Based on the current trajectory, the clear leaders in AI and, therefore, most likely eventual winners will be US and Chinese companies. The rest of the world risks being left behind. We explored in depth the US-China AI arms race in a two-part article in July 2025. In a nutshell, the US may be at the forefront of innovation, but China is ahead in its ability to deploy, at scale.
US companies will be AI winners … but valuations and ROI are a concern
The US’ innovation lead is no surprise. The nation has spent decades building a strong foundational ecosystem — deep academia/research institutions/industry linkages, dynamic public/private capital markets, and a meritocratic system that has excelled in driving innovation and attracting top talent from around the globe.
It is undeniably ahead in terms of raw compute power and has near-monopolistic control over critical upstream components of the supply chain, notably in chip design software and advanced lithography (as well as specialised components/technology such as laser-produced plasma EUV light sources) needed to produce the cutting-edge chips used in AI and data centres.
See also: The reasons for ‘unrational’ negative enterprise value stocks in exchanges
President Donald Trump’s MAGA agenda — shifting the country right through pro-growth policies, deregulation, private sector-led investment and limited government intervention — is expected to spur greater domestic investment from both US and foreign companies, boost productivity and strengthen the US’ competitive position globally. More jobs, higher wages and incomes. There will be some short-term pain, for sure.
Tariffs will raise consumer prices and inflation, weaken domestic consumption and economic growth in the near term. A recent Goldman Sachs report estimated that US consumers will shoulder 55% of tariff costs by end-2025, with US companies (22%) and foreign exporters (18%) sharing the remaining burden (5% of the cost would be evaded). We remain bullish on the long-term prospects of the US economy and its companies — a point we have consistently emphasised. The willingness to accept short-term pain for long-term gain may ultimately stand as one of Trump’s most positive legacies — a rare trait among leaders.
The only caution for us right now is valuations — the result of the AI frenzy and FOMO (fear of missing out). The growing risks are underscored by the recent spate of mega AI deals, involving just a handful of companies — Nvidia Corp, OpenAI, Oracle, Advanced Micro Devices (AMD), CoreWeave and Broadcom — investing tens of billions of dollars in each other (as investors, manufacturers and customers) and creating hundreds of billions in “free money” in terms of market valuations (scan the QR code to read our recent article, “When the stock market gives you more than free money, and revisiting Aokam Perdana” [The Edge, Oct 6, 2025]).
See also: Leading enterprises in China’s rapidly evolving AI ecosystem
This circular flow of money is fuelling concerns of an AI bubble. Some observers compare today’s capital recycling loop to the 1990s dotcom era, when interconnected investors, vendors, internet service providers and start-ups drove each other’s valuations higher. We all know how that ended — with many start-ups lacking viable business models and ultimately failing to deliver the revenues needed to sustain sky-high expectations, returns on investment (ROIs), and valuations.
OpenAI has so far announced plans for 26gw of capacity, targeted for full deployment by 2029 — 10gw with Nvidia, 10gw with Broadcom and 6gw with AMD. This represents roughly US$1.3 trillion ($1.7 trillion) in capital expenditure (capex), assuming a cost of US$50 billion per GW. (OpenAI CEO Sam Altman has reportedly told employees he aims to build 250gw of new computing power by 2033.)
Separately, Meta Platforms, Google, Amazon.com and Microsoft Corp have raised their planned 2025/26 capex in their latest 3QFY2025 earnings results, now expected to total a combined US$390 billion this year alone.
Meta indicated that capex will be “notably higher in 2026 than 2025” while Microsoft said capex growth in FY2026 will exceed that in FY2025, reversing its previous estimate for slower growth.
Despite rising demand, there is still a lack of strong use cases — in other words, not just innovation but innovation with tangible profits improvements — to justify widespread enterprise onboarding. While we believe the eventual returns on AI investments will materialise and are significant, they may not come soon enough for some. In fact, we think ROI will decline over the coming years, owing to the massive capex being invested in data centre infrastructure and slower-than-expected revenue growth. This is likely to result in some shakeout or consolidation among AI companies. For others, it could turn into a survival of the strongest balance sheets and cash-flow resilience.
Given that AI is still in the very early stages of evolution, it is not yet clear to us where the real value lies and which companies will be the eventual winners and losers. Case in point: Although we were previously positive on the software solution providers, many could be made obsolete as AI companies such as OpenAI expand aggressively both upstream (Stargate) and downstream (everything?), offering end-to-end solutions powered by proprietary AI models.
For instance, OpenAI has integrated affiliated shopping links such as Etsy and Shopify that include product recommendations and direct checkouts. AI and AI agents can replace many of their current functions (such as website, marketing and ad designs and inventory management). Could it, at some point, replace these platforms altogether (as the link between merchants and customers)? After all, OpenAI tied up with Walmart to offer users AI-assisted checkouts. Meanwhile, it has launched a web browser, Atlas, powered by ChatGPT — where paying subscribers also get “agents” that can perform/complete tasks on their behalf.
For more stories about where money flows, click here for Capital Section
We also expect AI companies will compete for a piece of the lucrative digital advertising pie, currently dominated by Google and Meta. OpenAI’s text-to-video platform Sora could turn out to be a formidable competitor for social media platforms (for consumer eyeballs, engagement and data), as well as traditional software ecosystems used by content creators. In this environment, standalone solution providers could be especially vulnerable while agnostic platform providers may be better positioned. Up next — AI-powered personal assistant and consumer devices (with former Apple star designer Jony Ive in its employ) to rival Apple?
Here’s our worry — say, AI companies were to succeed in generating the huge revenue needed to justify their massive capex and lofty valuations, at whose expense would that be? Right now, the market is pricing huge gains for AI companies but little to no loss for incumbents. All are trading at high valuations. It may not exactly be a zero-sum game, but all cannot win at the same time — yes, the pie will grow and, yes, new revenue streams could be found, but surely the total addressable market cannot double overnight. We intend to explore this further in a future article.
An even graver scenario would arise if AI companies falter and fail to generate the revenue needed before capital runs out. As we noted previously, if OpenAI (which is at the heart of the recent spate of mega deals) cannot meet its obligations, Oracle could be left with massive stranded assets and debt, while chip-system demand for and margins of Nvidia, Broadcom and AMD would compress rapidly. This demand contraction would ripple through the global supply chain and across equity markets, given the size and influence of these mega-cap stocks, the high concentration in key bellwether indices and the scale of passive funds. The broader US economy would also feel the impact, given how critical AI spending has been in supporting growth.
Thus, although we remain bullish on the US equity market over the long term, we are exercising caution. Accordingly, we are reallocating part of our investments into Chinese stocks, where current risk-reward propositions appear more attractive.
Upbeat on China’s economy and clearer monetisation path for AI
The fact that the US has thrown up every roadblock in its arsenal, including the proverbial kitchen sink, to slow China’s AI progress is tetstament to how fast the latter has caught up and how close the competition has become. Indeed, we think, by some metric and certainly for some sectors, China is ahead of the US. As the sayings go, “necessity is the mother of invention” and “what doesn’t kill you, makes you stronger”.
Chinese stocks, previously seen by many in the West as “uninvestable”, are trading at lower valuations compared to their US peers. Chinese companies are also much more focused on commercialisation and monetising AI investments. Yes, China intends to boost R&D aimed at achieving its own technological breakthroughs in the future. But China is also pragmatic where the Western world is often more inclined to chase after huge moonshot bets and frontier science. Case in point: OpenAI’s official mission statement is “to ensure artificial general intelligence (AGI) benefits all of humanity”. AGI comprises highly autonomous machines capable of superhuman general reasoning. China, on the other hand, is prioritising real-world industrial and commercial applications over higher AI capabilities such as clear macro policy directives and whole-of-nation approach, state funding support, heavy investment in education, infrastructure and domestic supply chains.
In August 2025, China’s State Council released the “AI Plus” policy directive, setting ambitious goals for “broad and deep AI integration” across the economy and society — to accelerate innovation. It aims to integrate next-generation intelligent terminals and AI agents in six key sectors (science and technology, industry, consumption, public services, governance and global cooperation), with the penetration rate expected to exceed 70% by 2027 and 90% by 2030. AI is already being integrated into both traditional and emerging sectors, from agriculture and manufacturing to robotics.
Chinese companies are adept at extracting more for less. DeepSeek was built and trained at a fraction of the cost of ChatGPT. Alibaba Group Holding plans to spend “only” US$65 billion on AI over the next three years and is already breaking even on its AI investments in e-commerce. More recently, the company disclosed an internal test in which its Aegaeon system successfully cut the number of Nvidia H20 GPUs (graphics processing units) used by 82% by dynamically allocating compute to serve multiple models. What this means: lower capex and operating costs for inference deployments.
China’s adoption of open-source AI models, combined with significantly lower cost barriers, is driving rapid and widespread AI adoption among small and medium enterprises. A large domestic market and scaleable applications, in turn, enable fast iteration and the generation of valuable real-world operational data, further improving AI models.
Case in point: China’s highly developed ecosystem in hardware and battery manufacturing, light detection and ranging sensors as well as AI models that have driven its electric vehicle sector are synergistic to the training and development for humanoid robots — at very competitive prices, compared to, say, the indicated price for Tesla’s Optimus robot (still in development). UBTech Robotics Corp’s humanoids, for instance, are already deployed in several auto factories, working in production tasks (assembly, quality inspection and materials handling). And Unitree Robotics has just unveiled its most-humanlike humanoid robot capable of performing fluid dancing and martial arts movements.
All of this will make already high-cost-efficient Chinese industries even more competitive in the global market. We believe that Chinese companies will see ROI on AI much sooner than in the US, on average. As Alibaba chairman Joe Tsai said: “The winner in AI should not be defined by who comes up with the strongest AI model, but by who can adopt it faster.”
This is one major reason we are bullish on the Chinese economy (for a more detailed analysis, scan the QR code to read our article titled “China, from one great economic transformation to another” [The Edge, Oct 20, 2025]). As we have said, not investing in China is not an option for serious investors.
As noted at the start of this article, we view both US and Chinese companies as winners in AI. In the shorter term, we favour Chinese stocks and are therefore overweight China-based companies in the global Absolute Returns Portfolio. We purchased Ping An Insurance as a proxy for the broader Chinese economy, while Trip.com and Kanzhun stand to benefit from rising domestic consumption in services. For tech sector exposure, we hold Alibaba, Tencent Holdings and the ChinaAMC Hang Seng Biotech ETF.
For the AI Portfolio, which is our decade-long investment bet on the future of AI, we have maintained a heavier weightage in US tech stocks — our current holdings include Amazon, Cadence Design Systems, Datadog, ServiceNow, Twilio and Marvell Technology. For China exposure, we are invested in Horizon Robotics and RoboSense Technology — which specialise in AI chips, hardware and software for autonomous vehicles that are also applicable to humanoid robots — in addition to Alibaba.
Avid readers might notice the absence of Intuit from our US holdings listed above. This is because we have exited the position in favour of Naura Technology Group Co, which we wrote about last week. Since then, the stock has traded lower amid short-term profit-taking across the Chinese semiconductor sector, making its current valuation more appealing. Naura is China’s largest domestic semiconductor equipment vendor. Over the past year, it has expanded its portfolio by acquiring a stake in Kingsemi Co. We are positive on Naura, given China’s ongoing push for domestic substitution across the semiconductor value chain. As at 2024, China’s domestic replacement rate in semiconductor equipment stood at roughly 25%, and Naura operates in segments expected to see the fastest localisation gains. This includes the cleaning equipment segment, where domestic replacement is projected to increase from 35% to 65% by 2026, according to MIR Databank projections.
Malaysia’s comparative advantage in plantations and high yields for domestic-centric stocks
Bursa Malaysia does not currently offer any exciting AI-related stock — that is, stocks with strong growth prospects that are commensurate with their relatively high valuations, which is due in part to the scarcity factor on the local bourse. The average trailing price-earnings multiple for all listed companies in the tech sector with a market cap above RM1 billion is 50 times. The five largest by market cap are Inari Amertron (44 times), ViTrox Corp (80 times), Frontken Corp (53 times), Zetrix AI (formerly MYEG; 9 times) and Malaysian Pacific Industries (39 times). Compare that with Nvidia (53 times), Microsoft (38 times), Apple (40 times), Alphabet (28 times) and Amazon (34 times).
How can Bursa Malaysia attract global investors when there is a smorgasbord of choices available in the world? What are the bourse’s competitive advantages? Most of its largest tech companies form part of the semiconductor supply chain — primarily OSAT (outsourced semiconductor, assembly and test) and ATE (automated test equipment) manufacturers — and EMS (electronic manufacturing services) providers.
Many of these companies operate at the lower end of the value chain and possess limited intellectual property — and thus limited pricing power — in the global market. This makes them particularly vulnerable to US tariffs, as they are forced to absorb part of the tariff costs or risk losing market share to competitors, further eroding already thin margins.
In fact, low value-add and limited pricing power are symptomatic across many of Malaysia’s manufacturing sectors, including gloves, furniture, plastic packaging and automobiles. These sectors have historically benefited from government policies such as tariff and non-tariff protection, cost subsidies (for fuel, electricity and water) and wage (foreign labour) suppression — measures that are ultimately unsustainable if Malaysia aims to become a high-income nation. We believe wages must be market-determined, reflecting worker productivity and skill sets, as competition for talent is global. Achieving this requires improving the quality of public education.
One of the clearest sectors in which Malaysia does have comparative advantage is plantations, hence our holdings in United Plantations and Kim Loong Resources. Medical tourism could be another, though listed healthcare stocks such as IHH Healthcare (33 times) and KPJ Healthcare (36 times) are also trading at high valuations relative to expected growth.
We recently invested in Southern Cable Group and Hiap Teck Venture, anticipating a resurgence in re-infrastructurisation, but have since cut our losses. Current policies appear focused on short-term results, lacking strategic or holistic vision.
Amid heightened uncertainties — including tariffs, global economic growth and geopolitics — and the absence of a compelling risk-reward proposition, we have decided to adopt a strategically defensive stance for the Malaysian Portfolio, allocating the bulk of our investments to high-yielding stocks such as Malayan Banking, LPI Capital and Hong Leong Industries. At least for now.
The FBM KLCI has performed very poorly in the past 10 years (see Table 1), registering negative returns. But the fact is that it has always underperformed relative to the MSCI Emerging Market Index, over the past 10, 20 and 30 years. The numbers look even grimmer when one factors in the secular decline in the value of the ringgit against the US dollar (that is, even lower returns in US dollar terms). It has also consistently underperformed the Indonesia and Singapore stock markets over all these time periods. Why should any investor with a global perspective buy Malaysian stocks? Indeed, why would anyone expect that Bursa’s performance will be any different going forward? We will explore this subject further in the near future — why the chronic underperformance and what needs to be done if Malaysia wants to perform and do well again.
The Malaysian Portfolio gained 1.6% for the week ended Nov 5 led by United Plantations (+4.9%) and Hong Leong Industries (+3.6%). As mentioned above, we disposed of our investments in Southern Cable Group and Hiap Teck, raising cash to about 47% of total portfolio value. This portfolio continues to outperform the benchmark FBM KLCI, with total returns of 187.5% since inception. By comparison, the FBM KLCI is down 11.4% over the same period.
The Absolute Returns Portfolio fell 1.7% for the week, however, paring total portfolio returns to 40.7% since inception. The top three gainers were Berkshire Hathaway (+2.7%), JP Morgan (+2.0%) and Goldman Sachs (+1.3%). Alibaba (-7.4%), Trip.com (-4.8%) and Kanzhun (-3.3%) were the biggest losing stocks last week.
The AI Portfolio also ended in negative territory, falling 0.8%. Total portfolio returns now stand at 6% since inception. The top gaining stocks were Twilio (+17.7%), Amazon (+8.6%) and Marvell Technology (+3.1%) while the big losers were RoboSense (-8.9%), Horizon Robotics (-8.6%) and Alibaba (-7.4%).
Unproductive wage policy
Malaysia’s progressive wage policy (PWP) was first proposed in late 2023 and underwent a pilot phase in mid-2024 before it was formally implemented nationally in January 2025. The voluntary programme is designed to be a monetary incentive for employers — specifically targeting small and medium enterprises (SMEs) — to raise wages of workers earning less than RM5,000 ($1,558) a month.
Without going into the finer details (and the computations do get convoluted), we can say that the government covers the additional costs — up to RM200 per new employee per month and up to RM300 per existing employee per month — if employers raise full-time salaries by at least 6%. Employees must also complete 21 hours of approved training — an attempt to tie higher pay to upskilling and productivity gains.
The government has budgeted RM200 million for the programme in 2025, targeting 50,000 employees. For perspective, the total number of people employed in Malaysia’s private sector is more than 9.4 million.
Sounds like a win-win situation for both employers (the government covers costs for worker increments) and employees (higher salaries). As at October 2025, however, only 20,737 workers had received salary increases in line with the PWP guidelines. Why the low take-up rate of only 41%?
We hazard a few guesses here. First, as we mentioned, the eligibility criteria, requirements and guidelines are convoluted, and application and approval are handled on a first-come, first-served basis. Perhaps companies are worried they will be rejected after they have hired a new worker at higher wages and/or raised salaries for existing workers. It would surely be difficult to reverse these actions. The programme runs from January 2025 to December 2027. In other words, the funding (subsidy) runs out after three years, and the companies will then have to bear the higher wage costs as well as mandatory training costs, which would dilute the incentive.
We suspect the companies that did register for the PWP programme are those that have already decided to hire at those specified salaries (spelt out in the “Guidebook of Starting Basic Salaries and Annual Salary Increments for All Sectors”, published by the Ministry of Human Resources) and pay the minimum 6% annual increment to their existing employees anyway. In other words, the PWP was not the motivating or deciding factor. Therefore, it contributed little to the broader aim of lifting salaries in the nation. To call it a “policy” is grandiloquent (or, in simple English, bombastic). It is yet another case of what we call “Singapore whitewashing”, or SWW, that is, borrowing a Singaporean concept out of context, mimicking its form without its original intent or substance.
The core idea of Singapore’s “Progressive Wage Model” (PWM) is to raise wages specifically for lower-income workers — starting with the cleaning sector in 2014 and progressively expanded to other traditionally low-wage sectors such as security, retail, landscaping, lift and escalator maintenance and food services. Each sector has a wage ladder — higher pay as the worker progresses from basic to skilled to supervisory roles (meeting training requirements). Notably, PWM is mandatory (not voluntary) and is applicable only to citizens and permanent residents. Employers can hire cheap foreign labour only if they pay PWM-compliant wages for their local workers, thus ensuring high compliance (effective enforcement). In short, Singapore adopted a productivity-linked sustainable wage growth model for low-income Singaporeans instead of adopting a blanket minimum wage policy (which, in Malaysia, includes all foreign workers).
Yes, we need a strategy to overcome wage stagnation and move towards a high-income economy, but subsidising a few people for a short period is not the solution. It is wasteful spending that delivers little beyond publicity.
Disclaimer: This is a personal portfolio for information purposes only and does not constitute a recommendation or solicitation or expression of views to influence readers to buy/sell stocks, including the particular stocks mentioned herein. It does not take into account an individual investor’s particular financial situation, investment objectives, investment horizon, risk profile and/or risk preference. Our shareholders, directors and employees may have positions in or may be materially interested in any of the stocks. We may also have or have had dealings with or may provide or have provided content services to the companies mentioned in the reports.
