Does the present pullback present a compelling case of buying the dip, or will it leave investors clutching a falling knife? Sceptics note that much of the market’s momentum is built on faith in AI (with its promises of productivity gains) but AI’s history is littered with cycles of hype and heartbreak. On the other hand, buying the dip has undoubtedly been one of this year’s most effective strategies. One popular play, hinging specifically on the Trump administration’s penchant for using brinkmanship to extract concessions from trading partners, even has a nickname: the TACO (“Trump Always Chickens Out”) trade. In essence, it is a bet that tariff-induced sell-offs will swiftly unwind once the White House reverses from its own threats.
Whatever the case, stripping away the noise and distraction, every investment decision ultimately comes down to one deceptively simple concept: valuation.
The shape of growth
Put simply, the value of any business is the sum of its future cash flows, discounted at a rate that reflects both risk and the time value of money. The so-called “growth stock premium” (tech companies’ tendency to command richer price-earnings [P/E] multiples than other sectors) is exactly what it sounds like: a bet on growth. The market assumes their earnings will grow sharply over time, producing a larger stream of future cash flows. Such companies are, in a sense, like children in oversized clothes, stitched from the hopeful narrative that time will fill them out.
See also: Stablecoin — the US$2 trillion question
Palantir, for instance, trades at a forward P/E north of 200 times, implying that the market has priced in rapid and sustained growth that makes paying US$200 for every US$1 of today’s profits appear reasonable. The question, then, is whether such expectations are realistic.
In reality, growth rarely follows a straight line. More often, it traces an S-curve: a slow start, then rapid acceleration after an inflection point, and eventually a slow taper into maturity. For a company, long-term survival depends on leaping to new S-curves before the old one peaks. The key takeaway, then, is that understanding a company’s position on this trajectory is crucial for assessing whether its current valuation make sense (see Chart 1).
See also: Sustained economic success comes only from harnessing local private enterprises
Tesla
Tesla best illustrates the complexity of the S-curve across its full trajectory: a sharp climb powered by electric vehicles (EVs), now flattening as it pivots towards riskier, still-unproven bets. The company’s first S-curve was built on EV adoption during a period of accelerated growth fuelled by falling lithium-ion battery costs and the scaling of its Supercharger network. From 2018 to 2023, Tesla consistently topped charts as the world’s best-selling EV maker.
Today, however, Tesla’s growth story rests less on car sales and more on autonomous driving, robotics and new energy technologies. Revenue projections through 2028 reveal an interesting pattern (see Chart 2): Since 2022, Tesla’s actual performance has clearly indicated moderating growth, yet the market appears to have priced in a significant new S-curve from 2026. This kind of pivot is precisely what can extend an S-curve or create new ones. It also introduces risk, however, when the market prices in the second act before it has found solid footing.
For Tesla, the danger is clear: Investors increasingly seem to be paying for promises rather than performance. In 2024, rival carmaker BYD’s annual revenue surpassed Tesla’s for the first time. Production of Tesla’s humanoid robot, Optimus, has fallen behind schedule for 2025. The July launch of its robotaxi service has since spiralled into a contentious class-action lawsuit over alleged misrepresentation about the safety of the company’s proprietary Full Self-Driving technology. In August, the company shut down its Dojo supercomputer hardware team, which was once touted as the cornerstone of its autonomous driving ambitions.
And perhaps the clearest evidence of Tesla’s story-driven trading is the company’s elevated forward P/E of 184 times, despite the company’s falling sales. Revenue dropped 12% y-o-y in the latest quarter. The market is willing to pay a premium precisely because of future growth expectations, reflecting continued belief in the company’s growth narrative over current fundamentals.
Palantir
For more stories about where money flows, click here for Capital Section
Palantir, by contrast, is still firmly on the upward slope of its first S-curve (see Chart 3). In its latest quarter, revenue reached US$1 billion ($1.3 billion), up 48% y-o-y — a significant acceleration from 27% in the same period last year. Accordingly, management has lifted its full-year revenue guidance to as high as US$4.15 billion, implying 45% growth for the full year. On the S-curve, this places Palantir firmly in its expansion phase, where growth is both rapid and compounding.
Viewed in isolation, Palantir’s forward P/E of more than 200 times is certainly brow-raising. Framed by the S-curve, however, the justification for such a rich valuation can be resolved over time through sustained growth. Should Palantir live up to consensus earnings growth expectations for the next five years, its valuation will shrink to a still-rich, but much more manageable, forward P/E of 55 times, where Nvidia’s trailing P/E stands today. If Palantir then manages five more years of consensus growth, it will trade at a forward P/E under 20 times by 2034. Such a trajectory is ambitious but not unprecedented. Nvidia, for example, posted growth above 50% in four of the past five years alone, peaking at 126% in 2024. With margins improving, Palantir also benefits from operating leverage: Each dollar of revenue growth translates into a disproportionately larger increase to its bottom line.
But herein lies the rub. Palantir’s lofty forward P/E hinges on flawless execution of its growth runway. The stock is, in other words, priced to perfection. As the S-curve reminds us, this phase of accelerated growth cannot last indefinitely. Growth will inevitably taper, whether from market saturation, mounting competition or the law of diminishing returns. Analysts already flag potential headwinds, particularly in Palantir’s commercial segment, which contributed 45% of total sales last year. Data warehousing players such as Snowflake and Databricks continue to expand large language model capabilities on their platforms, allowing customers to perform advanced analytics without migrating to Palantir’s platform. Moreover, while US commercial sales surged 93% y-o-y in the most recent quarter, international revenue declined 3%, underscoring weak traction abroad. On the government front, competition is intensifying, with reports suggesting the Pentagon and other US agencies are piloting AI models from Microsoft Corp and OpenAI.
If these pressures persist, Palantir will need to seek new growth vectors to maintain its expected high growth rate. Its long-term trajectory depends not only on scaling its current markets but also on its ability to “jump” S-curves successfully, driving fresh waves of demand before the existing one flattens. The trajectories of the next two companies, Nvidia and Intel Corp, exemplify opposite ends of the spectrum when it comes to timing and executing the leap to a new curve.
Nvidia
Today, Nvidia has become synonymous with AI, but it has not always been this way. Since its inception, Nvidia has consistently managed to identify and capitalise on new waves of demand, evolving from gaming to high-performance computing to AI acceleration, and now providing a full-stack platform serving AI, robotics and other emergent tech fields.
Founded in 1993, Nvidia was initially a relative latecomer in a computing landscape dominated by Intel and Advanced Micro Devices (AMD). The company carved out its first S-curve by targeting a niche that CPUs (central processing units) struggled to serve: rendering complex graphics for gaming and multimedia. In 1999, Nvidia launched the world’s first GPU (graphics processing unit), the GeForce 256, sparking rapid growth on the back of multiple tailwinds: Microsoft’s DirectX framework, which enabled richer graphics on Windows; hit 3D games such as Quake and Half-Life; and the broader PC gaming boom. Its early success culminated in rapid growth in the early 2000s and Nvidia was even recognised as Forbes’ Company of the Year in 2007.
Crucially, even as gaming continued to represent at least 50% of total sales through 2020, CEO Jensen Huang had begun quietly building the company’s next growth vectors. As early as 2004, Nvidia began adapting its GPUs for non-graphical workloads, culminating in the 2006 release of CUDA (Compute Unified Device Architecture). CUDA enabled GPUs to perform parallel computations, initially finding adoption mainly in research and academia but later becoming central to AI workloads.
Through much of the 2010s, however, the commercial market largely overlooked these advances. Nvidia’s academic and scientific applications were deemed too niche, and AI remained a neglected discipline. It was only in 2017 that Nvidia experienced meaningful reacceleration in growth, driven by multiple AI breakthroughs that led numerous publications to crown the year as “The Year of AI”. Deep-learning techniques saw adoption across healthcare and finance for tasks such as image recognition and predictive analytics. Concurrently, the cryptocurrency mining boom acted as an unexpected catalyst, though the volatility of crypto markets created a double-edged sword.
As AI demand accelerated in the 2020s, Nvidia finally crystalised the growth vector it had long been positioning for. Between 2017 and 2025, its revenue from data centres surged from roughly 12% to 88% of total sales.
Looking ahead, Nvidia’s growth is expected to moderate to the teens by 2028 (see Chart 4), and the company’s S-curve projects similar flattening. Nonetheless, Nvidia is already exploring new growth engines: humanoid robots, silicon photonics networking switches and autonomous driving, among other initiatives. Should the company successfully engineer these new growth vectors as it has in the past, this would enable it to create fresh S-curves and launch new growth cycles.
Intel
Conversely, Intel’s case illustrates the perils of failing to jump S-curves. Despite pioneering the x86 architecture that underpinned the vast majority of PCs, the company has struggled to adapt as markets shifted, stubbornly trailing its original curve down an inevitable decline (see Chart 5).
In the 2000s, Intel stood at the zenith of its power. Alongside Microsoft, the “Wintel” duopoly defined the PC era, and when Apple transitioned its Macintosh line to Intel chips in 2005, the company’s dominance appeared unassailable. Soon after, however, Intel made its first major misstep. In 2006, former CEO Paul Otellini famously declined to develop processing chips for iPhones, believing smartphone sales would not justify development costs. The smartphone market subsequently grew almost exclusively on competing ARM Holdings-based architecture.
Ironically, ARM has since emerged as a formidable threat even in Intel’s “most profitable cash cow” — servers and data centres. The shift towards AI workloads, which prioritises GPU-driven acceleration, has further eroded Intel’s dominance. ARM-based CPUs, once derided as “underpowered”, have gained favour for their superior power efficiency and customisability, allowing for lower energy consumption and operational costs.
An even bigger strategic miss, in hindsight, was passing on a potential 30% stake in OpenAI in 2017. Then-CEO Bob Swan doubted generative AI would reach commercial scale. In 2024, the year OpenAI hit a US$157 billion valuation in its Series E, Intel’s second-quarter earnings collapse triggered its worst trading day since 1974. The company ended the year down 60%, one of the worst-performing stocks on the Nasdaq. In November that year, it also concluded its 25-year run in the Dow Jones, ceding its position to none other than Nvidia.
In recent days, however, the market appears to be tentatively pricing in hopes for a turnaround, underpinned by two recent developments. One, SoftBank Group Corp will make a US$2 billion capital injection in return for an equity stake of about 2%, telegraphing a vote of confidence from the Japanese conglomerate. Two, the Trump administration will take a nearly 10% equity interest in return for US$5.7 billion in grants from the CHIPS and Science Act and US$3.2 billion under the Secure Enclave programme, both of which were previously awarded to Intel during the Biden era but unpaid. Notably, the shares will be issued at a discount to prevailing prices and, therefore, dilutive. White House economic adviser Kevin Hassett indicated that the Intel stake will form part of a sovereign wealth fund that could include other companies.
Intel’s stock rebounded more than 20% in August this year. Its shares are trading at a forward P/E of 172 times despite projections of single-digit revenue growth, a valuation that is in part distorted by depressed earnings but perhaps also reflects investor faith in an ambitious turnaround. Much of that belief is pinned on two factors: the Trump administration’s push for a “home-grown” semiconductor champion; and optimism about new CEO Lip-Bu Tan’s leadership.
The bottom line
So, what does all this mean for investors weighing whether to buy the dip?
Valuation is, by definition, forward-looking. It rests on future cash flows — numbers that do not yet exist. The gap between forecast and reality is bridged by belief: the conviction that today’s assumptions about a company’s growth, margins or technology will ultimately materialise.
The S-curve offers a useful framing, because every growth story eventually traces its shape — a slow start, rapid acceleration, and an eventual taper. As Chart 1 illustrates, each company discussed in this article sits at a different point on the curve. They also trade on different levels of belief or hope.
For Tesla, it is the promise of a new curve (for us, we have never believed the Tesla story; Musk is a great salesman); for Palantir, it is the durability of its first curve (we bought and, yes, sold too early — we should have listened to our own advice of “ride a winner”); for Nvidia, it is the credibility of management’s track record in creating the next growth vectors (we like Nivida tremendously but always thought its stock prices ran ahead of fundamentals — we were wrong; we need to reframe for very high-growth stories); and, for Intel, it is the foundations of a turnaround that may finally be gaining traction after more than a decade of false starts (we will watch and wait for new catalysts, if any; with a 10% stake and as the single-largest shareholder, will Trump and the US government end up helping or hurting Intel and other shareholders?).
We may also have been slightly “too fundamental” in our investment strategy. When liquidity drives markets, good stories go a long way in driving short-term valuations. Hope always sells … until reality hits.
When it comes to valuing a company, numbers set the floor and stories set the ceiling. But it is the S-curve that bridges the gap between the two and ultimately determines the trade.
The Malaysian Portfolio gained 0.8% for the week ended Sept 3. Insas Bhd – Warrants C (+25.0%), Hong Leong Industries (+3.9%) and Malayan Banking (+1.9%) were the top gainers for the week; the two losing stocks were LPI Capital (-2.6%) and Kim Loong Resources (-0.4%). Total portfolio returns now stand at 185.1% since inception. This portfolio is outperforming the benchmark FBM KLCI, which is down 13.7% over the same period, by a long, long way.
The Absolute Returns Portfolio also advanced for the week, up 2.5% and boosting total portfolio returns to 36.2% since inception. The three biggest gainers were Alibaba Group Holding (+10.2%), Trip.com (+8.7%) and ChinaAMC HangSeng Biotech ETF (+8.5%). At the other end, shares in Goldman Sachs Group (-2.5%), CrowdStrike Holdings (-2.2%) and Tencent Holdings (-0.3%) ended lower on profit-taking.
The AI Portfolio outperformed, surging 3.9% last week. The gains lifted total portfolio returns to 3.2% since inception. The top gainers were Horizon Robotics (+21.1%), Alibaba (+10.2%) and ServiceNow (+3.5%) while SAP (-1.5%) and RoboSense Technology Co (-0.4%) ended in the red.
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.
