Unprecedented Indeed! The Contrarian Take on Mary Meeker's 2025 AI Report
When you see the winds of change, acknowledge and align. And in this particular case, move fast!
It is not every day one looks forward to reading a 340-page report! And if that happens, it talks of passion for a subject and the authority of the author. In the world of technology analysis, few names carry as much weight as Mary Meeker. Her annual reports have been essential reading for investors, entrepreneurs, and technology enthusiasts for decades. This year's "Trends - Artificial Intelligence" report, spanning a massive 340 pages, represents her first major analysis since 2019 and focuses exclusively on the AI revolution reshaping our world.
What makes this report particularly striking is Meeker's repeated use of the word "unprecedented”, appearing over 50 times throughout the document, underscoring her belief that we're witnessing something truly extraordinary in the technology landscape (a note to the naysayers in 2022 and 2023 - I told you so!) The sheer scale and pace of AI's development and adoption are unlike anything we've seen before, even compared to previous technological revolutions.
Let's dive into the key sections of Meeker's report and explore what they mean for our collective future.
The Unprecedented Speed of AI Adoption
Meeker's data paints a clear picture: AI adoption is happening at a pace that dwarfs previous technological revolutions. ChatGPT reached 800 million weekly active users in just 17 months – a growth trajectory that makes even the meteoric rises of platforms like Instagram and TikTok look sluggish by comparison.
What's most fascinating is how this adoption isn't limited to tech-forward regions.
While the internet took 23 years for 90% of its users to be outside North America, ChatGPT achieved this milestone in just three years. This global simultaneous adoption represents a fundamental shift in how technology diffuses across borders and cultures.
The accelerated adoption curve means businesses and educational institutions have significantly less time to adapt than during previous technological shifts. The luxury of a gradual transition simply doesn't exist with AI. Organizations that fail to integrate AI capabilities quickly risk becoming irrelevant at a pace we've never seen before.
The Economics of AI: A Tale of Two Cost Curves
One of the most revealing sections of Meeker's report examines the diverging economics of AI development. While training costs for advanced models have skyrocketed, increasing approximately 2,400 times between 2016 and 2024 and now approaching billions of dollars per model, inference costs (the expense of actually using these models) have plummeted by 99% over just two years.
This bifurcation creates a fascinating dynamic: building cutting-edge AI requires enormous capital investment, but deploying and using AI is becoming increasingly affordable.
NVIDIA's 2024 Blackwell GPU, for instance, uses 105,000 times less energy per token than its 2014 predecessor.
This economic reality is reshaping the competitive landscape across industries. Only the largest tech companies and best-funded startups can afford to build foundational models, but virtually any organization can leverage these models through APIs and fine-tuning.
This democratization of AI usage, coupled with the concentration of AI development, is creating new power dynamics in the technology ecosystem.
The Rise of Open Source and China's Growing Influence
Perhaps the most geopolitically significant section of Meeker's report addresses the growing competition between proprietary and open-source AI models, with China emerging as a leader in the latter category.
In 2025 alone, China released three large-scale open-source AI models: DeepSeek-R1, Alibaba's Qwen-32B, and Baidu's Ernie 4.5 and they are rapidly closing the performance gap with proprietary Western models. DeepSeek, in particular, has captured 21% of the global LLM user share in just a few months.
This shift represents more than just technical competition, it's reshaping global technology influence. Open-source models are becoming the "garage labs" of AI innovation, fast, chaotic, global, and fiercely collaborative. For businesses and governments, this means strategic decisions about which AI ecosystem to invest in have geopolitical implications beyond mere technical considerations.
India's Critical Need to Catch Up
While China and the US race ahead in AI development, India is lagging behind and finds itself at a critical juncture. The disparity is even more pronounced in generative AI, where China has filed over 38,000 patents, the US around 6,500, and India only 1,350. Despite being ranked sixth globally in AI patents, India's total represents just 3.5% of China's advancements and about a fifth of the US's output.
India faces significant structural challenges in its AI development journey. A recent report highlighted key hurdles including a shortage of high-end AI hardware, limited access to advanced GPUs and cloud computing, and insufficient funding for research and development. These factors hinder India's ability to create cutting-edge AI systems that can compete globally.
The time for collective action is now. India's government has taken initial steps with the IndiaAI Mission, allocating ₹10,372 crore (approximately $1.25 billion) over five years to strengthen the country's AI ecosystem. This includes developing India's foundational AI models, establishing data centres, and improving AI infrastructure. However, this investment pales in comparison to China's $137 billion annual investment in AI.
Despite these challenges, there are promising signs. India leads the world in AI adoption among knowledge workers, with 92% of them using AI at work, far above the global average of 75%. The country is also showing the highest average annual growth rate in AI patents among the top five leaders at 56%, but that is expected at the initial stages.
These indicators suggest India has the potential to accelerate its AI development if it can address its infrastructure and funding gaps.
The Infrastructure Challenge: Data Centers and Energy
Meeker dedicates significant attention to the physical infrastructure supporting the AI revolution, particularly data centres and their growing energy demands. Global IT company data centre capital expenditure reached $455 billion in 2024 and continues to accelerate.
The report highlights that data centre electricity consumption has tripled since 2005, with the U.S. alone accounting for 45% of global data centre energy usage. This surge in power demand is creating new bottlenecks in AI development, with energy availability potentially becoming a more significant constraint than chip supply.
The environmental and infrastructure implications of AI are often overlooked in discussions about the technology's potential. As AI workloads grow increasingly power-hungry, we'll need to confront difficult questions about energy production, distribution, and sustainability. Organizations building AI strategies must consider not just computational resources but also their energy footprint and geographic proximity to power sources.
AI's Impact on Work and Employment
Meeker doesn't shy away from addressing AI's profound implications for the workforce. The report documents a 448% increase in AI job postings over seven years, while non-AI IT jobs declined by 9% during the same period. This shift reflects a fundamental restructuring of labour markets around AI capabilities.
The report acknowledges that jobs relying on processing large volumes of structured data to produce rules-based decisions are particularly vulnerable to automation. However, Meeker remains cautiously optimistic, noting that "history and pattern recognition suggests the role of humans is enduring and compelling" and that "technology-forward leaps have typically driven productivity and efficiency gains and more, but new, jobs”.
While I appreciate Meeker's historical optimism, the pace and scope of AI-driven workplace transformation will be unlike anything we've experienced before. The distinction between physical and cognitive automation is crucial - previous waves of automation primarily affected manual labour, while AI targets knowledge work that was previously considered safe from technological displacement. This will require fundamentally new approaches to education, career development, and social safety nets.
AI in Education: Promise and Peril
Meeker's analysis of AI's impact on education highlights both transformative potential and significant challenges. She notes how AI can personalize learning journeys, analyze vast amounts of student data, and create more engaging educational experiences.
However, critics have pointed out that Meeker's vision may overemphasize efficiency and productivity at the expense of the human elements that make education meaningful. There's tension between viewing education through a lens of measurable outcomes versus understanding it as a deeply personal, relational process.
AI will undoubtedly transform education, but we must be careful not to reduce learning to a purely transactional, automated process. The most valuable aspects of education, critical thinking, creativity, ethical reasoning, and human connection, cannot be outsourced to algorithms. Instead, we should focus on how AI can augment human teachers, freeing them from administrative burdens to focus on these higher-order aspects of learning.
The Emergence of AI Agents
One of the most forward-looking sections of Meeker's report examines the evolution from AI assistants to autonomous AI agents capable of performing complex, multi-step tasks independently. This shift represents a fundamental change in how we interact with technology, from tools we actively use to systems that proactively work on our behalf.
The report documents a 1,088% increase in Google searches for "AI agent" over just sixteen months, reflecting growing interest in this emerging paradigm. Major companies like OpenAI, Salesforce, Anthropic, and Amazon have all launched agent platforms with varying capabilities.
The transition from passive AI tools to proactive AI agents will fundamentally reshape human-computer interaction. As these systems become more capable of understanding context, maintaining memory, and coordinating across applications, they'll increasingly function as digital extensions of ourselves rather than separate tools. This will raise profound questions about agency, responsibility, and the boundaries between human and machine decision-making.
Conclusion: Navigating the AI Revolution
Mary Meeker's comprehensive report makes one thing abundantly clear: AI is not just another technology wave. It's a fundamental reset of business, society, and geopolitics.
The pace and scale of change are truly unprecedented, with implications that touch every aspect of our lives. As we navigate this transformation, several principles should guide our approach:
1. Adaptability over prediction: The speed of AI development makes precise predictions futile. Instead, we should build organizations, educational systems, and policies that can adapt quickly to emerging realities.
2. Human-AI collaboration: Rather than framing the future as a competition between humans and AI, we should focus on how the unique capabilities of each can complement the other.
3. Ethical guardrails: As AI becomes more powerful and autonomous, establishing clear ethical principles and governance frameworks becomes increasingly important.
4. Inclusive access: Ensuring that the benefits of AI are broadly shared across countries, socioeconomic groups, and industries will be crucial to preventing new forms of inequality.
The AI revolution is here, and its trajectory is clear. What remains to be determined is how we harness its potential while mitigating its risks. Meeker's report provides an invaluable roadmap for understanding where we are and now it's up to us to decide where we go from here.