Behind the bleeding earnings reports lies Google's "anchor" strategy.

After the market close on July 22, 2026, the capital market shackled Google with a contradictory narrative.

The Gemini app reached 950 million monthly active users, a 46% increase from 650 million in October last year. Yet in the same quarter, Alphabet recorded its first negative free cash flow since going public in 2004, with an outflow of $5.9 billion. Its capital expenditure guidance was raised to a maximum of $205 billion, and the flagship model Gemini 3.5 Pro, originally scheduled for release in June, is still trapped in partner testing with no public launch in sight.

The market immediately offered a smooth explanation: Google is burning cash wildly, but returns are failing to keep pace, and even its flagship model is falling behind. The stock price plummeted by more than 6% at the opening the next day, which is the most honest price tag for this line of logic.

This explanation is clear, straightforward, and very likely wrong.

Most people only focus on the "AI arms race" on the financial statements, completely failing to notice that Google is fighting a war of a fundamentally different nature. The goal of this battle is to plant an anchor in the minds of billions of people. All historically successful movements that have planted such anchors share a common trait in their early stages: the books are bleeding, while public sentiment is being captured.

01

The Misread "Lag": The Truth Behind the Model Delay

Let's set aside the ghost story of "Gemini 3.5 Pro facing a difficult birth" for a moment.

What Pichai repeatedly emphasized on the earnings call was not the delayed flagship, but the newly released Gemini 3.6 Flash this week — a lightweight model that sees a double-digit improvement in programming benchmarks while consuming fewer tokens. He also casually dropped a piece of information: Gemini 4 has already begun training, with a roadmap targeting a pace of "releasing a new model almost every month."

The core logic of this series of actions is by no means "building the strongest model," but "getting a sufficiently usable model into the hands of the maximum number of people at the fastest speed."

950 million monthly active users, 22 billion API tokens per day, and nearly 90% of the Fortune 100 companies are already using Gemini Enterprise. The picture pieced together by these numbers is clear: Google is using low-cost, fast-performing models like Flash as a stepping stone to get users accustomed to using Gemini first. The delay of Gemini 3.5 Pro stands out conspicuously on the financial statements; but in the landscape of user mindsets, it may not matter at all. 950 million people are already used to handling daily tasks with Gemini, and a three-month delay for the flagship model is nothing more than a small pebble thrown into a lake.

This is exactly the same logic as PayPal desperately burning cash to subsidize users before the bubble burst, getting the payment network up and running first. Wall Street kept pressing the question: Where is the profit model? Peter Thiel and Elon Musk's answer, translated, is just one sentence: Let 15 million people's wallets live on our network first, and we can talk about making money later.

Google's script today hasn't changed a single word. The subsidy target has shifted from "registered accounts" to "token consumption," and what it wants to lock in is the default entry point for billions of people in the AI era.

02

Burning Cash in the Wrong Place? First Check If What You've Spent It On Is Tangible

The number that the market fears the most — the $205 billion capital expenditure in 2026 — is a cash-flow-devouring monster in the traditional analytical framework. But if we shift our perspective and look at it through the lens of "using bubble money to build real businesses," the picture is completely different.

CFO Ruth Porat (Ashkenazy) omitted a detail that is easily overlooked on the earnings call: about 60% of the infrastructure expenditure in the quarter went to AI servers, and the rest was invested in data centers and network equipment. She added that Google is now "supply-constrained" — demand is so high that its own production capacity cannot keep up, so it has to temporarily rent third-party cloud resources to make up the gap.

This statement completely reverses the nature of the cash burn. When a company admits that "our capacity cannot keep up with demand," it is no longer betting on an illusory future; instead, it is filling the capacity gap of a toll highway that already has a long queue of vehicles.

The hundreds of billions of dollars flowing to NVIDIA and construction sites are turning into tangible, non-shutdownable, immovable physical assets. This is exactly the same underlying logic as Elon Musk raising $12 billion through three share issuances at Tesla's peak valuation, then pouring the money into factories in Shanghai, Berlin, Texas, and the 4680 battery production lines. When the bubble recedes, the factories remain, the production lines are still there, and they generate positive cash flow every day.

What Google is doing today is the exact same thing: taking advantage of the open valuation window of the AI narrative to lock in future funds in advance, and solidify them into indestructible physical infrastructure. With 950 million monthly active users for Gemini and Google Cloud's revenue surging 82% to $24.8 billion, the foundation of this castle has already been poured.

03

The Real Main Battlefield Is Not in Technology, But in the Minds of Billions of People

This is the most counterintuitive layer of this article.

Most people read Gemini's 950 million monthly active users as a "product metric." But they are reading the wrong line. The true identity of this number is the progress bar of a large-scale public sentiment transformation movement.

Generative AI has entered the public eye for no more than four years at most. In four years, Google has made nearly a billion people around the world get used to talking to an AI assistant, asking it to write emails, analyze documents, and debug code. This is not just technology diffusion — it is behavior reshaping. Every conversation and every API call is driving a default anchor into the user's brain: When you have a problem, open Gemini first.

Once this behavioral habit is solidified in billions of people, Gemini will no longer be an "optional product" — it will become a "basic setting" of digital life. Just like you don't say "I choose to use electricity from the grid" today, you just press the switch. Whoever owns that default option will be sitting on a never-stopping money printer.

Back then, the global warming narrative pushed Tesla to the altar along the same path. It didn't invent any new technology, but it spent 20 years giving the world a lesson in mindset transformation, making "electric vehicles equal the future" a factory-default setting in the minds of a generation. By the time traditional car companies realized it, the territory of public sentiment had long been no longer dominated by gasoline.

Every mindset transformation in human history has no surplus on its books in the early stages. In the 19th century, missionaries traveled around the world to build hospitals and schools, and they never relied on profit statements to sustain themselves; in the 20th century, when giant TV ad companies burned money like firewood during prime time, the returns took years to grow out of consumers' minds.

Pichai said Gemini 4 is a "very ambitious effort" that is being injected with "massive computing power." Google Cloud's revenue surged 82%. Capital expenditure guidance hits $205 billion. When these numbers are put together, they tell a story: a civilization-level infrastructure is being deployed globally, and the default thinking paths of billions of people are being paved along with it.

04

Conclusion: A Moat Deeper Than Financial Statements

The market sees negative free cash flow, sees the model delay, sees the stock price drop by 6%, and quickly labels it as "risk accumulation."

But truth rarely lives in the minds of the majority.

Google's bet is: when the AI era fully arrives, the default language, default thinking framework, and default interaction of billions of people's digital lives around the world will all run on the Gemini track. To obtain this "default right," the first negative cash flow report card in 20 years is just the beginning of the bills to be paid.

The math is simple: once the mental anchor is planted in the heads of a billion people, the path dependence that grows out of it will be the deepest moat on Earth. It is far more durable and harder to subvert than the benchmarking advantage of any single model.

History repeats one single truth: bubbles are fuel, and the castle is the destination. The most important construction blueprint for building the castle is not on the financial statements, but on the pre-laid default track in the minds of billions of people.

Analysts are still focusing on next quarter's capital expenditure and model release schedules. Google is already laying tracks for the next decade. Where these tracks ultimately lead is not decided by Wall Street's models, but by the 950 million people who open Gemini every day.

This article is compiled based on public materials, for information exchange only, and does not constitute any investment advice.

This article is from the WeChat Official Account "Silicon Starlight", written by Xingguang, and published with authorization from 36Kr.