Training Gemini 4 is too expensive! Barclays estimates that Alphabet Inc. Class C (GOOGL.US) will have negative free cash flow for two consecutive years.

Google's parent company Alphabet delivered a quarterly report that exceeded expectations, but the result is not relaxation, but a larger bill. According to Wind Trading Terminal, Barclays stated in a research report on July 22 that after Google announced its financial report for the second quarter of 2026, Barclays significantly raised Google's capital expenditure expectations to $350 billion for 2027, further increasing to $500 billion for 2028, both exceeding the operating cash flow by over 100% during the same period. This means that Google will continue to be in a negative free cash flow state in the next two years. The core driver of this surge in capital expenditure is that Google is accumulating computing power reserves for the large-scale training operations of the Gemini 4 models. Barclays believes that Google is currently leasing external GPU computing power at a premium higher than the market rate, providing "transitional" computing power support for some GCP clients, while releasing its own computing power for the training of Gemini 4. This arrangement costs approximately $1 billion per month and will put pressure on the cloud business profit margin in the fourth quarter and the first quarter of 2027. According to Wall Street News, the latest financial report shows that Google's Q2 cloud business revenue increased by 82% year-on-year, and the cloud business backlog orders increased by $50 billion to $514 billion in a single quarter, while search advertising revenue increased by 17% and YouTube advertising revenue accelerated to 13%. However, during the financial report conference call, the double inflation of operating costs and capital expenditures brought about by the comprehensive penetration of AI is becoming a core variable of concern to the market. CFO Ashkenazi expects that AI infrastructure investment will continue to put pressure on the profit and cash flow statements. Cloud business accelerating, but cost pressure rising synchronously Barclays attributed the highlight of this quarter to the continuous acceleration of Google Cloud. In the second quarter of 2026, Google Cloud revenue reached $24.768 billion, an 82% year-on-year increase (Barclays estimates about 77% after excluding TPU sales), and the cloud business operating profit margin increased to 36%. The cloud business backlog orders surged from $46.4 billion in the previous quarter to $514 billion, with a net increase of $50 billion in a single quarter, indicating a strong demand momentum from corporate clients for GCP. However, the rapid growth of the cloud business is accompanied by significant cost increases. The report pointed out that Google has signed a GPU leasing agreement with Colossus for approximately $1 billion per month to provide computing power to some GCP customers in a "transitional" manner, expected to last for about two quarters. Barclays believes that the essence of this arrangement is to release Google's first-party computing power for the large-scale training operations of Gemini 4. Barclays explicitly stated in the report that this significant upward adjustment of capital expenditure expectations is directly linked to the training demand of the Gemini 4 models. The report shows that Barclays has raised the capital expenditure expectation for 2027 from the previous $256.1 billion to $350.2 billion, an increase of 37%; the capital expenditure expectation for 2028 has also been significantly raised from $327.4 billion to $500.3 billion, an increase of approximately 53%. According to Barclays' predictive model, Google's capital expenditure from 2026 to 2028 will reach approximately $200.5 billion, $350.2 billion, and $500.3 billion, respectively. For two consecutive years, it will exceed the operating cash flow of the same period, with free cash flow turning into negative $6.5 billion and negative $13.52 billion. The report also pointed out that Google's management expressed clear confidence in the development pace of Gemini 4 and AI models during the financial report conference call. Currently, the call volume of the Gemini API has reached 22 billion tokens per minute, significantly higher than the 16 billion from a quarter ago, a year-on-year increase of 3 to 4 times. Barclays estimates that its market position is currently tied with Anthropic for second place. AI full-stack layout builds competitive barriers, but multi-line operations cost dearly Barclays qualitatively described Google as "the company with the highest level of AI advancement in the world currently", with its AI layout spanning the application layer (search, Gemini, etc.), the model layer (Gemini, Flash, etc.), the infrastructure layer (GCP), and the raw computing power layer (TPU), forming a complete full-stack coverage. Barclays believes that this breadth will become more prominent in the following quarters with the continuous acceleration of the cloud business. However, the cost of this all-encompassing competition is also significant. The Barclays report pointed out that Google faces competition from Anthropic and OpenAI at the model layer, as well as pressure from AWS and Azure at the infrastructure level, while also having to maintain the normal operation of 13 applications with over 1 billion monthly active users. This pattern of "multi-line operations" is the fundamental reason for the simultaneous inflation of operating expenses and capital expenditures. In terms of profit forecasts, Barclays raised revenue and operating profit expectations for 2027 by 5%, mainly reflecting the upward revision of Google Cloud revenue and profit. However, due to the impact of the Colossus agreement, earnings per share expectations for the fourth quarter of 2026 and the first quarter of 2027 were slightly reduced, with subsequent EPS expectations adjusted accordingly. Barclays raised Google's target price from $405 to $425 and maintained a "hold" rating based on an average of 25 times the estimated EPS for 2027 to 2028 and a multiple of 15 times EBITDA. Looking from a longer-term perspective, Barclays believes that Alphabet currently has few comparable competitors in capturing the full-stack capabilities of the AI value chain and that the strategic value of its cross-layer layout will gradually be released as the demand for AI continues to expand. This article is reprinted from "Wall Street News"; GMTEight editor: Chen Siyu