4.28 US Stock Market Recap — Is the Semiconductor Dream Coming to an End? Who Can Take the Baton Next?

I haven't been posting many reviews lately, mainly because there really isn't much worth reviewing. Every day, the market just goes up as soon as it opens, with the semiconductor sector taking turns leading the gains. Yesterday, NVDA hit a new all-time high after a nine-month consolidation period.

SOXL has been the absolute biggest winner in this rally. The concentration of capital within it is extremely severe. Influenced by the pullback in the optical module sector, it finally ended its 18-day winning streak yesterday. With several major tech giants set to report earnings this week alongside the FOMC meeting, capital is likely to lean towards risk aversion. Coupled with the fact that the CTA buying that fueled the earlier rally is finally nearing exhaustion, and the nearly $20 billion in pension fund rebalancing selling pressure expected by month-end, this could press the pause button on this frenzied market run.

The main dilemma of the current earnings season is essentially whether Capex can continue to expand. The logic behind the earlier semiconductor rally was that engineering teams would desperately consume all available computing power because everyone is afraid not of spending too much, but of not spending enough and missing the AGI window. This creates a perverse incentive: the more aggressively you spend, the more respect you command.

The market has directly extrapolated this logic chain: more tokens → more intelligence → ultimately leading to AGI. Personally, I'm cautious about this extrapolation, but I must admit that EPS is indeed accelerating at the fundamental level, and every link in the supply chain is tight—power, CPUs, GPUs, copper, fans, even engineers are in short supply.

But this earnings season, the market faces a similar problem to the one before last year's AI bubble: can Capex be raised further? The current predicament is that in an environment of rising input costs, if capital expenditures merely remain flat, that actually equates to a deceleration. But can companies' FCF still support it? They need real performance to back it up. At this stage, market logic remains linear: more Capex beats, more EPS beats = exceeding expectations.

But in the semiconductor industry, which has already consumed too much optimistic expectation in the earlier phase, how many companies can withstand the extremely demanding expectations for perfect earnings? Objectively speaking, INTC's earnings last week weren't perfect (they just told a good story), but capital was willing to buy into it. So, how many companies can replicate this glory going forward? I don't know, but at this current stage, the risk-reward ratio is already severely imbalanced, and gambling on the remaining upside isn't very meaningful.

So, where should the market's direction be next? I think this is the question everyone is most concerned about, especially for friends who feel they missed this rally. Today, I'll mainly talk about a few software layer companies I'm relatively bullish on.

According to the latest research from MLQ Research in April 2026, corporate AI budgets have fully transitioned from the "pilot exploration" phase to the "production deployment" phase. However, the dividends brought by AI are not evenly distributed; they exhibit a clear stepwise transmission effect.

Currently, the first wave of AI spending dividends has been captured by the infrastructure layer (hardware). Meanwhile, the data and analytics platforms that form the core foundation of AI architecture (Snowflake, MongoDB) and the observability platform (Datadog) are on the eve of an earnings explosion inflection point.

The widespread deployment of AI agents is actually beneficial for software stocks. As the complexity of enterprise IT architecture increases exponentially, the demand for underlying data governance, storage and retrieval, and system monitoring is expanding rapidly. For software companies like MDB, DDOG, and SNOW, the core investment logic is: without high-quality, retrievable enterprise-specific data, there are no reliable AI agents; without powerful observability tools, the stable operation of AI systems cannot be guaranteed. In my own use of AI, I've gradually realized that enterprise-grade databases will become the next bottleneck after tokens.

Snowflake (SNOW) The Most Flexible Turnaround Play

Despite weak technicals and facing fierce market competition, its fundamentals are undergoing a substantial reversal. Signing seven nine-figure deals in a single quarter, product revenue growth returning to 30%, combined with extremely bullish sentiment in the options market and Wall Street's expectation of up to 61% upside, make it the current investment play with the best odds.

Datadog (DDOG) The Most Certain Steady Growth Play

As the only company achieving GAAP profitability, DDOG demonstrates an excellent business model. The IT architecture complexity brought by AI is directly translating into orders for its observability products (e.g., an 11x increase in MCP tool invocations in a single quarter). Suitable as a core holding for the cloud software sector. However, it's nearing overbought territory and may face some profit-taking pressure in the short term.

MongoDB (MDB) The AI Infrastructure Poised for Takeoff

Although management admits AI hasn't yet contributed substantial revenue, its Atlas database has crossed the $2 billion ARR milestone. As enterprises shift from AI pilots to full-scale deployment, demand for high-performance NoSQL databases and vector search is bound to explode. Short-term bearish options sentiment may provide better opportunities to build positions on dips.

Overall, against the macro backdrop of AI capital expenditure shifting towards the software application layer, MDB, DDOG, and SNOW form the "iron triangle" of enterprise data and infrastructure. Looking forward to their performance next. As for the largest software company, MSFT, I posted about it in the second thread; interested friends can take a look.