Big Tech's "AI Money-Burning War": The current scale is underestimated, future depreciation is underestimated, and the earliest price war may break out in 2027.
```
U.S. tech giants are engaged in an unprecedented AI infrastructure arms race, with their capital expenditure intensity approaching the peak levels seen during the dot-com bubble.
According to news from Trading Desk, new research by Bank of America and Morgan Stanley shows the market has severely underestimated the true scale of current AI investments, while also being insufficiently prepared for the impact of future depreciation costs. Supply-demand imbalances could trigger a cloud services price war as early as 2027.
Morgan Stanley’s research indicates that "hyperscale" players—including Amazon, Google, Meta, Microsoft, and Oracle—are expected to have capital expenditures reaching 26% of sales revenue by 2027, close to the 32% peak during the dot-com bubble and exceeding the 20% seen during the shale oil boom. More critically, these public capital expenditure figures do not fully reflect the scale of investment, as off-balance-sheet instruments like finance leases are increasingly used to accelerate data center expansion, resulting in an underestimation of the real scale of current investment.
Bank of America’s analysis focuses on the long-term impact of these investments. Their research shows that the market generally underestimates future depreciation expenses. By 2027, for Google, Amazon, and Meta alone, market forecasts for depreciation expenses could be understated by nearly $1.64 billion. The bank further notes that if supply continues to outpace demand, the industry could see even more aggressive pricing strategies as early as 2027.
Capital Expenditure Race: A "Arms Race" with Underestimated Scale
Morgan Stanley's report compares the current wave of AI investment to two historical periods of capital frenzy: the telecom sector's fiber optic construction during the dot-com bubble, and the energy sector's drilling during the shale oil revolution. The report points out that the current capital intensity is approaching the former’s peak. Uniquely, tech giants are accelerating expansion through increasingly complex financial means, making traditional capital expenditure (Capex) data insufficient to fully capture their investments.
Morgan Stanley highlights two key factors that have led to the underestimation of actual investment scale:
First, the rise of finance leasing. Companies like Microsoft and Oracle are increasingly using finance leases to build data centers. Economically, this is similar to borrowing to purchase assets, but the initial investment often doesn’t appear in traditional capex, thereby bypassing the cash flow statement. The report finds that after factoring in finance leases, the capital intensity of Microsoft and Oracle rises significantly. For example, Morgan Stanley estimates that Microsoft’s ratio of capital expenditure to sales will jump from 28% to 38% in fiscal 2026, and Oracle’s from 41% to 58%. Additionally, the amount committed to leases not yet started by these giants has already exceeded $335 billion, indicating this trend will continue.
Second, the delayed effect of "Construction in Progress": Huge investments are accumulating on balance sheets as "Construction in Progress (CIP)". These assets do not start depreciating until officially put into use, so their costs have yet to impact the profit statement. According to Morgan Stanley data, the CIP balances for Google, Amazon, Meta, and Oracle have all surged over the past year; for example, Amazon's increased by about 60% ($17 billion) and Google's by about 40% ($15 billion). This means a massive amount of capital has already been spent, but the impact on profits is only beginning.
The “Time Bomb” in Financial Reports: Wall Street Underestimates Future Depreciation Costs
If Morgan Stanley uncovers the “under the iceberg” level of investment, Bank of America points out how these investments will translate into real cost pressures in the future. The core argument: Wall Street is “slow to react” to the growth of future depreciation costs.
BofA analyst Justin Post notes in the report that as the combined capital expenditures of Google, Meta, and Amazon grow by 56% and 63% in 2024 and 2025, their depreciation and amortization (D&A) expenses will inevitably accelerate from 2026 onward. By 2027, BofA’s depreciation forecasts for the three giants differ significantly from market consensus:
Alphabet (Google): a gap of about $700 millionAmazon: a gap of about $590 millionMeta: a gap of about $350 million
In total, a "forecast gap" of nearly $1.64 billion means these companies' future actual earnings could be much lower than current market consensus.
The report also pointed out another factor worsening depreciation risk: the “short lifespan” issue of AI assets.
Unlike traditional servers, hardware such as GPUs for AI computation face faster technological iterations and higher workloads, potentially cutting their useful lives to just three to five years.
BofA points out that in the first quarter of 2025, Amazon has already shortened the expected usage life for some servers and network equipment from six years to five, citing accelerated technology development in AI and machine learning. This is the opposite of the trend in previous years, in which tech giants were generally extending equipment lives to smooth out costs. If this trend reverses, depreciation expenses will be recognized more quickly, impacting short-term profitability.
Risks and Returns: Price War Could Erupt as Early as 2027
BofA warns that the AI infrastructure market could repeat the pattern of aggressive investment leading to overcapacity and price pressure. As tech giants accelerate AI infrastructure investment, there is a risk of overbuilding—i.e., computational supply exceeding demand for high-value AI services.
Also, as large language models converge in performance, product differentiation may weaken, causing infrastructure services to become commoditized. Meta is building several gigawatt-scale data centers slated to come online between 2026–2029; Oracle and OpenAI’s proposed $500 billion Stargate project is expected to add huge AI capacity from 2028–2029. If demand cannot keep up with this scale of supply deployment, hyperscale players may resort to aggressive pricing to maintain utilization, thereby compressing margins.
BofA believes that if supply exceeds consumption (which it sees as likely only from 2027 at the earliest), hyperscale players may adopt more aggressive pricing strategies to maintain utilization, which will erode profitability.
~~~~~~~~~~~~~~~~~~~~~~~~
The above content is from Trading Desk.
For more detailed interpretation, including real-time insights and frontline research, please join [Trading Desk ▪ Annual Membership]
Risk Warning and DisclaimerThe market has risks; investment needs to be cautious. This article does not constitute personal investment advice and does not take into account individual users’ special investment goals, financial status, or needs. Users should consider whether any opinions, views, or conclusions in this article fit their specific circumstances. Invest accordingly at your own risk. ```