Bloomberg reported that the Biden administration is preparing new export restrictions on Nvidia's AI chips, to be issued as soon as Friday, targeting adversarial nations. Through a tiered system, key allies (Tier 1) maintain unrestricted access to artificial intelligence chips, Tier 3 countries (such as China) faces near-total bans, while Malaysia and vast majority of countries are in the middle category i.e. Tier 2. The new rules additionally limit the export of closed AI model weights (computation power), although this may be overcome if entities obtain validated end-user (VEU) status which comply with US security standards. Thus once this condition is satisfied, we don't see impediments to the AI business operations of YTLPOWR (OP; TP: RM5.00). Likewise, the AI server assembly business operations of NATGATE (OP; TP: RM3.10) is also not impeded. While there are country level quotas on chips for Tier 2 countries that suggest a "ceiling" where the chips ultimately end up, there shouldn't be restriction to sales if their end-buyers comply with VEU status. We further draw comfort that NATGATE's customers would have already been approved by NVIDIA. We will continue to monitor for developments.

Meanwhile, proposed Trump tariffs could reduce US consumer tech purchasing power by up to USD143b annually, according to the Consumer Technology Association (CTA). The semiconductor sector has continued to thrive, with global sales growing 20.7% YoY as of November 24. All in, INARI (OP; TP: RM3.85), NATGATE, KGB (OP; TP: RM4.16), and PIE (OP; TP: RM6.85) remain our top picks for the sector, driven by strong earnings visibility and strategic positioning to benefit from the ongoing AI demand uptrend.

Biden to further limit Nvidia AI chip exports in final push. The Biden administration is preparing to impose a new round of export restrictions on artificial intelligence (AI) chips, specifically targeting Nvidia's GPUs, in a final effort to limit the spread of advanced technology to adversarial nations like China and Russia, according to Bloomberg. These restrictions (see Exhibit 1) would introduce a tiered system: Tier 1 countries (including key U.S. allies like Germany, Japan, and Taiwan) would have unrestricted access to American chips; Tier 2 countries would face limits on the total computing chip they can import, though companies in these countries could bypass restrictions by meeting U.S. security and human rights standards (via a Validated End User or VEU designation); Tier 3 countries, including China and those under U.S. arms embargo, would face near-total bans on chip imports. The new regulations also extend to the export of closed AI model weights (which refer to the calculations to process data to make predictions), further restricting their use in non-allied countries. While the measures are designed to limit geopolitical rivals' access to superior AI technologies, industry groups like Nvidia and the Semiconductor Industry Association argue that such a significant policy shift should not be rushed during the presidential transition without more consultation with the tech sector.

There are 2 levels of restrictions being company and country level. For Tier 2 countries, we believe that the reported move by US to place a Tier 2 country level restriction of 50,000 chips in 2 years is in indeed a low ceiling for a country. However, it also means that the big picture idea here would be to induce companies which are the owners of the chips to apply for Validated end-user (VEU) status given that companies can surpass the country quantity restriction level so long as they obtain the VEU status. We therefore believe that this also means that for companies that own the chips to provide a service such as GPU as a service, or hyperscalers, would be incentivized to qualify as a VEU which also ultiimately means being subject to US security standards. This should not preclude assemblers of AI-chip powered servers from being able to conduct business, but it does imply that there are higher guardrails being put up for the buyers/end-owners of these chips themselves.

The new rules also restrict the export of closed AI model weights, which are key parameters used by AI software to make predictions. Tier 3 countries like China and Russia are banned from hosting powerful closed AI models, while Tier 2 countries can access them only if they meet strict security standards or obtain special approval (VEU status) from the U.S. government. Open model weights, which are publicly accessible, are unaffected by the rules. However, if companies in Tier 2 countries want to fine-tune open models with significant computing power, they must apply for a U.S. government license.

These measures aim to control access to advanced AI technologies while allowing limited use in allied countries In this context, Natgate, the only OEM partner of Nvidia in Malaysia (see Exhibit 2) and a small OEM partner in the ASEAN region, believes it will be minimally impacted by the GPU export quotas as a significant portion of its business comes from Singapore and deliveries are made as per client requests. Additionally, NATGATE sells AI servers to Nvidia-approved cloud partners, which are likely already cleared by U.S. authorities in our view. As a result, we make no changes to our earnings forecast for NATGATE. For illustrative purposes, every reduction of 500 GPU servers from our current 4.4k assumptions for FY25 would lower our net profit by 6.6%. Seperately, YTLPOWR is likley to experience minimal impact as well, in view of its priority access to Nvidia's chips due to its cloud partner status. The optimistic scenario here is if providers such as YTLP from Tier 2 countries may be able to take some market share from Tier 3 countries since there will be now be restrictions for Tier 3 countries to export the computing power.

Proposed Trump tariffs could slash US consumer tech purchasing power. The Consumer Technology Association (CTA) has warned that proposed import tariffs under Donald Trump's administration could result in substantial price hikes for consumer electronics in the US. The CTA outlined two potential scenarios: in the first, a 10% global tariff with an additional 60% tariff on Chinese imports could lead to price increases of up to 45% for laptops and tablets, 40% for video game consoles, 31% for monitors, and 26% for smartphones. In the second scenario, a more severe global 20% tariff with a 100% tariff on Chinese goods could push prices even higher, with laptops and tablets rising by 68%, consoles by 58%, monitors by 48%, and smartphones by 40%. The CTA emphasized that these tariff hikes would significantly reduce American consumers' purchasing power, with an estimated USD90b to USD143b in annual losses. The CTA argues these tariffs would harm consumers and businesses, reduce spending power by up to USD143b annually, and likely lead to manufacturing shifting to lower-cost countries, rather than bringing production back to the US. We believe the rise in tariffs could prompt price- sensitive consumers to accelerate their spending on consumer products before price hikes take effect.

Sustained semiconductor uptrend cycle. Global semiconductor sales recorded a 20.7% YoY growth as of November 2024, continuing an uptrend that began in November 2023, driven by strong demand in AI, HPC, 5G, and AI smartphone replacements. This uptrend, which has lasted 13 months, indicates that approximately 50% of the cycle has been completed, based on historical cycles where uptrends typically last 22 months and downtrends 12 months (with the longest being 30 and 16 months, respectively). The cycle could extend into late CY25 or mid-CY26, supported by projected global semiconductor sales growth of 19.0% YoY to USD627b in CY24 and 11.2% YoY to USD697b in CY25, according to World Semiconductor Trade Statistics (WSTS). The growth is primarily driven by demand for advanced chips, GPUs, AI accelerators, and memory, particularly for DRAM and NAND flash used in AI infrastructure, data storage, and 5G system.

INARI, NATGATE, KGB and PIE remain our top picks for the sector, given their earnings visibility and strong positioning to capitalise on the data centre fit-out phase and growing demand for AI-related infrastructure. Among these, INARI and NATGATE are preferred for their higher capital upside and direct exposure to the AI demand cycle and the anticipated smartphone replacement cycle, making them standout performers during the ongoing semiconductor uptrend.

Source: Kenanga Research - 10 Jan 2025