Earnings season, the nearly 65 billion AI bill starts to be settled this week

Over the past two weeks, AI trading has been almost entirely driven by sentiment and narratives: first, TSMC delivered a perfect earnings report but was sold off due to excessive spending; then, Kimi K3 dealt a sharp blow; followed by Monday's rebound from oversold levels, recovering some ground, with the Semiconductor Index up 1.9% and memory stocks leading the charge. Next up are hard data points—starting Wednesday, tech giants Google, Tesla, and IBM will 陆续 release their results.

The market estimates that capital expenditures for data centers by the five hyperscalers this year will surge approximately 79% year-over-year, reaching $644 billion. Over the past two years, it was this escalating investment that supported the valuation of the entire chain, from NVIDIA to TSMC and Hynix. Now, the market is shifting from "pricing in promises" to "pricing in execution."

Following the industry chain makes it clear why these earnings reports are a stress test for the whole sector. Capital expenditure is the lifeblood at the upstream end of this chain: downstream are the cloud providers (Microsoft, Google, Amazon, Meta) throwing money around; midstream are the companies fulfilling orders for GPUs and custom chips (NVIDIA, Broadcom); and further upstream are foundries and memory manufacturers (TSMC, Hynix). Everyone's valuation rests on the assumption that "cloud providers will continue to increase capex." Therefore, what's truly being scrutinized during earnings season isn't just one company's EPS, but whether this assumption still holds: as long as capex guidance continues to be revised upward, oversold chip and memory stocks have ammunition for a counterattack; if any player starts contracting, or if massive depreciation eats into profits before revenue materializes, the foundational premise for valuations along the chain will loosen.

Specifically regarding these earnings reports, there are several figures more noteworthy than EPS: whether cloud business growth rates are dropping tiers, whether capex guidance indicates increased investment or caution, and whether profit margins can withstand the depreciation of newly built data centers. Google and Tesla, reporting after market close on Wednesday, happen to correspond to different aspects of this exam.

First, look at Google. The consensus expectation is Q2 revenue of approximately $116.9 billion, up 21% year-over-year, and net income of about $35.3 billion, up 25%. However, the real battleground is Google Cloud—in Q1, it saw 63% YoY high growth, and most analysts set the passing bar for Q2 around 60%. If it falls below this, it will be interpreted as a signal that enterprise AI demand has peaked. Its capex guidance for this year has already been raised to $180–190 billion. Bulls argue this investment builds a wider moat for cloud and models, and cloud order backlogs will gradually convert to revenue; bears worry that data center depreciation will hit before revenue does, squeezing margins. Another often overlooked aspect: will AI-driven search erode ad exposure efficiency? In Q1, search ads generated $60.4 billion and YouTube ads $9.9 billion. The stability of this core base directly determines whether Google has the confidence to keep burning cash. The market's divergence boils down to two answers to one question: "Is high investment a moat or a bottomless pit?"

Next, look at Tesla, which corresponds to another question: can high sales growth translate into profit. In Q2, it delivered a record 480,000 vehicles, up 25% YoY, far exceeding previous expectations. Yet, the divergence here lies precisely in this—whether this volume reflects genuine demand or was achieved through aggressive promotional pricing. Thus, gross margin is what's truly under scrutiny: investors want to see automotive gross margin (excluding regulatory credits) hold steady at around 12.5% from Q1, proving that volume growth wasn't just stacked via discounts. Consensus expects revenue of ~$26.4 billion and EPS of $0.54, but more cautious institutions like Deutsche Bank have lowered EPS expectations to $0.36 due to promotional pricing. The options market is pricing in a single-day move of roughly 7.6% post-earnings. As for its sky-high P/E ratio of 349x, it's no longer supported by car sales alone but bets on the imagination space of Robotaxi and Cybercab—so substantive progress on the call conference often matters more for stock price direction than the quarterly numbers themselves.

Both sides need to be laid out clearly. On the bullish side, hard evidence of demand is on the way: S&P's overall Q2 earnings growth expectation has been revised up to 26%, semiconductor sales grew 79% YoY in Q1 with even higher expectations for Q2, all pointing to expanding demand. On the bearish side, market tolerance for error is already very low—"pricing in execution" means that any weakening in capex guidance or margin erosion from depreciation could trigger a sell-off led by profit concerns. This is not an earnings season where you can get away with muddling through. Wednesday's Alphabet Google Cloud growth rate and capex guidance will be the first positive answer from a hyperscaler following the questions on Kimi's efficiency, and it will also determine whether the chain from foundries to memory can secure ammunition for a short-term counterattack.

The above content represents personal views only and does not constitute investment advice. Follow Vivian, wishing everyone gains and sellers lose~