Big Four's AI Capital Spending Soars to Record $725 Billion

$725 Billion: Tech Giants' AI Arms Race Sets Another Record

AI investment by the world's four largest tech giants is climbing at a staggering pace. According to the latest earnings reports and planning data, Google, Amazon, Microsoft, and Meta plan to invest a combined $725 billion in capital expenditure in 2026 — a 77% surge from last year's already record-breaking $410 billion. Even more noteworthy is the hidden force behind this astronomical figure: the sharp rise in component prices has become an impossible-to-ignore driver.

Microsoft: $25 Billion Swallowed by Price Hikes

Amid this capital spending frenzy, Microsoft's disclosure stands out. The company explicitly stated that $25 billion of its AI budget is directly attributable to rising memory and chip costs. In other words, this massive sum has not translated into additional computing power or infrastructure expansion — it has been consumed by supply chain price increases.

This figure reflects the harsh reality of today's AI hardware market. As global demand for high-bandwidth memory (HBM), advanced GPUs, and custom AI chips grows exponentially, supply-side capacity expansion has fallen far behind the pace of demand growth. The resulting supply-demand imbalance has directly driven up prices for critical components.

The Big Four's Distinct Spending Strategies

While rising prices represent a shared challenge, each company has its own strategic rationale for ramping up investment:

• Microsoft: As OpenAI's largest investor and the provider of Azure cloud services, Microsoft must continuously expand data centers to support the training and inference demands of GPT-series models while deeply integrating AI capabilities into its Copilot product line.

• Google: Fighting on three fronts — Gemini model iteration, in-house TPU chip development, and cloud computing infrastructure — Google faces relentless capital expenditure pressure. The company is attempting to partially hedge against external supply chain price hikes through its custom chip strategy.

• Amazon: As the world's largest cloud computing platform, AWS must not only meet customers' rapidly growing AI workload demands but is also aggressively pushing mass deployment of its in-house Trainium chips to reduce dependence on NVIDIA.

• Meta: To support the continued training of its Llama open-source models and upgrades to its AI-powered ad recommendation systems, Meta is undertaking a massive expansion of its AI infrastructure — what Mark Zuckerberg has called "the most important technology investment of this generation."

Behind the Price Surge: Structural Supply Chain Tensions

The spike in component prices is no accident but rather the result of multiple compounding factors:

First, HBM is in severe short supply. High-bandwidth memory is a core component of AI accelerators, and global production capacity is currently concentrated in the hands of just three companies: SK Hynix, Samsung, and Micron. As NVIDIA's H200 and B200 series GPUs ship in large volumes, HBM demand far outstrips supply, driving prices ever higher.

Second, advanced process node capacity is constrained. TSMC's 3nm and soon-to-be mass-produced 2nm process capacity is being fought over by major customers including Apple, NVIDIA, and AMD. Foundry fees have risen accordingly, directly pushing up chip costs.

Third, geopolitical risk premiums persist. U.S. chip export controls on China, along with various countries' push to localize semiconductor supply chains, have further increased uncertainty and costs across the global chip supply chain.

Concerns: A 'Leap of Faith' on Investment Returns

The $725 billion capital expenditure figure already exceeds the GDP of many countries. Wall Street is divided — optimists believe AI will reshape the entire economic system and that current investments will ultimately yield outsized returns; skeptics worry that if AI application monetization falls short of expectations, such massive capital spending will severely drag down corporate profits.

Even more concerning is the fact that when a $25 billion budget increase merely covers price hikes rather than capacity expansion, the "real value" of capital expenditure is declining. The actual computing power gains these tech giants are achieving may be far less impressive than the headline numbers suggest.

Outlook: The Arms Race Is Far From Peaking

Based on current trends, 2027 capital expenditure figures could be even more staggering. On one hand, next-generation AI models continue to demand exponentially more computing power; on the other, component supply chain capacity expansion requires two to three years of construction, meaning short-term pricing pressure is unlikely to ease.

For the entire AI industry, this capital spending surge led by the Big Four represents both a firm bet on the future of AI and an endurance test of patience and resolve. Whoever can most efficiently convert capital into technological advantage and commercial returns in this spending race will ultimately determine the industry landscape of the AI era.