morning_wind

Assumption:

Will it open high and close low again today, just to trick us? $KLA(KLAC.US)$Direxion Semicon Bull 3X(SOXL.US). On the other hand, in the US stock market, it feels like the focus of urgent value is shifting even more towards hardware.

Current positions, for reference only.

Hold on! Hold the door!

$XL2CSOPHYNIX(07709.HK) Data shows SK Hynix's depositary receipts rose 14% on the Frankfurt Stock Exchange in Germany. Let's see how much we can recover tomorrow.

The number of daily trading orders is very high today. Sold $Meta Platforms(META.US), further increasing exposure to AI. Part of the buying is for short-term rebound catching, and part is for building positions in $KLA(KLAC.US) and $Applied Materials(AMAT.US) as long-term holdings. If the rebound goes smoothly, subsequent short-term positions will continue to be moved into $VanEck Semiconductor ETF(SMH.US) for style distribution.

The first step is a significant increase in capex, occurring in the first 1-2 years. The second step is to continue increasing capex substantially, with free cash flow nearly exhausted or even turning negative, and dividends reduced, which is a matter for 2026. Now it seems the third step has begun. $Alphabet(GOOGL.US) wasn't satisfied with just long-term bond financing and even kicked off equity financing, and $Meta Platforms(META.US) surprisingly followed suit. It's very clear how FOMO the industry leaders are, showing a firm bullishness on AI. Considering the upcoming IPOs of three behemoths, the absorption of liquidity (whether by draining non-tech sectors, the crypto circle, or consuming macro liquidity) will be very fierce. Fluctuations in liquidity expectations will amplify market volatility. The overnight market is turbulent, just recording this.

$XL2CSOPHYNIX(07709.HK) After a big rally, scary stories are bound to come. To preserve the interaction, nvda halved the memory allocation. First, it didn't cut HBM. Second, it did so because the memory was insufficient. Why is the memory insufficient? Because besides GPUs, now CPUs/XPUs also need it. This is a sign of shortage.

$NVIDIA(NVDA.US) can't stand it anymore. Sell a bit of $Alphabet(GOOGL.US) to support Dazi. The storage stocks next door are rising so happily, can HBM shipments directly turn into computing power? Without GPU shipments, where would HBM shipments come from?

$Taiwan Semiconductor(TSM.US) Making a separate post to explain why I think it's rising: After the agent, the overall focus of AI chips has become more decentralized and diversified. Mobile, Apple, CPU, AMD, ARM, various ASICs, and HBM are becoming more balanced in terms of weight. On one hand, this has led to a broad-based rise; on the other hand, TSMC's production capacity is facing increasingly diverse demands from customers, all of which are urgent. Although TSMC has expanded capacity, after several months of transmission by the agent, the most critical bottleneck has finally reached the manufacturing and packaging stage. The demand for capacity is urgent, and it's not just NVDA dominating the market. What to do? Raise prices!

$JOYY(JOYY.US), cash equals market cap, a very aggressive dividend plan. Even if it's a cigarette butt, it's made of gold. Bought a position last night. Hope this can set an example for the old guard of Chinese concept stocks and kick things off. The $PDD(PDD.US) lottery results are out tonight, learn something from it.

$AMD(AMD.US) AMD and OpenAI are both racing towards trillion-dollar valuations. A few months ago, the two companies signed a bet, and now they are in the final stage of mutual commitment: OpenAI will deploy up to 6GW of AMD Instinct series GPUs in phases, with the first 1GW batch expected to start in the second half of 2026; in return, AMD will grant OpenAI up to 160M equity subscription rights (approximately 10%), linked to delivery/market cap milestones. The market reacted strongly to this deal, bringing a lot of upside potential, but it also comes with significant dilution, execution, and infrastructure risks.

In the CPU sector, almost all institutions missed the market rally in April$AMD(AMD.US)$Intel(INTC.US)

$Taiwan Semiconductor(TSM.US)$Alphabet(GOOGL.US), thanks to the 13F, the quality of the two StockPro picks the OP is responsible for has been verified once again.

$AMD(AMD.US) The 410 call expiring mid-month also looks like it's about to be called away 😅, and as expected, $XL2CSOPHYNIX(07709.HK) is also scary today. It's time to start a phased reduction of positions.

Undoubtedly, AI will affect and replace a large portion of current knowledge/information-based white-collar jobs (including mine). The speed of this process may mainly depend on how quickly the entire social organization can rebuild for AI. Fortunately, in some cases, behind such massive transformations, there are also opportunities for compensation, like a decent severance package for buying out seniority.$Alphabet(GOOGL.US)$Taiwan Semiconductor(TSM.US)$SMIC(00981.HK)

$Microsoft(MSFT.US), $Alphabet(GOOGL.US), and $Amazon(AMZN.US) — these three cloud giants' capex have all reached $200 billion. The primary market valuations of the two unlisted companies, OpenAI and Anthropic, are already very close to one trillion dollars. As long as AI continues to develop at a high speed, any pullback caused by concerns about demand, macro policies, or geopolitics is a good opportunity for rebalancing and enhancing returns.

$XL2CSOPHYNIX(07709.HK)$SMIC(00981.HK) Take a group photo for memory

$Alphabet(GOOGL.US) Are you starting up too? Sold some 370 calls expiring on the 51st, welcome to the meltdown.

$NVIDIA(NVDA.US) is catching up on gains, bullish on new highs.

Portfolio bought 5% $SoftBank(SFTBY.US), optimistic about the business of its two sectors: 1) OpenAI will reverse the temporary product disadvantage against Anthropic. 2) AI chips require a large amount of CPUs, and Arm will capture the largest share here.

Deepseek v4 is here. If we include the next-generation models from Anthropic/OpenAI that haven't been publicly released yet, China's cutting-edge models have made significant progress and are catching up closely, but there's still a gap of about 3-6 months.$SMIC(00981.HK)$Taiwan Semiconductor(TSM.US)

$Taiwan Semiconductor(TSM.US)The signal released by the interaction between the two parties is very clear. The previous rumors about something like '27 Wutong' are baseless. This signal means TSMC has at least eliminated a major geopolitical risk, worth at least two limit-up days. An all-time high is inevitable.

Did not sell any $Alphabet(GOOGL.US).