Since 2017, Stanford’s AI Index has served as the industry’s most comprehensive pulse check. At 400+ pages, the ninth edition published in 2026 offers a massive amount of data on the current state of AI - but for those of us building production-grade systems at GoML, the real value lies in the 'signal.' We’ve cross-referenced these global trends with our own findings on enterprise deployment to help separate the hype from what scales.
Here's what stood out.
AI capabilities are accelerating faster than most people expected
Contrary to popular belief, AI progress is not slowing down. In just one year, model performance improved from 60% to nearly 100% on SWE-bench Verified. This pace is far from gradual. Today’s models can surpass human baselines in PhD-level scientific questions, multimodal reasoning and competitive mathematics.
However, the problem is that the very models capable of winning in the International Mathematical Olympiad cannot reliably tell the time on an analog clock 50% of the time. This phenomenon is referred to as the "jagged frontier of AI" in the AI Index Report and is among its more valuable notions.
AI adoption has accelerated rapidly
Generative AI adoption reached 53% in just three years, outpacing the early growth of PCs and the internet, as highlighted in the AI Index Report. Adoption among organizations hit 88%, and four in five students now use AI in their studies.
The value of generative AI in the US is estimated at $172 billion annually. As noted in the AI Index Report, the gap between AI users and non-users continues to widen.
The U.S-China gap has effectively closed
The United States and China have swapped the lead position several times since early 2025. In March 2026, Anthropic's leading model takes first place with an advantage of only 2.7%. It is practically a tie. China has higher publication output, citation frequency, patent generation, and industrial robot usage, while the USA still generates more high-impact patents and leading models, according to the AI Index Report.
There is one notable statistic related to investment activities: worldwide corporate AI investment was estimated at $581.69 billion in 2025, which marks a 129.9% growth compared to the previous year. While the USA still invests more than others, the pool of talents is drying up. According to the AI Index Report, the amount of AI experts relocating to the USA has decreased by 89% since 2017.
The environmental cost is real and growing
The emissions from training Grok 4 were estimated at 72,816 tons of CO2 equivalent, which translates to approximately the amount emitted by 17,000 vehicles annually. The electricity consumption capacity of AI data centers has increased to 29.6 GW, which equals the power consumption of the whole State of New York.
Safety and policy are lagging behind
The level of trust in the government’s ability to regulate AI is only at 31% in the United States, which is the lowest among all surveyed countries in the AI Index Report. Europe scores much higher in terms of its public’s confidence in the regulation of AI compared to both the United States and China.
It is clear that there is a significant discrepancy when it comes to perception. While 73% of experts from the United States see the effects of AI in the workforce positively, only 23% of ordinary citizens think the same way.
What this means for teams building with AI
AI is no longer a distant opportunity. It is already delivering tangible benefits. Increases in productivity have ranged from 14% to 26% for software development and customer service, and up to 72% in marketing, as shown in the AI Index Report. However, usage declines where decision-making skills are required. AI-based solutions are currently being employed only marginally in virtually all organizational functions. The technology is robust, but still far from ready to be deployed off-the-shelf for complex processes.
The problem lies in implementation. Many organizations fail to move beyond the proof-of-concept stage because production requires the right technology stack and expertise. This is a gap we address in our whitepaper on production-grade AI systems. The key takeaway is clear. The real issue is not the model, but the gap between a functional demonstration and a reliable production-ready solution.
GoML offers six accelerators for key use cases, as well as a unique framework, called AI Matic for moving from POC to production. Its success is evident, such as an increase of 99% faster portfolio analysis and listing time reduction of over 80%. The difficult part is not building something but implementing it. We help businesses navigate this landscape with practical AI implementations grounded in what works. Reach out to explore what's possible for your team.
.jpg)
