I favour the uranium supply chain for durability. AI data centres may accelerate nuclear demand, but uranium benefits from the broader reactor fleet and fuel-security needs, rather than depending on any single SMR design reaching commercial scale. BE is interesting because fuel cells can address the nearer-term problem: data centres need reliable power before new nuclear plants can realistically arrive. SMR offers the biggest upside if deployments scale, but also the greatest execution, financing and regulatory risk. My ranking: uranium for the strongest long-term risk/reward, BE for the nearer-term AI power bottleneck, and SMR as speculative optionality. The AI electricity shortage looks structural, but that does not make every power stock a structural winner. I would rather own the bott
B. GOOGL for me. Meta may have the larger social distribution opportunity, but Google already sits much closer to commercial intent through Search, Shopping, YouTube and its broader ecosystem. If Gemini agents can move users from “search” to “act and transact”, Google has a clearer path to monetisation. META could ultimately surprise if it converts its huge user base into commerce, but that requires changing user behaviour and proving transaction economics. For now, I would rather own GOOGL for the AI-agent trend. It has both distribution and an existing monetisation engine to build upon.
I favour the uranium supply chain for durability. AI data centres may accelerate nuclear demand, but uranium benefits from the broader reactor fleet and fuel-security needs, rather than depending on any single SMR design reaching commercial scale. BE is interesting because fuel cells can address the nearer-term problem: data centres need reliable power before new nuclear plants can realistically arrive. SMR offers the biggest upside if deployments scale, but also the greatest execution, financing and regulatory risk. So I see it as: uranium for the durable structural thesis, BE for nearer-term AI power demand, and SMR as the higher-risk optionality. Given the sector's high beta, I would expect plenty of momentum-driven volatility even if the long-term power-demand thesis remains intact.
B. GOOGL for me. Meta may have the larger social distribution opportunity, but Google already sits much closer to commercial intent through Search, Shopping, YouTube and its broader ecosystem. If Gemini agents can move users from “search” to “act and transact”, Google has a clearer path to monetisation. META could ultimately surprise if it converts its huge user base into commerce, but that requires changing user behaviour and proving transaction economics. For now, I would rather own GOOGL for the AI-agent trend. It has both distribution and an existing monetisation engine to build upon.
Memory keeps running: SK Hynix +7.05% to a record, Micron +2.75%, SanDisk +1.51%. The bull case remains powerful: AI/HBM demand is squeezing supply, while SK Hynix's massive shareholder-return programme adds another tailwind. But at record highs, expectations matter more than the story. Much of the good news may already be priced in, so any disappointment in pricing, HBM demand or shareholder returns could trigger a sharp pullback. Micron reports on Sept 30, giving investors a major read on DRAM/HBM pricing, margins and AI demand. My pick: wait for Micron rather than chase SK Hynix at records. The memory cycle still looks strong, but I would rather sacrifice some upside for confirmation than pay peak expectations today.
A. DELL gets my vote, and DELL is the ONE I would put on my watchlist. AI infrastructure still looks like the theme with the strongest runway. Dell's $60.9bn of AI-server orders and $95bn backlog suggest this is not merely an AI narrative anymore; customers are actually committing huge amounts of capital. The interesting part is that demand is also spreading into storage, networking and traditional servers. The catch is valuation. DELL has already had an enormous run, so I would not chase a vertical move simply because it made another high. Fresh highs backed by rising earnings and guidance can keep making fresh highs, but the margin for disappointment gets smaller. Healthcare royalties such as RPRX and HALO are attractive for their recurring cash flows, but for growth momentum, I still f
C: AAPL stays between $310 and $330. Apple events often produce plenty of excitement without producing an equally dramatic stock move. Much of the foldable iPhone story is already anticipated, so simply confirming what the market expects may not be enough to push AAPL decisively above $330. At the same time, I would not bet heavily on $310 either. The new product cycle, potential foldable iPhone and AI developments provide enough catalysts to support sentiment unless Apple seriously disappoints on pricing or execution. The interesting part is that AAPL already touched $330.81 last week before pulling back. That suggests $330 is meaningful resistance, while roughly $310 remains an important downside area. So my pick is C. I expect volatility during and after the event, but probably m
I think the supply-chain rally has further to run, but leadership may shift from Nvidia to the bottlenecks. Nvidia's $96.2bn quarter and 117% Data Center growth confirm that AI infrastructure demand is still accelerating. More importantly, Nvidia's supply commitments have surged to $279bn, primarily for memory. That makes the margin pressure revealing. If scarce memory is expensive enough to compress Nvidia's margins, the same cost pressure can translate into pricing power for memory suppliers. DRAM and HBM demand already exceeds supply, strengthening the case for $MU and $SKHY. Connectivity and optical names can benefit too as ever-larger GPU clusters require more bandwidth. I would therefore avoid chasing the whole basket after an earnings spike. Nvidia has proved the demand story; now I
A: Chase the Winner. I would rather pay a fair premium for a business whose earnings, cash flow and competitive position are still strengthening than buy a falling stock simply because it looks cheaper. Momentum backed by fundamentals can persist far longer than expected. The key is distinguishing expensive from overvalued. For names like $NVDA, $GOOG or $META, I would watch earnings growth and forward guidance rather than the share price alone. A 30x multiple with rapidly rising earnings can ultimately be cheaper than a 15x stock with deteriorating fundamentals. Buying the dip works when the market has overreacted. But a falling price by itself is not a thesis. Sometimes the dip keeps dipping because the business outlook has genuinely changed. So A for me, but only when the fundamentals
I would wait for Micron rather than chase Friday’s memory rally. The sector’s fundamentals remain attractive, but Friday’s moves were unusually strong relative to the broader market. Micron closed above $1,000 after gaining 6.1%, while SanDisk jumped 11.9%. At these valuations, good news is increasingly priced in. September 30 matters more. Micron’s results should tell us whether AI-driven HBM/DRAM demand, pricing and margins are still accelerating. Meanwhile, the Taiwan labour dispute is a genuine tail risk: unions representing nearly 10,000 workers are considering strike action, and Taiwan is Micron’s largest manufacturing base. I would not short the momentum, but I would not chase it either. Passive flows can push prices higher temporarily; earnings ultimately have to validate them. I w