AI capital expenditure reaches 3.2% of GDP in 2027, surpassing the U.S. defense budget for the first time

By 2027, AI capital expenditures by Alphabet, Amazon, Meta, Microsoft, and Oracle will account for 3.2% of U.S. GDP, surpassing the defense budget's 2.7%. From Taiwan's semiconductor perspective, this computing power gamble is reshaping the global economic structure. Is Taiwan ready?
(Previous context: Without AI, U.S. GDP growth would be only 0.66%)
(Background supplement: Kobeissi Letter: Is AI Doomsday the Biggest Short in History? Cognitive Costs Collapse, the Era of Abundance GDP Begins)

Table of Contents

Toggle

• From an Arms Race to a Computing Power Race: A Historical Shift in Capital Flow
• How Significant Is 3.2% of GDP? Viewing This Computing Power Gamble from Taiwan's Perspective
• Cross-Country Comparison: The U.S. Leads the Way, Can Asia Catch Up?
• The Future of AI Investment: Bubble or New Normal?

The U.S. AI investment boom is reshaping the economic landscape at an unprecedented pace. According to the latest forecast data from The Kobeissi Letter, by 2027, capital expenditures on AI by the five tech giants—Alphabet (Google's parent company), Amazon, Meta, Microsoft, and Oracle—will rise to about 3.2% of U.S. Gross Domestic Product (GDP). If this forecast comes true, it will mark the first time in U.S. history that annual AI capital expenditures surpass national defense spending, which is expected to account for only about 2.7% of GDP in 2027.

This figure alone is shocking enough, but the growth curve is even more astonishing. The Kobeissi Letter points out that AI capital expenditures will account for about 1.5% of GDP in 2025, leap to about 2.5% by 2026, approaching that year's defense budget of 2.7%. By 2027, the total AI capital expenditures of the five giants are expected to exceed $1.1 trillion (already projected to exceed $800 billion in 2026). This is not just a shift in corporate investment behavior; it is a deep restructuring of the national economic structure.

From an Arms Race to a Computing Power Race: A Historical Shift in Capital Flow

For a long time, the U.S. federal defense budget has been one of the largest items in federal spending and a synonym for "national priority." Now, private sector investment in AI infrastructure is about to cross this symbolic defense red line. This means that market forces are spontaneously allocating resources to areas perceived as more promising than military defense—artificial intelligence and computing power infrastructure.

This trend is driven by profound structural forces. Since the explosion of generative AI at the end of 2022, major cloud service providers and social media giants have launched data center expansion plans known as "the largest infrastructure race in history." Microsoft's collaboration with OpenAI, the deployment of Google's Gemini model, Meta's open-source AI strategy, and Oracle's cloud expansion—each is making capital investments on a scale of hundreds of billions of dollars. These expenditures cover not only hardware procurement like Nvidia GPUs but also data center construction, energy infrastructure, and the hefty salaries of AI R&D talent.

How Significant Is 3.2% of GDP? Viewing This Computing Power Gamble from Taiwan's Perspective

For Taiwanese readers, the figure of 3.2% has a more direct reference point. Taiwan's defense budget in 2024 accounted for about 2.5% of GDP, while the overall technology industry (including semiconductors) accounted for about 15% of GDP. The AI capital expenditure alone from the five U.S. tech giants is equivalent to 1.28 times Taiwan's annual defense budget (calculated as a share of GDP). In other words, the AI investment intensity of just these five companies exceeds the entire defense budget of a medium-sized country.

More importantly, this wave of AI investment is closely tied to Taiwan's semiconductor industry. TSMC, as the core foundry for custom AI chips from Nvidia, AMD, Google, and even Microsoft, is one of the biggest beneficiaries of this torrent of capital expenditure. When U.S. giants spend $1.1 trillion building AI infrastructure, a significant portion of those chip orders flows to Taiwan. However, this also raises an unavoidable question: Is Taiwan keeping pace with the world in terms of AI investment itself?

Compared to the trillion-dollar level of spontaneous AI investment by U.S. private companies, Taiwan's AI budgets—both at the government level and in private enterprise computing power investments—appear relatively cautious in scale. Although TSMC's capital expenditure on advanced processes is also staggering (expected to exceed $32 billion in 2025), Taiwan as a whole lacks a large-scale AI application ecosystem like that of the five U.S. giants. This means Taiwan may hold a key position in the "hardware manufacturing" link of AI infrastructure but lag far behind in investments at the "computing power deployment" and "application scenario" levels.

Cross-Country Comparison: The U.S. Leads the Way, Can Asia Catch Up?

From a global perspective, the U.S. leads in AI capital expenditure almost unchallenged. The Kobeissi Letter data covers only five U.S. listed companies, excluding AI-related investments from Apple, Tesla, and other startups. If the scope is expanded to the entire U.S. tech industry, the share of AI capital expenditure in GDP could be even higher.

In comparison, China's BAT (Baidu, Alibaba, Tencent) and ByteDance are also investing heavily in AI infrastructure, but due to differences in total GDP, their AI capital expenditure as a share of GDP is much lower than that of the U.S. Japan and South Korea have made some arrangements in semiconductor manufacturing equipment but lag behind the U.S. expansion rate in hyperscale data center construction. This creates a noteworthy pattern: The U.S. is redefining national competitiveness in the AI era through capital expenditure, while other economies risk falling behind.

Returning to Taiwan's situation, the implications of this trend are twofold. On one hand, Taiwan's semiconductor supply chain plays an irreplaceable role in global AI infrastructure, meaning related companies will continue to benefit. On the other hand, if the core value of AI comes from the multiplicative effect of "computing power × data × application scenarios," then Taiwan might fall into a structural imbalance of "strong hardware, weak software" if it remains only in chip manufacturing without doubling down on AI applications and computing power deployment in the long run.

The Future of AI Investment: Bubble or New Normal?

Of course, there are cautious voices in the market regarding this unprecedented wave of capital expenditure. Citrini Research once issued a report warning that overinvestment in AI infrastructure could trigger a "global intelligence crisis" by 2028, leading to excess computing power and a sharp drop in returns. Goldman Sachs also pointed out in a report that the super profits from AI chips are distorting Taiwan's and South Korea's current account structures, potentially forcing central banks to raise interest rates.

However, the Kobeissi Letter takes a more optimistic view. The organization has previously proposed the "cognitive cost collapse" theory, arguing that AI is going through a process similar to the post-dot-com bubble period when internet infrastructure truly realized its productivity. As computing costs continue to decline and AI applications permeate all industries, the current capital expenditure will prove to be a reasonable long-term layout.

Whether bubble or new normal, one indisputable fact is that AI capital expenditure surpassing the defense budget symbolizes a fundamental shift in the logic of resource allocation in human society. In the Cold War era, national security was the highest priority for resource allocation. In the AI era, computing power and intelligence, driven by private enterprises, are becoming the new core of "national security." For Taiwan, finding its position at this turning point will be the most important strategic issue for the next decade.