Welcome to STATE OF AI REPORT 2025.
Published by Nathan Benaich on 9 October 2025.
👋 Read the 2025 Report
👀 See Live Survey Results
If 2024 was the year of consolidation, 2025 was the year reasoning got real. What began as a handful of “thinking” models has turned into a global competition to make machines that can plan, verify, and reflect. OpenAI, Google, Anthropic, and DeepSeek all released systems capable of reasoning through complex tasks, sparking one of the fastest research cycles the field has ever seen.
The business of AI finally caught up with the hype. The large labs and leading AI companies are now generating close to $20 billion in annual revenue. The absolute capabilities achieved by flagship models continue to climb reliably while prices fall precipitously, with the capability-to-price ratio doubling every 6–8 months. According to Ramp, 44% of U.S. businesses now pay for AI tools, up from just 5% in 2023. Average contracts have reached $530,000, and AI-first startups are growing 1.5× faster than their peers. Adoption has gone mainstream: in our new survey of 1,200 AI practitioners, 95% say they use AI at work or home, and three-quarters pay for it themselves. You can participate here.
This commercial reality has collided with infrastructure and geopolitics. Multi-gigawatt data centers like Stargate mark the start of the industrial era of AI, as the U.S., UAE, and China compete to build national compute backbones. Power and land are now as important as GPUs. In parallel, China’s DeepSeek, Qwen, and Kimi have closed the gap with GPT-5 across reasoning and coding, giving China a credible claim to second place in global AI capability.
The tone of the global AI conversation has also changed. The existential risk debate has cooled even further than it did last year, even though evidence of fragile alignment abounds. Safety now focuses on tangible problems: deceptive reasoning, model monitorability, and the balance between capability and control. Researchers are discovering that models can feign alignment, forcing labs to consider what one paper called a “monitorability tax”—accepting slightly weaker systems in exchange for transparency.
Meanwhile, politics has hardened. The U.S. has embraced an “America-first AI” agenda and turned industrial strategy into national security policy. Europe’s AI Act is mired in implementation woes. China has leaned into domestic open-weights ecosystems, overtaking Meta’s in developer share and fine-tunes, as Meta relinquished its mantle as the leader of open-source AI to China.
It’s clear that AI is the most important economic growth driver we have. It’s a system of production, reshaping energy markets, capital flows, and policy frameworks. What began with scaling laws is now being governed by physics, geopolitics, and colossal sums of capital.
- Key takeways from the 2025 Report include::
- OpenAI retains a narrow lead at the frontier, but competition has intensified as Meta reliquinshes the mantle to China’s DeepSeek, Qwen, and Kimi close the gap on reasoning and coding tasks, establishing China as a credible #2.
- Reasoning defined the year, as frontier labs combined reinforcement learning, rubric-based rewards, and verifiable reasoning with novel environments to create models that can plan, reflect, self-correct, and work over increasingly long time horizons.
- AI is becoming a scientific collaborator, with systems like DeepMind’s Co-Scientist and Stanford’s Virtual Lab autonomously generating, testing, and validating hypotheses. In biology, Profluent’s ProGen3 showed that scaling laws now apply to proteins too.
- Structured reasoning entered the physical world through “Chain-of-Action” planning, as embodied AI systems such as AI2’s Molmo-Act and Google’s Gemini Robotics 1.5 began to reason step-by-step before acting.
- Commercial traction accelerated sharply. Forty-four percent of U.S. businesses now pay for AI tools (up from 5% in 2023), average contracts reached $530,000, and AI-first startups grew 1.5× faster than peers, according to Ramp and Standard Metrics.
- Our inaugural AI Practitioner Survey, with over 1,200 respondents, shows that 95% of professionals now use AI at work or home, 76% pay for AI tools out of pocket, and most report sustained productivity gains, evidence that real adoption has gone mainstream.
- The industrial era of AI has begun. Multi-GW data centers like Stargate signal a new wave of compute infrastructure backed by sovereign funds from the U.S., UAE, and China, with power supply emerging as the new constraint.
- AI politics hardened further. The U.S. leaned into “America-first AI,” Europe’s AI Act stumbled, and China expanded its open-weights ecosystem and domestic silicon ambitions.
- Safety research entered a new, more pragmatic phase. Models can now imitate alignment under supervision, forcing a debate about transparency versus capability. External safety organizations, meanwhile, operate on budgets smaller than a frontier lab’s daily burn.
- The existential risk debate has cooled, giving way to concrete questions about reliability, cyber resilience, and the long-term governance of increasingly autonomous systems.
The State of AI Report is a team effort. I’m deeply grateful to everyone in the AI community who continues to create the breakthroughs that power this work. Special thanks to Zeke Gillman, Ryan Tovcimak, and Nell Norman, and to our many reviewers across research, industry, and policy for keeping us honest.
We write this report to share the most interesting things we’ve seen and celebrate the achievements of the AI community while informing conversation about the state of AI. I’d love to hear your thoughts on the findings, your take on our predictions, or any suggestions for next year’s edition.
Enjoy reading!
Nathan
🏡 Home 📧 Sign up to Air Street Press
Authored on the interwebs by:
This work is licensed under a Creative Commons Attribution 4.0 International License.