U.S. tech giants are unafraid of the impact from DeepSeek, with Google, Microsoft, and Meta all increasing their investments, while there are some discrepancies within Morgan Stanley regarding cloud spending forecasts.In the capital expenditure battle in the AI field, the three giants—Google, Microsoft, and Meta—are burning cash at an astonishing rate. Despite facing challenges from DeepSeek, these tech giants firmly believe that large-scale investments will lay the foundation for future AI competition.Media reports indicate that the intensity of capital in this wave of AI investment surpasses that of the last oil boom. In 2024, the average capital expenditure of Microsoft, Meta, Amazon, and Alphabet is expected to account for 17.2% of their revenues, up from 12.7% in 2021, with this figure projected to rise further to 22% by 2025.Previously, there were some discrepancies within Morgan Stanley. The Greater China team at Morgan Stanley believes that the growth cycle of the cloud computing market may continue until the first half of 2025, with an expected year-on-year growth rate of 11% for the capital expenditure of the top ten cloud vendors before 2025. However, the North American hardware research team at Morgan Stanley expects a year-on-year growth rate of 24% for the capital expenditure of the top ten cloud vendors before 2025 (up from a previous estimate of 21%).The analysis from the Information Equality public account suggests that the direction of computing power investment in the AI industry itself has not changed. The emergence of DeepSeek r1 has only caused some structural changes in the narrative of training, but has significantly strengthened the reasoning narrative in terms of scale. The reduction in model costs has instead improved the ROI of AI investments, making capital expenditures more solid.The Three Giants Continue to Increase Capital ExpenditureGoogle previously announced a significant increase in its investment scale. Sundar Pichai, CEO of Google and its parent company Alphabet, announced that the company plans to invest $75 billion in capital expenditures in 2025, a substantial increase from last year's $52.5 billion expenditure. This figure is equivalent to 19% of Wall Street's expected total revenue for Google this year, far exceeding the average of less than 13% over the past decade.Although Google started late in AI investment, it is quickly catching up. Sundar Pichai emphasized that Google is focused on reducing the cost per query and has achieved significant efficiency improvements in AI development. Pichai stated regarding AI opportunities, "It is as big as it gets, which is why you see us investing to meet this moment."Microsoft is the king of capital expenditure, consistently maintaining a leading position in AI investment. According to Visible Alpha's forecast, Microsoft's capital expenditure in 2025 is expected to exceed $90 billion, accounting for more than 30% of its revenue.Meta has also significantly increased its investment plans, recently announcing a more than 60% increase in capital expenditure for 2025, reaching $65 billion, which also accounts for more than 30% of its revenue. Meta CEO Mark Zuckerberg stated last week during the earnings call that substantial capital expenditure will be a strategic advantage for capable companies.Additionally, Google has an advantage in capital expenditure, as its profitable search advertising business remains strong, with the operating profit margin of its services segment reaching 39% in the most recent quarter. Its net cash balance is also the highest among peers, reaching $99 billion, while Microsoft and Meta have net cash balances of $42 billion and $49 billion, respectivelyIn the competition for AI spending, the companies with the deepest pockets are not sitting idle.Capital Intensity Surpassing the Oil BoomSome media outlets have summarized that the capital intensity required for technology companies to develop artificial intelligence projects is even higher than the peak of the supercycle investment by oil and gas companies in the mid-2010s.Analysts expect that in 2024, the average capital expenditure of Microsoft, Meta, Amazon, and Alphabet will account for 17.2% of their revenue, up from 12.7% in 2021, and this figure is expected to rise further to 22% by 2025.In contrast, from 2012 to 2016, BP, Chevron, ExxonMobil, and Shell spent an average of 10.2% of their revenue on capital expenditures each year. During this period, after several years of rising oil prices, large energy companies invested heavily in capital expenditures to increase production.Discrepancies in Morgan Stanley's Internal Cloud Spending ForecastsPreviously, Morgan Stanley's Greater China semiconductor research team released a report stating that they have significantly lowered the expected shipment volume of NVIDIA's GB200 in 2025 from the previous estimate of 30,000-35,000 units to 20,000-25,000 units, with the worst-case scenario potentially falling below 20,000 units. This downgrade could lead to a market impact of $30 billion to $35 billion on the GB200 supply chain, putting immense pressure on related supply chains and semiconductor companies.Additionally, the growth rate of capital expenditures in the cloud computing industry is expected to slow to single digits in the fourth quarter of 2025, which will also negatively impact the shipment volume of the GB200.In the aforementioned research report, Morgan Stanley's Greater China semiconductor research team expects the capital expenditure growth rate of the top ten cloud vendors before 2025 to be 11% year-on-year (previously 8%).However, there are certain discrepancies within Morgan Stanley. In the latest report from Morgan Stanley's North American hardware research team, the expected capital expenditure growth rate of the top ten cloud vendors before 2025 is 24% year-on-year (previously 21%).In response, the information equity public account analyzed that the direction of computing power investment in the AI industry itself has not changed. Moreover, the emergence of DeepSeek r1 has only caused some structural changes in the narrative of training, but has significantly strengthened the reasoning narrative in terms of scale. CSPs in North America and China are genuinely benefiting from this open-source initiative. "Google emphasized in its earnings call that the reduction in model costs has made their application deployment more cost-effective, and the computing power usage more related to Google's revenue is reasoning rather than training. The reduction in model costs has actually improved the ROI of their AI investments, making capital expenditures more solid."