AI Capital Spending by Google, Microsoft, Amazon, and Meta Set to Surpass $7.25 Trillion in FY2026

The AI landscape this week is characterized by fast‑moving technology, escalating capital commitments, and evolving regulatory frameworks.

Regulatory guidance: China’s Central Cyberspace Administration announced the release of the "Artificial Intelligence Application Ethics and Safety Guidelines (Version 1.0)" during the 2026 China Cyber Civilization Conference.

Capital spending surge: Google, Microsoft, Amazon, and Meta disclosed in their latest earnings calls that combined AI‑related capital expenditures for FY2026 are expected to exceed $7.25 trillion, a growth of more than 77% year‑over‑year (source: Ti Media / Caixin).

US policy move: The White House revealed a draft executive order aimed at providing a clear roadmap for AI development.

AI‑driven media: Hunan Economic TV launched AI anchors “Shengsheng” and “Shuangshuang,” using an "AI + human" collaboration model with all broadcasts labeled as AI‑generated.

Defense collaboration: The US Department of Defense signed agreements with eight AI vendors—including OpenAI, Google, Nvidia, Microsoft, AWS, SpaceX, Reflection AI, and Oracle—to deploy cutting‑edge AI technologies on classified military networks.

Model upgrade: OpenAI announced that the default ChatGPT model will be upgraded to GPT‑5.5 Instant for all users, reducing hallucinations by 52.5% in high‑risk domains and cutting factual‑error statements by 37.3%.

Enterprise AI agents: Anthropic released a new AI agent tailored for financial services, aiming to provide more professional automation solutions.

Safety reviews: Google, xAI, and Microsoft agreed to submit their new AI models to US government safety reviews as the White House considers forming an AI working group.

Research breakthrough: A study published in Science showed OpenAI’s o1‑preview model achieved a 67.1% accuracy rate in emergency case triage, outperforming two expert physicians (55.3% and 50.0%).

SpaceX‑Anthropic partnership: SpaceX will grant Anthropic access to its Colossus 1 supercomputer, adding over 300 MW of compute capacity—equivalent to 220 000 Nvidia GPUs—within a month, with plans for gigawatt‑scale orbital AI compute.

Medical AI approval: China’s National Medical Products Administration approved a chest‑CT AI‑assisted diagnostic software for fast‑track review, marking the first large‑model medical product in the special review channel.

Funding rounds: Chinese AI startup Kimi completed a $2 billion financing round, reaching a post‑money valuation of $20 billion; DeepSeek secured a $450 billion valuation in a new round led by the China Integrated Circuit Industry Investment Fund; and other AI firms such as Kunlun Chip and others announced IPO preparations.

EU regulatory update: European governments agreed to postpone the enforcement date for high‑risk AI system rules to December 2 2027 and to ban AI‑generated non‑consensual pornographic images effective December 2 2026.

Academic publication: Researchers from Tsinghua University and New Bird Trust published "Hyper‑RAG: Combating LLM Hallucinations using Hypergraph‑Driven Retrieval‑Augmented Generation" in Nature Communications , proposing a hypergraph‑based framework to improve factuality in complex knowledge‑retrieval scenarios.

Overall, the week underscores a intensifying global AI race, with massive fiscal commitments, tighter governance, and rapid advancements in model capabilities and applications across industry, defense, healthcare, and research.