State of AI 2025: Infrastructure, Adoption, and the Geopolitical Race between the USA and China

The new report “State of AI 2025” by Messari captures a sector growing simultaneously in three dimensions: adoption, investment, and computational capacity, with an impact that now transcends industrial, national, and political boundaries. Artificial intelligence is no longer an emerging technology but a structural force reshaping the global economy, geopolitical power, and digital infrastructures.

The result is a clear picture: AI is entering a phase of accelerated maturity, where competition no longer concerns just the models, but especially energy, hardware, data centers, and geostrategic alliances.

AI Adoption Booms Across Every Sector

According to the report, AI is now present “everywhere, simultaneously.” The collected statistics show an unprecedented cross-sector adoption:

• Education: 63% of K-12 teachers and 49% of university instructors are already using GenAI tools.
  
• Healthcare: 66% of doctors report using AI in clinical practice (compared to 38% in 2023).
  
• Finance: 17% of teams use “AI agents” and over 75% anticipate daily use by 2028.
  
• Media: over 700 outlets have signed agreements with AI companies (Guardian, FT, Washington Post, Condé Nast, etc.).
  
• Software development: 61% of developers use AI in their workflow.
  
• Defense: the US Army has signed a decade-long contract with Palantir valued at up to 10 billion dollars.
  
• Energy and infrastructure: data centers were already consuming 415 TWh in 2024, a figure expected to double by 2030.
  

It is a contemporary, rapid, and cross-sectional adoption: a pattern not even seen in the early years of the internet.

The Rise of Big AI Labs: OpenAI, Google, Microsoft, Anthropic

The report confirms a clear trend: major American labs are accelerating towards an increasingly scale-oriented model.

• ChatGPT grew from 1 million users in 2022 to 700 million weekly users in 2025.
  
• Google reports 450M monthly active users for Gemini.
  
• Microsoft Copilot has surpassed 100M MAU.
  
• Claude by Anthropic estimates over 30M users.
  

Simultaneously, paid adoption is skyrocketing: from 5% of business users in 2023 to 44% in 2025, with retention close to 80%.

It’s no longer just hype: the market is paying to use AI.

The Arms Race: USA vs China

One of the strongest chapters of the report concerns the geopolitical aspect. By 2025, the competition between the United States and China for AI leadership has become openly strategic.

United States

On July 23, the White House unveiled “Winning the Race: America’s AI Action Plan”, a plan with over 90 initiatives to:

• accelerate AI innovation and research;
  
• build domestic infrastructure (federal cloud, supercomputers, national TEEs);
  
• define global standards and security guidelines;
  
• consolidate leadership in chips and semiconductors.
  

Simultaneously, the USA maintains and extends export controls towards China, limiting access to NVIDIA’s most advanced chips.

China

Three days later, China responded with the Global AI Governance Action Plan, focused on:

• technological self-sufficiency;
  
• development of domestic hardware (Huawei Ascend series);
  
• state-led open-source models;
  
• AI applied to strategic industries;
  
• alternative multilateral cooperation to the US model.
  

The USA focuses on private frontier models and the market.
China focuses on national capacity, controlled hardware, and vertical alignment.

It’s a two-speed race but with the same destination: AI sovereignty.

The Real Battle: Compute, Energy, and Data Centers

According to Messari, the competition is not about the models, but about the infrastructure.

CapEx in “new industrial era” style

Citigroup estimates that global spending on AI infrastructure will exceed 2.8 trillion dollars by 2029, aligning with NVIDIA’s forecast of 3–4 trillion by 2030.

Concrete examples:

• OpenAI is planning a 10GW data center (Project Stargate).
  
• A single training run by 2028 will require between 1 and 2GW.
  
• OpenAI, NVIDIA, and Oracle have forged a mega-deal worth 400 billion, described as “circular finance”.
  
• Oracle itself has issued 18 billion in bonds to strengthen its cloud infrastructure.
  

We are witnessing the largest infrastructure expenditure in the history of software.

The VC Market: AI Dominates Everything

In Q3 2025:

• 62% of all US VC funds have been allocated to AI.
  
• Investments in the first 9 months of 2025 already exceed the entire 2024 by 62%.
  
• +135% compared to 2023.
  

It’s an unprecedented market capitalization. And it shows no signs of slowing down.

The Role of Crypto: Not Hype, but Complementary Infrastructure

The report highlights a crucial point for those working in Web3:

Crypto does not “compete” with AI: it enables it.

In this new intelligence economy, blockchain becomes essential for:

• verify the integrity of the models;
  
• ensure data provenance;
  
• coordinate decentralized compute networks;
  
• provide instant and global payments;
  
• build identities and reputations for AI agents;
  
• create markets for data, compute, and inference.
  

The core of the report is precisely this: the future of AI will not be completely centralized nor entirely decentralized. It will be hybrid.

Conclusion: AI as the New National Infrastructure

“State of AI 2025” does not merely describe a growing sector: it outlines a new industrial era.

• Adoption is growing in every sector.
  
• Investments are at a record high.
  
• Infrastructures are expanding exponentially.
  
• Nations compete as in the space race.
  
• Energy and compute become strategic resources.
  

And above all: AI becomes a national infrastructure, on par with electricity, internet, and defense.The next major frontier — which we will analyze in the second article of the series — concerns the growing demand for decentralization, privacy, verifiability, and coordination: the birth of Decentralized AI.