About two years ago, I called Nvidia overvalued. Investors were clamoring over the stock, and it went on to become the most valuable company in the world. All along, my instinct has been to get wary when everyone is singing a stock’s praises and buying it hand over fist. That caution has saved me countless times from getting swept into the speculation. Staying out of the herd has been useful. A recent example: in December 2025, I sold platinum into the gold and precious-metals craze ther than buying it. I still remember the queues snaking outside the shops. That was the contrarian in me talking. And I think a lot of investors are running that same contrarian instinct on the memory stocks today. But here’s the thing. Two years on, Nvidia didn’t crash. It broke $5T and stayed the most valuab
Nvidia is the obvious example. It used to be a gaming company with a brief moment of fame during the crypto-mining boom. Cyclical at best. In 2022, revenue jumped 61% year on year; in 2023, it decelerated to just 0.2% growth. Then AI demand sent its revenue and profits, and the share price, to the sky. Revenue doubled in FY24 and again in FY25. Even in FY26 it grew 65% year on year, which is phenomenal at that size. That is what a business inflection point looks like. And this kind of sudden, fundamental change is exactly what most investors struggle to wrap their heads around. We’re talking about a real, significant leap in the business itself, not just the share price. Most investors are trained to analyze steadily growing companies and extrapolate historical trends. That’s why they keep
Tesla needs AI for autonomous driving, robotaxi dispatch, Optimus robots, manufacturing automation, and service diagnostics. SpaceX needs AI for rocket design simulation, Starlink routing optimisation, factory automation, and anomaly detection. Tesla’s Megapacks can support SpaceX launch sites, Starlink ground stations, xAI data centres, and high-power factory operations. This is already happening. SpaceX and xAI reportedly bought hundreds of millions of dollars of Tesla Megapacks in 2024–2026. This is one of the clearest synergies as SpaceX needs resilient power, while Tesla Energy needs large industrial customers. The common layer is compute, data centres, AI talent, chips, models, and internal tools. Reporting has also flagged collaborations involving Tesla’s voice assistant, chip facto
There are plenty of less obvious synergies if SpaceX and Tesla were one company. Data is one. Between them, they have access to many kinds of data Tesla vehicles generate mobility, road, and environment data like starlink provides connectivity and network-performance data and space X has aerospace telemetry and operation data Their factories generate robotics and manufacturing data , this can improve AI systems for autonomy, logistics, network optimisation, and robotics Together, they can secure better supplier terms, priority access, and deeper supplier relationships. Talent. They can attract an elite engineering pool, and culture and talent can move across the system, between rockets, AI, robots, and energy. And imagine a disaster area where roads, power, and communications are all
So how do you identify sustainable growth? Buffett looks for businesses with durable competitive advantages—companies that competitors can't easily chip away at. If he doesn’t have a high degree of confidence that a company will be significantly larger in the future, he won’t buy—even if it looks cheap. Aside from the popular MOAT ETF, another ETF that reflects Buffett’s philosophy is the Dividend Aristocrats. These are companies that have raised dividends for at least 25 consecutive years. To do that, the business must be growing steadily. But my issue with Dividend Aristocrats is that most of the companies grow slowly—usually in the single digits. Consistency is good, but consistent growth in the 10–15% range over decades is even better. That’s the sweet spot. Also, not all great g
Meta went public at US$38 per share, raising over US$16 billion. However, the social media giant’s honeymoon period didn’t last long, and the stock struggled almost immediately.By early September 2012, less than three months after listing, the stock price had declined to less than half its IPO price, at a low of US$17.55. If you bought US$10,000 worth of Meta stocks at the IPO price of US$38 per share, you would’ve secured around 263 shares. With the stock now trading at US$614.23 per share, that initial US$10,000 would be worth approximately US$161,500 today, just based on capital appreciation. Moreover, the tech giant declared its first dividend on 1 February 2024.With a current dividend yield of 0.34%, your cumulative dividend would amount to roughly US$1,216, bringing the total investm
TOTO is one of the top producers of electrostatic chucks used in NAND memory production. You guessed it. The boom in memory prices and shortages, driven by AI data centres stockpiling storage, has brought loads of business to TOTO. They don’t sell these to Samsung or SK Hynix directly. Instead, TOTO is reportedly a dominant supplier to Lam Research, whose cryogenic etching machines are then used by every major NAND maker including Samsung, SK Hynix, Micron, Kioxia, and YMTC, to carve channel holes through 200+ layer 3D NAND chips It comes down to ceramics. TOTO has been making advanced ceramic parts for over 40 years — and it turns out, those same ceramic capabilities translate into electrostatic chucks, the precision plates that hold silicon wafers in place during chip manufacturing. In o
Tencent’s first-quarter 2026 results show a company in transition. Its core business is still highly profitable and generates strong cash flow. At the same time, it is making an aggressive push into artificial intelligence. This pivot has yet to be proven. Tencent continues to position AI as its next major growth driver. But execution is still a work in progress. The launch of the Hy3 preview large language model is a sign of progress in its in-house capabilities. A revamped AI infrastructure supports this. Even so, Tencent trails its domestic peers. Alibaba Group and ByteDance are ahead. Global leaders like Microsoft and Alphabet are further ahead still, both in frontier model performance and ecosystem scale. Tencent’s hybrid approach uses both proprietary models and external partners lik
Modern GPUs can crunch numbers much faster than standard memory can feed them. If the data transfer rate is too slow, an incredibly expensive GPU sits idle, waiting for information. High Bandwidth Memory (HBM) solves this by stacking DRAM chips vertically and connecting them directly to the processor via a microscopic highway, delivering terabytes of data per second. Without HBM, scaling frontier AI models is physically and economically impossible. Memory is the bottleneck in AI right now. As AI models have grown exponentially—moving from training to continuous, high-context inference and multi-step agentic reasoning—the constraint has shifted from processing data (compute) to moving data (memory bandwidth). And supply is short. Producing HBM requires roughly three times the wafer capacity
SanDisk (SNDK) tops the S&P 500 for the highest return in 2026 YTD. The stock was already the best performer in 2025, and it has continued this year. We’re only four months in, and it’s already more than tripled. SanDisk was a spinoff from Western Digital just last year. It was loss-making initially, but Q3 FY2026 saw net profits soar to $3.6 billion. Revenue was up 251% year-on-year. From zero to hero, quite literally. Western Digital reported 45% revenue growth year-on-year. But what’s more impressive is that its gross margin expanded 10 percentage points—from 40.1% in Q3 FY2025 to 50.5% in Q3 FY2026. That’s a sign WDC is raising prices as it sells more storage. Supply shortages are driving up prices, and these companies are the direct beneficiaries. Finally with AI, cloud growth exp