I think infrastructure offers the clearer opportunity today, but software may ultimately capture more value. AI infrastructure benefits first because every model and application needs compute, memory, networking, power and data centres. Demand is tangible and spending is already enormous. But infrastructure is capital-intensive and eventually risks overcapacity and commoditisation. Software has the opposite challenge: competition is intense today, yet successful applications can scale with much lower marginal costs and become deeply embedded in business workflows. My preference would be infrastructure while AI capex remains strong, then gradually shift attention towards software companies that demonstrate genuine pricing power, recurring revenue and measurable productivity gains. The next

C. Valuation and market expectations. AI chip demand and infrastructure spending can remain exceptionally strong, but stocks trade on the gap between reality and what is already priced in. A semiconductor company can deliver record revenue and profits yet still fall if growth, margins or guidance disappoint elevated expectations. That becomes especially important after a strong rally, when valuations assume years of continued AI expansion. I would still watch revenue growth, margins and hyperscaler capex closely, but valuation determines how much good news is already embedded in the share price. Great company ≠ great investment at every price.

I think the bigger opportunity for platforms like Robinhood is providing crypto services rather than holding large amounts of Bitcoin on their own balance sheets. Trading, custody, staking, tokenisation and stablecoin-related services can generate recurring revenue while letting customers bear most of the underlying crypto price risk. A large corporate Bitcoin position is different: it effectively turns part of the company into a leveraged bet on BTC and could increase earnings and balance-sheet volatility. Holding some Bitcoin may make strategic sense for liquidity, treasury diversification or supporting crypto operations, but I would not expect it to become the core strategy for most financial platforms. The real prize is probably becoming the infrastructure through which millions of cu

I want clarity on one thing: does Warsh see the surge in long-term yields as monetary tightening that reduces the need for another immediate hike, or as a separate fiscal/term-premium problem the Fed should largely look through? My base case is an October hold, which already appears well priced. The bigger market mover would be his December signal. Recent Fed commentary suggests policymakers can afford to wait, even though another hike this year remains on the table. For equities, especially long-duration tech, I would rather hear Warsh acknowledge that financial conditions have tightened and that policy remains data-dependent. If he stresses persistent inflation while dismissing the rise in long yields, markets could start pricing a more aggressive path. The key question is no longer simp

A company’s own story stops dominating when the market decides the bigger risk sits above the company level. Intel falling 5%+ despite favourable news suggests investors were trading the semiconductor sector rather than Intel itself. When concerns shift to higher yields, stretched valuations, AI capex or a broader chip-cycle slowdown, good company news can easily be overwhelmed. That does not make Intel’s progress irrelevant. If its fundamentals continue improving while the share price is dragged down mainly by sector sentiment, that divergence could eventually become an opportunity. The key question is whether Thursday was simply “sell the group first, differentiate later”, or whether investors see risks in Intel that the positive headlines have not addressed.

I think memory may be approaching a cyclical peak, but the evidence is not strong enough to call the top yet. Samsung’s record profit shows the underlying demand and pricing environment remain exceptionally strong. The warning sign is the second derivative: profit is still growing, but sequential growth has slowed sharply, while DRAM and NAND price increases are moderating. For cyclical stocks, markets usually price the turn before earnings actually fall. That explains why MU, SNDK and SK Hynix can sell off even when current profits look spectacular. My base case: memory fundamentals remain strong into 2027, especially HBM and AI-related demand, but valuations could become increasingly sensitive to any sign of slower pricing, capacity expansion or weaker AI capex. Record earnings do not ne

I think the bigger force at the long end is the term premium rather than just expectations for another Fed hike. Persistent fiscal deficits and heavy Treasury issuance mean investors need to absorb more duration, while inflation uncertainty makes them demand higher compensation for holding 10Y and 30Y bonds. That can push long yields higher even if the Fed eventually pauses. So I am watching Treasury supply, auction demand and the term premium closely. If those pressures persist, a Fed pause may bring limited relief to long yields, keeping valuation pressure on equities, especially long-duration growth stocks.

I think the long end can stay elevated, or even move higher, even if the Fed stands pat. The key issue is that the 10Y/30Y are increasingly being driven by more than the expected Fed path. Fiscal deficits, heavy Treasury issuance and a rising term premium can keep pushing long-term yields higher without another rate hike. A Fed pause may relieve some pressure at the front end, but it does not automatically solve the supply-demand imbalance further out the curve. If investors demand more compensation to hold duration, the curve could steepen through higher long-end yields. For equities, that matters because a 5%+ 10Y keeps the discount-rate hurdle high, particularly for expensive growth stocks. My base case: Fed pause ≠ long-end relief. I would watch Treasury auctions, term premium and infl

The higher FY2028 target is encouraging, but I do not think US$20B alone is enough to carry MRVL significantly higher. The key question is whether earnings can catch up with the AI narrative. I would watch AI-related revenue growth, margins and whether hyperscaler demand remains strong enough to justify the current expectations. A raised target improves confidence, but once the market prices in strong growth, execution matters more than guidance. If Marvell keeps beating estimates and raising forecasts, the rally can continue. If growth merely meets the new target, valuation could become the bigger constraint. For me: bullish on the business, but increasingly selective on the share price.

A. AI & Technology 🤖 I’ll be watching AI policy most closely, especially any changes around chip export controls, data-centre power infrastructure, AI regulation and government incentives for domestic semiconductor production. These could have significant implications for the entire AI supply chain, from Nvidia and AMD to hyperscalers, utilities and data-centre operators. Around the midterms, even shifts in expectations for future policy could move valuations before any legislation actually changes.

C for me: ⚡ Data centres, power & infrastructure. AI models may change leaders quickly, but every serious competitor still needs compute, memory, networking, cooling and electricity. That makes the infrastructure layer particularly interesting because it can benefit regardless of whether OpenAI, Meta, Anthropic or another player ultimately wins the model race. I’m also watching D closely. The scale of AI capex is becoming enormous, so eventually revenue and free cash flow must justify it. Spending hundreds of billions is bullish for infrastructure suppliers, but not necessarily for the companies writing the cheques. My preferred approach is therefore to follow the bottlenecks: GPUs → HBM → networking → cooling → power. As one constraint gets solved, capital tends to move towards the n

I’d rather own the index here. The handful of mega-cap leaders may continue outperforming, but buying them after a strong run means accepting much greater concentration and valuation risk. If earnings or guidance disappoint, the same stocks carrying the market could also lead the correction. An index lets me participate in the AI and tech rally while retaining exposure to financials, industrials, healthcare and other sectors that could take over leadership if the rally broadens. I wouldn’t completely avoid the winners, but I prefer them as part of a diversified portfolio rather than making a concentrated bet. At record highs, diversification may look boring, but I’m happy to trade some upside for less single-stock risk. 📈

The answers are: 1A, 2B, 3C, 4B, 5A, 6B, 7B, 8B, 9C, 10B For Q9, the 2× ETF rises about 20% on Day 1, then falls about 18.18% on Day 2. Starting from 100: 100 → 120 → about 98.18, resulting in a small loss despite the underlying index returning to roughly its starting point. This demonstrates volatility drag from daily resetting.

I’d separate these into earnings-backed and expectations-backed record highs. My top three are TSMC, NVIDIA and Johnson Controls. TSMC and NVIDIA remain at the heart of AI compute demand, while JCI shows how AI spending is spreading into cooling and physical data-centre infrastructure. Its $21 billion backlog is particularly attractive. I’m more cautious on Lumentum after its huge run, and on MPC and VLO because today’s exceptional refining margins may not last indefinitely. CRWD and FTNT have strong fundamentals too, but their valuations leave less room for disappointment. 🥇 TSMC 🥈 NVIDIA 🥉 JCI I wouldn’t chase a stock simply because it is making new highs. At these valuations, I want earnings, cash flow and guidance to keep justifying the price. A great company can still be a poor inves

I think the market is worried about both sides of the equation at once: future supply rising while AI demand may arrive later than expected. Toshiba’s expansion raises the possibility that today’s HDD scarcity and pricing power eventually weaken. But analysts argue the reaction may be excessive. Morgan Stanley’s industry checks suggest Toshiba’s expansion is unlikely to eliminate the HDD shortage through 2028. The bigger risk may actually be demand timing. Morgan Stanley estimates a sizeable US data-centre power shortfall through 2028. If data centres cannot get powered on schedule, customers could delay equipment deliveries, hitting memory, storage and optical suppliers before Nvidia or Broadcom. That creates an awkward combination: more supply being planned for the future, but uncertaint

For now, I’d call it a headline running ahead of the order. Musk confirmed that TSMC and Terafab are in discussions, but there is still no disclosed contract value, capacity commitment or firm timeline. That makes TSMC’s record high partly a bet on what the relationship could become rather than revenue already secured. That said, I can understand why the market likes TSMC here. If Terafab needs enormous leading-edge capacity, TSMC is difficult to avoid. Even a multi-foundry strategy could leave it as a major beneficiary. Intel is the more interesting side of the trade. Musk previously said Terafab planned to use Intel 14A, so bringing TSMC into the discussion weakens the idea that Intel has a privileged position. My verdict: positive signal for TSMC, negative negotiating signal for Intel,

I’m not chasing the S&P 500 above 7,800. 📈 Record highs alone are not a reason to sell, but valuations and market concentration make Q3 earnings especially important. I want to see whether earnings growth and guidance can justify the latest repricing, particularly across AI, memory, optical communications and power infrastructure. The easing 10-year Treasury yield is supportive, but if yields reverse higher or mega-cap guidance disappoints, the market could quickly test how much optimism is already priced in. My approach: keep DCA-ing into broad-market ETFs rather than trying to time the top, while keeping some cash ready for a meaningful pullback. I would rather add more aggressively after a correction than chase a euphoric rally. So I’m still participating, just not accelerating. Ear

I think compute is resting on the more fragile assumption. The bullish compute thesis assumes AI capex can keep growing rapidly and, more importantly, that customers will eventually generate enough economic value from AI to justify all that infrastructure. If monetisation disappoints, hyperscalers could moderate spending surprisingly quickly. Memory is cyclical and vulnerable to oversupply, but demand is increasingly tied to real hardware requirements. AI accelerators need large amounts of high-bandwidth memory, while servers still need DRAM and storage. So I see memory’s risk as more about supply, pricing and cycles, whereas compute carries a bigger valuation and AI-ROI assumption. Both can fall, but if the market starts questioning whether every extra dollar of AI capex produces adequat

If I had $10,000 to invest today, I wouldn’t try to time the perfect entry. I’d put around 50% into broad-market ETFs, 15% into quality financials/dividend stocks, 10% into gold, 15% into short-term fixed income or money-market funds, and keep 10% cash ready for opportunities. “Higher for longer” is both risk and opportunity. Expensive growth stocks and highly leveraged companies could remain under pressure, while banks, insurers and cash-generating businesses may hold up better. At the same time, higher yields make cash and short-duration bonds genuinely useful again. I’d expect rates to stay relatively restrictive until inflation is convincingly under control, so I wouldn’t rush to go all-in. But if the market fell 10–20% without a major deterioration in fundamentals, I’d gradually depl