By 2027, capital expenditures on artificial intelligence by five US tech giants—Alphabet, Amazon, Meta, Microsoft, and Oracle—are expected to reach 3.2% of US GDP, surpassing the projected defense budget of 2.7%. According to data from The Kobeissi Letter, AI capex stood at 1.5% of GDP in 2025, jumped to 2.5% in 2026, and is forecast to exceed $1.1 trillion in 2027. This marks a historic shift where private-sector spending on AI infrastructure overtakes military spending, redefining national priorities through market forces.
From Arms Race to Compute Race: A Capital Flow Milestone
The US federal defense budget has long symbolized national priority. Now, Big Tech's AI investments cross that symbolic line. Since the generative AI explosion in late 2022, cloud giants have launched what many call the largest infrastructure race in history. Microsoft partners with OpenAI, Google deploys Gemini, Meta pushes open-source AI, and Oracle expands its cloud. Spending covers Nvidia GPUs, data centers, energy infrastructure, and AI talent—each company committing tens of billions annually.
Taiwan's Perspective: Chip Benefits vs. Application Gap
For Taiwan, 3.2% is a stark reference point. Taiwan's 2024 defense budget is about 2.5% of GDP, and its entire tech sector contributes about 15%. The five US giants' AI capex alone equals 1.28 times Taiwan's defense spending (as a share of GDP). TSMC, the core foundry for Nvidia, AMD, Google, and Microsoft's custom AI chips, is a prime beneficiary. Yet Taiwan's own AI application and computing deployment investments remain modest. Lacking a large-scale AI ecosystem like the US, Taiwan risks a structural imbalance—strong in hardware manufacturing but weak in software and deployment.
Global Comparison: US Dominance, Asia Struggles to Catch Up
The Kobeissi Letter data covers only five publicly traded US companies. Including Apple, Tesla, and startups would push the share even higher. In China, BAT and ByteDance invest heavily, but as a percentage of GDP, their AI capex is far lower due to the larger GDP base. Japan and South Korea focus on semiconductor equipment but lag in hyperscale data center builds.The US is redefining national competitiveness through capital expenditure in AI, leaving other economies at risk of falling behind.
Bubble or New Normal? Divergent Views
The unprecedented capex surge draws mixed opinions. Citrini Research warns that overinvestment could trigger a "global intelligence crisis" by 2028, leading to compute oversupply and collapsing returns. Goldman Sachs notes that AI chip super-profits distort Taiwan's and South Korea's current accounts, possibly forcing central banks to raise rates. In contrast, The Kobeissi Letter is optimistic, drawing parallels to the post-dot-com era when internet infrastructure eventually became productive. Falling compute costs and pervasive AI applications could validate today's massive outlays.
AI capex overtaking defense spending signals a fundamental shift in how society allocates resources. During the Cold War, national security commanded top priority. In the AI era, computing power and intelligence—driven by private enterprise—are becoming the new core of "national security." How Taiwan positions itself at this inflection point will define its strategic trajectory for the coming decade.