In 2026, the AI arms race among tech giants is accelerating: the combined AI-related capital expenditures announced by Amazon, Google, Microsoft, and Meta for 2026 have surpassed $630 billion—a figure equivalent to Sweden's entire annual GDP.

This AI infrastructure competition, described by The Wall Street Journal as "more expensive than the moon landing," has triggered significant turbulence in capital markets. Tech giants' AI arms race is under pressure to transition from burning money to generating profits.

Giants Invest Tens of Billions in Core AI Sectors

The AI investments of tech giants have reached the tens of billions of dollars, focusing on core sectors such as AI infrastructure, chip R&D, and technology implementation, showcasing their respective AI strategies and ambitions.

In 2026, Amazon leads the AI race with a planned expenditure of $200 billion, a 50% increase from its actual investment of $125 billion in 2025, marking a new record for the company's single-year capital expenditure.

This massive investment is characterized by "aggressive expansion": the majority is allocated to AWS computing power expansion, large-scale deployment of Trainium2 in-house chips, over 100,000 new AI robots to reduce logistics order processing time by 30%, and the advancement of Project Kuiper satellite network construction.

According to the latest financial report, Amazon's annual AWS revenue reached $127.1 billion, a 20% year-over-year increase; operating cash flow in Q4 2025 grew by 20% to $139.5 billion. These robust financial figures underpin its multi-billion-dollar investment.

Google parent company Alphabet has set a spending range of $175-185 billion, nearly double its 2025 capital expenditure. Funds are primarily directed toward data center expansion, NVIDIA chip procurement, and the full-scale advancement of Gemini model iteration and in-house TPU chip R&D, aiming to maintain its core search business advantages through vertical integration of chips and large models.

Microsoft plans to invest over $120 billion, with its latest quarterly capital expenditure in FY2026 reaching $37.5 billion, a 66% year-over-year increase. Funds are mainly allocated to AI computing power construction to support OpenAI's computing demands and Azure cloud service global expansion, deepening its partnership in artificial general intelligence R&D.

Meta plans to invest $115-135 billion, with the upper limit of $135 billion representing an 87% increase from 2025. Meta's spending focuses on building self-owned hyperscale computing clusters to sustain the iteration of its Llama open-source model and upgrade AI recommendation systems for its core social platforms.

According to Bloomberg, the total investment of over $630 billion from the four giants accounts for more than 75% of global AI infrastructure investment, creating a formidable financial moat around AI and intensifying the industry's Matthew effect.

This investment has evolved from a mere technological competition into a "strategic defensive battle" among the giants. As a senior Silicon Valley analyst noted, the risk of not investing in AI in 2026 far outweighs the risk of overinvesting. Falling behind could mean losing access to core markets like cloud services and search within three years.

Why Are Massive AI Investments Facing Investor Resistance?

Despite the giants' confidence in the AI sector, their multi-billion-dollar investments have faced "voting with their feet" from capital markets: after Amazon announced its $200 billion investment plan, its stock price plummeted 11% in after-hours trading; Microsoft, despite strong cloud business growth, saw its stock price drop 18% in a single week due to a 66% surge in data center spending. Google and Meta also experienced varying degrees of stock price declines.

In contrast, Apple, which maintained a cautious approach in the AI race, saw its stock price rise 7.5% after announcing robust (robust) financial results, becoming one of the few market winners during earnings season.

This divergence in stock performance essentially reflects capital markets' direct response to the uncertainty of AI investment returns. The reality of "high inputs, slow returns" in AI investments has placed the giants under dual pressure from financial constraints and uncertain returns.

The financial pressure primarily stems from large-scale infrastructure investments significantly straining the giants' cash flows. For example, Amazon's free cash flow dropped to $11.2 billion in Q4 2025, while net purchases of property and equipment increased by $50.7 billion, directly reflecting the squeeze on cash flow from AI investments. According to a JPMorgan report, the four giants' AI investments in 2026 will consume over 90% of their operating cash flows. Additionally, the accelerated hardware update cycle—shrinking from five years to three—means rapid asset depreciation for servers and data centers, hanging like a "Sword of Damocles" over the giants' financial reports.

The uncertainty of investment returns is evident in the mismatch between AI investments and revenue returns—a key factor driving "bubble" concerns.

While AWS's AI service revenue accounts for about 30% of Amazon's total, a report from the McKinsey Global Institute reveals that nearly 80% of businesses deploying AI have failed to improve net profits, and 95% of generative AI pilot projects have not yielded direct financial returns. This has left investors deeply skeptical about AI investment ROI, shifting capital markets' focus from "whether AI stories can be told" to "whether AI can sustainably improve return structures."

From Burning Money to Generating Profits: Is AI Agent the Key?

This multi-billion-dollar AI arms race also marks the official entry of tech competition into the "heavy asset era"—where the lightweight asset myth of driving industry growth through coding alone has ended, replaced by a frantic scramble for data centers, chips, and power resources. The key to breaking this deadlock lies in the commercialization and implementation of AI technologies.

2025 was dubbed the "Year of AI Agents," and 2026 has seen accelerated maturation of intelligent agents. For instance, the recent surge in popularity of OpenClaw represents breakthroughs in product form, interaction methods, and open ecosystems for intelligent agents, even boosting Mac mini sales.

The maturation and explosion of Agentic AI may become the core variable in resolving the contradiction between input and return for the giants. Unlike generative AI, Agentic AI possesses the ability to autonomously execute complex workflows. If implemented at scale, it could fundamentally alter the commercial value logic of AI, upgrading it from a "tool" to an "autonomous workforce."

The giants are also focusing on Agentic AI as a core breakthrough: Amazon aims to cut logistics costs by 30% and computing power costs by 40% through intelligent agents; Google and Microsoft are deploying Agentic AI technologies and planning to implement them across millions of enterprise clients. These goals all point to a fundamental shift from "selling computing power" to "selling efficiency."

The $630 billion investment represents a collective bet by tech giants on the AI era and a significant milestone in global AI industry development. Behind this gamble lies an industry consensus that computing power will become a core production factor in the future. However, only computing power that translates into actual commercial value constitutes true core competitiveness.

This competition also carries substantial risks. Analysts predict that if AI applications fail to achieve scaling (scalable) profitability by 2027, the tech industry may face a "deleveraging" pain similar to the aftermath of the 2000 dot-com bubble.

Clearly, for the giants, 2026 is both a year of AI infrastructure buildout and application implementation. The transition from "selling shovels" to "selling efficiency," from infrastructure deployment to commercial value realization, requires achieving a balance between input and return. Only by converting massive capital investments into tangible commercial value and operational efficiency can they ultimately prevail in this silent arms race.

#ArtificialIntelligence#IntelligentAgents#AIArmsRace