According to Woofun AI, after Alphabet's earnings report was released, Google's stock price plummeted more than 10% after the market, almost erasing the gains of most of the past year. Market panic spread, but deep logic points to valuation mistakes during the transition period rather than a basic market crash. The slowdown in traditional search advertising growth and negative free cash flow led to short-term sell-off. However, the doubling of Google Cloud's revenue, the explosion in TPU export sales, and the peak of the Gemini 4 model's capabilities revealed strong support for the new engine.
This huge divergence between short-term financial data and long-term strategic value is reminiscent of similar market miscarriages experienced by Microsoft in the early stages of cloud and AI transformation. Currently, Google is facing a similar test of “perception time difference”. The long-term logic has not changed; on the contrary, it has been further consolidated by the accelerated implementation of new businesses. The market's lack of patience with the 2B business monetization cycle has led to this round of sharp price fluctuations, but this has provided a window for long-term funding to reevaluate the closed-loop value of Google's AI full stack.
Breaking the focus of market fears, the pressure Google faced in the third quarter was indeed reflected in the fact that several key financial indicators fell short of expectations, but this was not a complete slowdown in the basic market. Alphabet's total revenue for the third quarter reached $127.6 billion, up 25% year over year. Although overall services were still growing, it was slightly lower than the market's agreed expectations; of this, advertising revenue was about $82 billion, which also failed to meet Wall Street's high expectations. The performance of the core advertising business was particularly critical. Search and other advertising revenue increased 12% to US$63.3 billion year over year, which was not only lower than the market forecast of US$64.9 billion, but also far below the bottom line of 14% growth set by buyer agencies. In contrast, the growth rate remained at 17% in the second quarter. What worries the market even more is management's guidance on the future. The cloud business, which has just set a record high profit margin of 36.2%, is expected to face profit margin pressure in the fourth quarter due to increased use of third-party computing resources; on top of the capital expenditure of 19.5 billion to 205 billion US dollars this year, capital expenditure in 2027 will also “increase significantly.”
This means that Google is simultaneously under triple pressure: the growth rate of its old cash business is slowing down, monetization of new AI models is lagging, and infrastructure investment continues to increase. The combination of these three forms a standard “kill valuation package” in the current US stock market, but the question is, is the market overly obsessed with current figures while ignoring the structural reasons behind these pressures?
Deeply dismantling the search business, the so-called “AI kills Google search” is still a false proposition. The truth is hidden in the misalignment of user behavior migration and commercialization strategies. The growth rate of Google's overall search query volume (including business queries) this quarter hit the highest level in the past five years. Under the AI model, visual searches surpassed 50 billion per month, and the vast majority were “incremental searches” rather than a simple replacement for traditional searches. CEO Sundar Pichai repeatedly emphasized during the conference call that just like switching to mobile devices ten years ago, Google is focused on creating the best search user experience first and then monetizing it. As a result, although the average query length of AI has reached three times that of traditional searches, monetization efficiency is relatively low, and more interactions have yet to bring in advertising revenue comparable to traditional searches.
Notably, the ad experiment has shown positive signs, and the click rate of sponsored product lists and organic product listings in AI Mode is almost equal. Until recently, Google announced that it will launch AI-mode advertising internationally for the first time and launch multiple rounds of conversation ads in the US. This shows that its strategy is to first let users get used to new search methods and then gradually expand proven ad formats. As user habits shift from “knocking on a few keywords” to “a long chain of AI multi-round interaction,” the traffic pool has indeed grown, but in order to maintain entry-level stickiness, Google actively chose to slow down commercialization by half a beat. From a valuation perspective, AI search at this stage should not be viewed as a burden that is dragging down performance, but rather as a 'bullish option' that has not been fully realized. If subsequent ad load rates, conversion rates, and ability to monetize a single query gradually increase, Google has every chance of unleashing new revenue growth in the existing search ecosystem.
The sharp fluctuation in market sentiment is due in large part to the conflict between Wall Street's impatience and CEO Sundar Pichai's engineer mentality. Current funding is extremely utilitarian, and requires seeing an immediate surge in EPS and advertising revenue in the current quarter. Especially in the context of rising AI capital expenditure, every unrealized growth point may be viewed as evidence of a decline in future returns. Sundar Pichai himself has always been a typical engineer-thinking CEO, emphasizing product experience, technological progress, infrastructure, and long-term returns. His personality is extremely lacking in the genes to “paint the big cake” and use grand narratives to incite market sentiment. Faced with questions from analysts, he honestly explained Q4's base pressure and computing power bottlenecks. The combined deceleration pressure of the old business was reflected in the financial statements before the commercialization results of the new business. Naturally, the stock price gave negative feedback on the spot.
This conflict between short-term performance and long-term narratives made it difficult for the market to understand Google's strategic strength during the transition period, leading to a sharp correction in stock prices after the earnings report was released. However, this pricing bias based on short-term financial data often overlooks the painful period that tech giants will inevitably experience during the transformation process, as well as substantial breakthroughs in the underlying structure of new businesses.
Looking away from the old, slightly fatigued business, Google Cloud's boom has become the brightest part of Google's new engine. The double increase in revenue and profit and the huge backlog of orders show the true intensity of demand on the enterprise side. In the third quarter, Google Cloud's revenue surged 103% year over year to 30.7 billion US dollars. As a comparison, the growth rate in the previous quarter was about 82%, and violence can continue to accelerate in the amount of 10 billion US dollars, which in itself shows strong market demand. What is more noteworthy is the profit performance. Google Cloud's operating profit for the quarter reached 11.1 billion US dollars, more than tripling year on year, and the operating profit margin directly jumped from 23.7% in the same period last year to 36.2%, reaching a record high.
Meanwhile, the cloud business's backlog of unfulfilled orders reached $561 billion, of which more than 55% are expected to be recognized as revenue within the next 24 months. On the corporate side, Google revealed that sales of 'knowledge workplaces' in all of its AI services have surpassed 100 million, and nearly 90% of the Fortune 100 companies use Gemini Enterprise. For example, BNP Paribas deployed the system for more than 65,000 employees, and Tata Steel launched more than 300 professional AI agents within nine months to optimize the supply chain and customer service. These cases prove that AI is actually taking over the actual workflow within the enterprise from early demo demonstrations and partial pilots. Once enterprises start using AI for programming, knowledge work, data analysis, and business process automation, it will drive the demand for computing power, storage, database, and security services, and help Google turn AI into a complete set of infrastructure and services that can be continuously sold to enterprises.
According to data compiled by Woofun AI, the truth behind the deterioration of free cash flow is not a simple 'AI bottomless hole', but rather conceals the rapid expansion of TPU's export business and the impact of one-time expenses. Many people saw free cash flow negative for two consecutive quarters and far below expectations. The first reaction was a deterioration in fundamentals, but the dismantling of CFO Anat Ashkenazi and cloud business CEO Thomas Kurian revealed a different picture. In the third quarter, Alphabet reported negative free cash flow of $9.065 billion, which worsened further from the negative $5.855 billion in the second quarter, but this included a final payment of $5.2 billion in the EU's historic fines in 2018.
Excluding this $5.2 billion cash payment of EU fines, the adjusted free cash flow was about negative $38.65 billion, which is an improvement of about $1.99 billion compared to the second quarter. Additionally, conventional infrastructure capital expenditure is $51.4 billion, which is mainly used to support the expansion of cloud business and the preparation of TPU. The “negative $9.1 billion” free cash flow that frightened the market actually overshadowed the huge opportunities offered by TPU export sales. Google's TPU is changing from an “internal chip that you use to save money” to a “platform-based product that sells computing power.” This logic completely fulfilled the real money order in the third quarter, but the financial confirmation cycle was completely different from traditional cloud leasing, which led to huge cash outflows in the early stages.
The explosion of TPU export sales marks the success of Google's strategy to shift from internal chips to platform-based products. Inventory backlogs actually prepare for contracts rather than slow sales. The performance of the latest training chip TPU 8T is 2.7 times higher than the previous generation, and the 8I inference chip is 80% more cost-effective. It has strong pricing power in the current environment where accelerated computing power is scarce. In terms of software ecology, Google and Meta collaborated to launch TorchTPU to improve the training, inference, and deployment experience of PyTorch workloads on TPU and reduce developer migration costs.
In terms of distribution channel expansion, the TPU system is no longer just locked in Google Cloud; it has begun to be deployed directly into the customer's local data center through cooperation with Oracle and Crusoe AI (a joint venture with Kuroishi). Currently, nearly 100 heavyweight customers (including Apple, Anthropic, Citadel, Jump Trading, etc.) are using TPU. Why did TPU sell better and have worse cash flow during the season?
Due to the different financial confirmation cycles for overseas sales of complete TPU systems (including hardware, storage, optical internet, and software stacks), Google revealed on a conference call: “The inventory of all TPU systems corresponds to a signed contract, and we only purchase parts to build inventory when we have a contract.” The customer has already signed a dead contract, and Google must first spend real money to buy parts and save inventory. This is reflected in a huge cash flow outflow for the third quarter; until the 2027 system delivers and bills are issued on a large scale, this money will only turn into massive revenue and cash flow. This is a different cash flow cycle from self-built data centers. Early inventory accumulation will depress operating cash flow. As the system is delivered and customer payments are made, the corresponding working capital occupation can only be gradually released.
The peak of the Gemini 4 series model capabilities has added the most critical piece to Google's full-stack AI puzzle, and the surge in API calls has verified the reality of market demand. Since the launch of the Gemini 4 series, Google has not only stabilized its position, but has even gone straight to the front in some key reviews. In addition to Gemini 4 Argon, which was released two weeks ago, the latest Gemini 4 Carbon directly topped the world with a score of 59 in the latest update of the authoritative Artificial Analysis Intelligence Index, completely crushing Argon in full-stack web development, multi-modal understanding, scientific reasoning, and cybersecurity.
Coupled with the newly released world model Genie 4, the real-time interactive environment simulation capabilities are further expanded, providing a foundation for robot and intelligent body training. Reflected in actual call data, the number of tokens processed by Google's first-party model API is close to 32 billion per minute, an increase of 40% over the same period last year, which is more than four times that of the same period last year; the number of users of the Gemini app has also officially passed the 1 billion mark.
With stronger model capabilities and lower inference costs, Google has more opportunities to deeply embed AI into search, Workspace, and Gemini applications, and can further promote enterprise customers to use Google Cloud, call models, and deploy agents; in turn, cloud infrastructure and self-developed TPUs can also support model training, inference, and large-scale deployment. This is full-stack logic, which allows the model, product entry, cloud service, and the underlying chip to form a closed commercial loop that promotes each other. The model itself is not an isolated business, but the core driving force of the entire ecosystem.
The core of the mistaken logic lies in the “perceived time lag” between the 2B business and the 2C business, causing the market to underestimate the value of long-term contracts. On the one hand, the search advertising business, which grew 12% but fell short of expectations, Google Cloud, which surged 103%, and the cloud business backlog of orders of up to 561 billion dollars; on the other, stock prices plummeted 10% after the market.
This strong sense of division makes it hard not to think of Microsoft before. At a critical stage in Microsoft's transformation to the Azure cloud and early Copilot, Wall Street also played almost exactly the same script: as long as the traditional Windows OEM business or personal computing sector lagged behind a bit during the season, or a short-term surge in capital expenditure suppressed free cash flow, even when Azure's underlying subscription volume skyrocketed, the market still blew up. Today's Google is facing a set of contradictions that are not exactly the same, but the underlying logic is highly similar. The rules of the 2B business have always been 'slow heat, lock orders, and post-cash'. Whether enterprise customers are migrating workloads to Google Cloud, whether they are starting to use TPUs in production environments, and whether they are willing to sign multi-year computing power contracts, these changes are difficult to judge from everyday experience alone.
The first step is a lengthy internal process. For example, Bank France and Pakistan deployed Gemini and Tata Steel to launch more than 300 industrial agents for 65,000 employees. The second step was to lock in orders and pre-empt costs, and computing power and system requirements were locked into a backlog of up to 100 billion US dollars. In order to fulfill the contract, Google even had to pay its own real money to advance TPU inventory and bear the negative pressure of free cash flow; the third step was a commercialized centralized blowout, waiting until the computing power was delivered and the company actually took over In the daily workflow, customers move from “signing a contract” to “actually consuming tokens and continuing to renew the contract,” and only then will revenue and profits be released centrally at a certain node in the future.
Looking back at Microsoft's history, it is confirmed that Google's long-term logic has not changed, and that the current decline is a long-term funding opportunity. Sundar Pichai defined the upcoming 2027 as Google's 'golden year of productivity', and the CFO also clearly designated 2027 as a 'harvest year' for large-scale recognition of TPU revenue. Admittedly, the slowdown in search ad growth, negative free cash flow, and a significant increase in capital expenditure in 2027 are unavoidable real issues, and they are also the focus of repeated market games in the fourth quarter. However, if we look back at how Microsoft has gone through the deep-water zone of 'old business decelerating and new infrastructure', we may be able to be a little more certain.
When a giant with multi-billion C-side entrants, big models returned to the throne, and had a backlog of 561 billion dollars of orders on the B-side, it was knocked out of a 10% deep hole simply because of “actively delaying search and monetization” and “cash flow time lag for signing and preparing the goods.” This is likely another misstep that market sentiment sends to long-term capital under “perceptual time differences.” The increase in Gemini 4.0's capabilities, Google Cloud's 103% growth rate, and the explosion of TPU export sales together form a solid foundation for Google's AI counterattack. The current valuation adjustment is not a deterioration in fundamentals, but rather a market misjudgment of the new business monetization cycle, which provides long-term investors with an opportunity to restructure.