--- title: "Will Google be the first to cut AI spending?" type: "News" locale: "en" url: "https://longbridge.com/en/news/293482373.md" description: "An analyst predicts Google may cut AI spending due to cash flow constraints and a potential AI bubble burst, citing high capital expenditures and increased debt. However, this bearish view is contested by Google's substantial cash reserves and rising operating cash flow. The article questions whether the analyst's past predictions hold merit, noting significant errors in previous market forecasts." datetime: "2026-07-22T13:38:15.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/293482373.md) - [en](https://longbridge.com/en/news/293482373.md) - [zh-HK](https://longbridge.com/zh-HK/news/293482373.md) --- # Will Google be the first to cut AI spending? The following has been widely circulated today: an overseas analyst believes that Google may be the first to cut capital expenditures, and that the AI ​​bubble will burst. We will analyze his views below, but regarding capital expenditures, the answer will be revealed early this morning. Google will release its Q2 2026 earnings report after the market closes on July 22nd (US time). The conference call is scheduled for 4:30 PM ET on July 22nd, which corresponds to 4:30 AM Beijing time on July 23rd. The earnings report will be released before the conference call. We will discuss Google's latest big data model and internal computing power in tomorrow's live stream, as well as the supernodes at WAIC. In his analytical framework, AI infrastructure investment has exceeded the capacity of companies' cash flow, and the prices of models and computing power are continuing to decline, ultimately leading to insufficient return on investment, peak capital expenditure, and a transmission to the valuations of semiconductor and technology stocks. Some of the data in the article is accurate, but this article is actually more of a continuation of Damir Tokic's bearish view on the AI ​​industry over the past two years. Alphabet's cash flow is no longer sufficient to cover its AI expansion. Damir Tokic's assessment of Alphabet is based primarily on three facts. First, Alphabet's capital expenditures are growing rapidly. In the first quarter of 2026, the company's operating cash flow was $45.79 billion, while capital expenditures reached $35.67 billion, leaving only about $10.1 billion in free cash flow in the conventional sense. As investments in servers, data centers, power, and network equipment continue to increase, capital expenditures are consuming more and more operating cash flow. Secondly, Alphabet began increasing external financing. In the first quarter of 2026, the company issued approximately $31.38 billion in net debt, bringing its long-term debt to $77.5 billion. Subsequently, the company announced a large-scale equity financing round to expand its AI infrastructure and global computing power. Thirdly, Damir Tokic believes that Alphabet's reported free cash flow still overestimates the company's true cash generation capabilities. He argues that equity incentives and share buybacks used to offset equity dilution should also be deducted. According to this calculation, the free cash flow left to shareholders by Alphabet may be far lower than the figures in its standard financial statements. Therefore, he concluded that Alphabet can no longer rely entirely on internal cash flow to sustain its AI expansion and must continue to raise funds through bond issuance and stock offerings. If AI investments do not generate sufficient returns quickly, the company will eventually need to reduce capital expenditures. This logic is not without basis. Alphabet's free cash flow has indeed been squeezed, and its financing methods have changed. However, directly deriving "Alphabet will be the first to cut AI spending" from these facts still lacks crucial evidence. Increased financing does not necessarily mean that the company is facing financing constraints. At the end of the first quarter, Alphabet still held approximately $126.8 billion in cash and marketable securities, and its operating cash flow was also increasing. The company's choice to raise funds in the capital market could be due to insufficient cash flow or a desire to maintain liquidity, adjusting between the balance sheet, financing costs, and shareholder returns. Similarly, capital expenditures lowering free cash flow does not equate to unsustainable investment. What truly needs to be assessed is whether these investments will lead to future growth in cloud revenue, advertising efficiency, and AI product revenue, rather than whether the capital expenditures themselves are high. This is not the first time he has declared the AI ​​bubble bursting. To understand this article, one cannot only look at Alphabet's financial data but also at the author's past judgments. Damir Tokic has consistently believed since at least 2024 that the AI ​​and semiconductor markets are peaking. In mid-2024, he judged that semiconductor ETFs were nearing their peak and that the generative AI bubble was bursting. In August 2024, he again believed that the S&P 500's rebound was about to end, and that Nvidia's earnings report might trigger the next round of decline, leading to a recessionary bear market. After the emergence of DeepSeek, he further argued that low-cost models proved AI training no longer required such massive capital investment, and other companies would shift to cheaper models and lower-cost inference solutions, resulting in a significant decrease in AI capital expenditure. By 2025, his year-end target for the S&P 500 was 4100 points, predicting a potential market decline of about 30%. However, by the end of 2025, the S&P 500 closed at approximately 6901 points, setting a new all-time high. This represents a significant margin of error. He also maintained a long-term pessimistic view of Alphabet. Previously, he published an article titled "Sell Alphabet Before ChatGPT Replaces Google," arguing that generative AI would disrupt Google Search's business model and giving Alphabet a significantly lower valuation than its market price. Entering 2026, his views have been further reinforced. Since June, he has published a series of articles arguing that the decline in the price of large models indicates that AI capital expenditure has peaked, Meta's sale of surplus computing power indicates an oversupply in the industry, companies' shift towards open source and small models means that high-priced cutting-edge models have lost their commercial basis, and the operating cash flow of technology companies is being overestimated by equity incentive accounting. Alphabet's potential reduction in capital expenditure is just the latest link in this "AI bubble bursting" narrative. Risk observation is valuable, but the credibility of inflection point judgments is low. Damir Tokic's problem wasn't that he completely misjudged the risks. Whether the return on AI investment can cover the cost of capital is indeed one of the most important issues for the tech industry in the coming years. Declining prices for large models, changes in computing power leasing prices, increased depreciation, debt financing, and equity dilution can all affect the profit distribution of the AI ​​industry chain. The higher the current capital expenditure, the higher the future revenue and return requirements of companies will be. However, he has long exhibited a clear tendency: equating "accumulating risks" directly with "the bubble has burst." These two are not the same thing. An industry may have a valuation bubble, but capital expenditures will continue to grow; model prices may fall, but usage will grow even faster; the cost per inference may decrease, but this will stimulate more applications and computing power demand. Cloud vendors selling computing power may be an expansion of their business model, and does not necessarily mean that computing power is severely oversupplied. The cash flow, financing, and ROI issues raised by Damir Tokic are worth noting, but these issues are insufficient to prove that the AI ​​bubble has burst. Considering his long, repeated, and often prematurely bearish record, the credibility of his latest judgment still needs to be significantly discounted. ### Related Stocks - [GOGL.US](https://longbridge.com/en/quote/GOGL.US.md) - [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md) - [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md) - [.SPX.US](https://longbridge.com/en/quote/.SPX.US.md) - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [NVD.DE](https://longbridge.com/en/quote/NVD.DE.md) ## Related News & Research - [Ralph Schackart Reiterates Buy on Alphabet as Gemini-Powered AI Strategy Drives Record Search Usage and Broad Enterprise Adoption](https://longbridge.com/en/news/293572599.md) - [AI leaders sign statement asking the government to do something about automated AI](https://longbridge.com/en/news/294105828.md) - [Quanta Services surges on blowout earnings for AI power infrastructure giant](https://longbridge.com/en/news/294368119.md) - [The AI Hangover Has Arrived](https://longbridge.com/en/news/294191198.md) - [Wavestone joins forces with AI Builders in deal to expand AI and data consulting](https://longbridge.com/en/news/294238044.md)