Written by:潮向研究

At 4 AM Beijing time on July 23, Alphabet will release its Q2 2026 financial report, followed by a conference call at 4:30 AM. This is the first financial report among the "Seven Giants," with Microsoft, Meta, Amazon, and Nvidia following behind. The options market has already priced this night: an implied volatility of 5.3%, corresponding to a company with a market value of $4.5 trillion, with daily fluctuations exceeding $200 billion.

The market treats it like a corporate earnings report to trade; in reality, it is a midterm exam for the entire AI capital expenditure cycle.

If the answer is good, the earnings report seasons for the other giants will go smoothly; if it fails, no one from Nasdaq to A-shares optical modules can remain unscathed.

Six days ago, the market tore open a gap

On July 16, Bloomberg reported, citing ten current and former Google employees, that Gemini 3.5 Pro is months behind its original schedule due to programming capabilities not meeting internal standards. On that day, Alphabet's stock price fell by 4.4%, with about $200 billion in market value evaporating.

The reason for the market sell-off is unrelated to revenue, unrelated to profit, and only concerned with one launch date.

Pichai personally promised in May at the I/O conference that it would be released in June; June has passed, and Google left only the comment "testing with partners." Bloomberg's report added the remaining information: training data was updated once at the end of June, targeting programming, and the results disappointed internally; engineers were frustrated, and some researchers moved to competitors.

Programming is currently the battlefield with the highest gold content in the commercialization of large models, directly determining whether AI agents can independently complete multi-step tasks and also determining where corporate clients will allocate their funds. OpenAI and Anthropic release products weekly, while Gemini 3.5 Pro is only rolling out once a week, deepening the binding between enterprise developers and competitive platforms.

This drop exposed a new contract for AI trading in 2026: the market tolerates burning nearly $200 billion for a year, provided that the model must arrive on time and be capable enough upon arrival. Once the launch date is missed, capital expenditure shifts from "moat" back to "bill."

Tonight's financial report is Google's first performance review under this new contract.

Three numbers that will truly determine the direction tonight

The market consensus is clear: revenue is around $116.8 billion, a year-on-year growth of about 21%; earnings per share are $2.86 to $2.89, a year-on-year growth of about 24%. Google has exceeded expectations in all four previous quarters; simply beating expectations is no longer enough to drive stock prices; the real variables lie within three clues.

Cloud growth and the conversion of backlog orders.

In the first quarter, Google Cloud revenue grew by 63% year-on-year, significantly outpacing Azure's 40% and AWS's 28%, with backlog orders nearly doubling to $462 billion.

CFO Ashkenazi left a key comment last quarter: if production capacity could keep up with demand, cloud revenue could have been even higher. This statement simplifies the issue; what needs to be validated tonight is the speed of capacity ramp-up and how much of that $462 billion is turning into real revenue at the pace promised by management (around half within 24 months). If the growth rate stays above 60%, the story of AI infrastructure continues; if it slows down significantly, the market will immediately revalue all cloud vendors.

Where the money is going and hints for 2027.

Google's capital expenditure guidance for 2026 is $180 billion to $190 billion, more than five times 2023's $32.3 billion, with management already forecasting a "significant increase" in 2027. The cost is reflected in the cash flow statement: free cash flow declined nearly 47% year-on-year in the first quarter, and in June, an additional $80 billion in stock was issued to fund AI expansion. Any specific numbers regarding 2027's expenditure during tonight's conference call could sway stock prices more than the Q2 performance itself. An upward revision of expenditures accompanied by strong cloud growth is aggressive; an upward revision of expenditures accompanied by cloud slowdown is disastrous.

TPU shifting from cost advantage to a second revenue stream.

This is the most imaginative underground line of this financial report. Google has begun selling TPU computing power to external clients, with a computing power leasing company backed by Blackstone planning to rent out TPUs to AI labs starting in 2027; Anthropic signed a letter of intent for a million Ironwood chips; Meta is reported to be considering deploying TPUs in its data centers in 2027. Management previously guided for hardware sales to start contributing revenue in the second half of 2026, with a major focus on 2027. Even a single confirmation of TPU sales progress tonight could be enough to reprice the "Google chain."

Conduction chain: from Mountain View to A-share optical module workshops

The radius of influence of Google's financial report has long exceeded the U.S. stock market.

On the U.S. stock market level, as the first giant to submit its report, Google's conference call will set the tone for the entire earnings season. Cloud growth and capital expenditure guidance will be directly factored into the expected models for Microsoft and Meta; any progress with TPU is a direct stress test for Nvidia. Last November, when the rumor of Meta acquiring TPU emerged, Nvidia's stock fell in a single day, compelling it to publicly claim that "its technology is still a generation ahead of the industry," a scene still fresh in memory.

On the A-share market level, the "Google chain" has become one of the highest-weight narratives in computing power trends.

The deployment mode of TPU clusters (each cluster contains 9216 liquid-cooled chips, interconnected through OCS all-optical switching) means that its unit usage of optical modules, optical switching, liquid cooling, and high-speed PCBs is far higher than traditional architectures. Zhongji Xuchuang and New Yisheng rank among the top three in the Growth Enterprise Market, with order visibility extending to 2027; Saiwei Electronics exclusively manufactures MEMS micro-mirror chips for Google’s OCS. Tonight's capital expenditure guidance serves as the waterline for these companies' orders over the next six quarters. If expenditures are revised upward, the valuation disputes over optical modules will automatically settle; if expenditures fall short of expectations or cloud growth slows, the crowded computing power transactions will face the harshest sell-off of the year.

Three scenarios

Cloud beats expectations, expenditure pace is controllable. Google Cloud's growth maintains above 60%, 2027's expenditure gives direction but is not alarming, and TPU sales confirmation lands. Stock prices challenge the previous high of $408, AI transactions fully recover, and A-share computing chains rise with it. This is the bullish scenario, representing the path implied by the current 43 analysts who give buy ratings, with an average target price of $428.

Performance beats expectations, but feelings of expenditure out of control intensify. Revenue and profit both exceed expectations, but the 2027 expenditure numbers are too aggressive, free cash flow continues to deteriorate, and the new launch date for Gemini still remains unclear. Stock prices surge then retreat, and market sentiment shifts towards a judgment of "revenue growth without profit," passing pressure onto Microsoft and Meta a week later.

Cloud slowdown. Growth falls below 55%, or backlog order conversion is worse than expected. This is a tail risk; once it appears, the impact will cover all cloud vendors and the entire computing power supply chain, leaving Nvidia unable to escape.

Based on the current positioning structure, the probability of the third scenario is the lowest, while the second is underestimated. The market is overly focused on whether there will be a beat or not, while the true odds lie hidden in the Q&A session of the conference call.

At 4:30 AM, the conference call connects, and the first question from analysts is highly likely to be unrelated to Q2 figures. Everyone wants to know one thing: what is the new launch date for Gemini 3.5 Pro?