value chain, driven by global memory shortages, foreign exchange (FX) volatility and customer-led cost-sharing or price-down mechanisms, with the impact most visible in more commoditized segments, especially electronics manufacturing services (EMS), where utilization remains sub-optimal.
While global semiconductor upcycle remains intact, Kenanga highlighted that the gains are becoming more concentrated.
“The global semiconductor cycle remains firm, with industry sales still accelerating and the market on track to approach $1 trillion by 2026,” it said.
According to the research house, January 2026 sales rose 3.7 percent month on month and 46.1 percent year on year to $83.7 billion, while World Semiconductor Trade Statistics (WSTS) expects the market to grow by more than 25 percent in 2026 to about $975 billion, led by continued strength in logic and memory.
“Unlike previous cycles, this upturn has been supported by more structural drivers, including AI, high-performance computing (HPC), 5G and next-generation device upgrades, which should help extend the cycle, although the benefits are becoming increasingly uneven across the sector,” said Kenanga.
The research house also noted that AI-driven memory tightness is reshaping the industry, but at the expense of non-AI demand.
“AI workloads are significantly more memory-intensive, prompting suppliers to prioritize HBM and high-capacity DDR5 for AI servers over conventional DRAM and NAND for smartphones, personal computers (PCs), vehicles and other consumer devices,
“This has created a zero-sum environment, where hyperscalers and AI server customers are absorbing a disproportionate share of industry output, while non-AI end-markets face tighter supply, higher costs and weaker affordability,” it said.
It opined that the AI boom is also making phones, PCs and electronics more expensive.
It is noted that IDC has sharply cut its 2026 smartphone shipment forecast, while Gartner now expects global PC shipments to experience a double-digit decline, both citing surging memory costs.
“In short, the AI build-out is no longer only lifting demand for AI infrastructure, but is also pushing up input costs across the broader electronics chain, which could squeeze margins, weaken affordability and place structural pressure on lower-end devices,” said Kenanga.
According to the research house, memory remains the key driver of the next WFE upcycle.
“Against this backdrop, memory is emerging as the clear leader of the next wafer fabrication equipment (WFE) upcycle,
“We expect memory WFE to have grown 13percent in 2025 followed by 23 percent in 2026, before moderating to 9 percent in 2027, making it the strongest growth segment across semiconductors,” it said.
This reflects the highly equipment-intensive migration in 3D NAND and DRAM; we estimate memory will account for roughly 65 percent to 70 percent of total WFE capital expenditure (capex) over 2024 to 2027, it added.
By comparison, it opined that logic spending should recover more gradually, supported by continued migration towards 3nm-class and sub-3nm nodes.
According to Kenanga, AI applications are evolving rapidly, but monetization is now the key differentiator.
It noted AI has progressed from basic automation to assistants, copilots, agents, and now agentic AI, where systems can increasingly reason, use tools and execute tasks with limited human intervention.
“In practical terms, AI is moving from a reactive helper to a more autonomous digital worker, extending the spending cycle beyond model training into real-world deployment,
“From an investment perspective, however, the focus is shifting from AI capability alone to sustainable monetization,” it added.