Imitating AT&T's "building network against the trend" during the Great Depression? Super-large scale operators such as Microsoft, Meta, and others are betting on the AI survival battle with an annual expenditure growth rate of 72%.

Wall Street is becoming increasingly nervous about the capital expenditures of super-sized data center operators. The total capital expenditures of the four super-sized data center operators - Amazon.com, Inc. (AMZN.US), Microsoft Corporation (MSFT.US), Alphabet (GOOGL.US), and Meta (META.US) - in 2024 were slightly over $200 billion. At the current trend, this number will almost reach $700 billion in two years. The combined free cash flow of these four companies last year dropped to $200 billion, lower than the expected $237 billion for 2024, mainly due to spending on data center construction. Investors can't help but wonder: how will these expenditures change if the economic situation reverses? History provides part of the answer from an unexpected source. In 1930, the telecom giant AT&T (T.US) was in the midst of its largest private infrastructure construction period in history, with annual construction costs reaching a staggering $585 million, which was almost unimaginable at the time. Even when faced with the worst economic depression in history, with GDP collapse and skyrocketing unemployment rates, the Bell System never stopped its construction projects, and even continued to pay a dividend of $9 per share from 1929 to 1942. This countercyclical behavior was not based on optimism about profits, but on a survival instinct rooted in a "monopoly logic": as the embodiment of network effects, if the Bell System stopped expanding, its reason for existence and moat would immediately crumble, making continued construction the only choice to maintain its system. This "stopping is more dangerous than continuing" structural logic is echoing precisely in today's AI arms race, as the behavior of the giants shifts from proactive growth to defensive investment. Amazon.com, Inc. CEO Andy Jassy's recent description of AI opportunities is essentially not a growth report, but a statement of competitive risks - in the current era of technological explosion, the risk of falling behind is far greater than the cost of over-investment. Currently, the capital expenditures of these five giants (including Oracle Corporation) have maintained a staggering average annual growth rate of 72% since mid-2023, with total spending accounting for 2.2% of the U.S. GDP. This investment is no longer about optimizing existing operations, but about occupying a central Hub Group, Inc. Class A role in the future digital landscape similar to the "telecom infrastructure" of yesteryear. However, modern super-sized computing centers face a more severe financial depreciation test compared to the Bell system of the past. The depreciation rate of AI assets is very rapid, with hardware equipment usually depreciating at a rate of 20% per year. This means that by 2025, annual depreciation expenses could devour all of these companies' profit growth. The financial condition of free cash flow is deteriorating, but spending commitments have not changed. Unlike the Bell system of the past, which had government concessions and fixed return guarantees, today's tech giants operate in a completely open, multi-party global oligopoly market. Faced with tariff fluctuations, interest rate uncertainties, and weak market sentiment, these companies may express a firm stance in their guidance, but their financial cushion has significantly dwindled. When the logic of infrastructure evolves into a survival competition, capital expenditures become the last item to be cut - not the first, and whether this high-leverage expansion is a ticket to a new era or a spark for the next financial crisis is becoming the focus of global investors.