AsianFin -- Mary Meeker, the renowned venture capitalist once dubbed the Queen of the Internet, has returned to the spotlight with a sweeping 340-page report titled Trends — Artificial Intelligence, marking her first major trend study since 2019.
The report, which uses the word "unprecedented" 51 times, outlines in data-rich detail how the acceleration of AI development, adoption, and commercialization is reshaping the technological landscape faster than any previous shift in history.
Meeker, founder and general partner at Bond Capital and formerly the head of growth investments at Kleiner Perkins — where she backed companies like Facebook, Spotify, Ring, and Square (now Block) — now turns her full attention to the generational disruption that is artificial intelligence.
The speed and scope of AI development are truly unprecedented — and the data proves it, Meeker wrote in the report.
According to Meeker, AI's ascent is marked by a series of compounding forces:
· User Growth: ChatGPT reached 800 million weekly active users in just 17 months — a scale and speed unmatched even during the internet boom. Its annual recurring revenue (ARR) growth has also exceeded that of any product from the Web 1.0 or 2.0 eras.
· Cost Collapse: While model training can still cost upwards of $1 billion, inference costs — the cost of using AI — have plummeted 99% over the past two years, according to Stanford data.
· Global Competition: As Nvidia's latest Blackwell GPU boasts a 105,000-fold energy efficiency gain over 2014's Kepler chips, Chinese companies are rapidly catching up with open-source approaches. Meeker says this dynamic has unleashed a global technology arms race.
The report highlights a broader power shift: AI innovation has migrated from universities and research labs to companies and proprietary platforms. Meeker identifies 2019 — the year GPT-2 launched with limited parameters — as a turning point when closed-source models began to dominate, driven by profit incentives, competitive advantage, and safety concerns.
Models like OpenAI's GPT-4 and Anthropic's Claude are trained on massive private datasets in opaque environments, requiring months and millions of dollars to develop. These models perform exceptionally well and are widely used by enterprises and governments — but they lack transparency. Their inner workings, including weights and training data, remain inaccessible to the public.
AI's evolution from open research to closed product experiences — governed by APIs and legal barriers — is a paradigm shift, Meeker notes.
The Rise of Open Source — And a New Wave of Competition
As closed models solidify dominance in consumer and enterprise markets, open-source models are regaining momentum due to their accessibility, cost-efficiency, and customizability. Developers, startups, and researchers are increasingly favoring models like Meta's Llama and DeepSeek, which offer full download access and local deployment capabilities.
Open-source AI has become the garage lab of the modern tech era — fast, chaotic, global, and fiercely collaborative.
According to the data, China leads the world in the number of open-source large model releases as of Q2 2025, including DeepSeek-R1, Alibaba's Qwen-32B, and Baidu's Ernie 4.5.
Open models are closing the performance gap with closed models faster than expected, with inference, code generation, and multilingual capabilities nearing parity. Critically, these gains are achieved at a fraction of the cost.
Behind the AI boom lies a massive wave of infrastructure investment — from chips to cloud services. Meeker draws a direct line between AI's ascent and capital expenditure:
· Token Costs: From 2022 to 2024, the cost per token in LLMs dropped an estimated 99.7%, driven by gains in hardware and algorithm efficiency.
· Hardware Efficiency: Nvidia's Blackwell, Google's TPU, and Amazon's Trainium are accelerating the transition to purpose-built AI computing. These are not side projects — they are strategic bets on the future, Meeker said.
· Cloud Pressure: The compute needs of LLMs are pushing enterprise IT budgets, cloud providers, and chipmakers into an unprecedented flywheel of growth — and strain.
The report spotlights India as a major consumer market for AI platforms. The country accounts for 13.5% of ChatGPT's mobile app user base, surpassing the US (8.9%) and Germany (3%). India is also the third-largest market for China's DeepSeek.
Meeker notes India's strategic role in AI adoption, calling it a bellwether for platform-level growth in emerging economies.
AI's ecosystem is bifurcating. On one path are proprietary models like GPT-4 and Claude, favored by large enterprises for their performance and ease of use. On the other are open models like Llama and Mixtral, valued for flexibility, transparency, and sovereignty.
Closed models dominate in user experience and enterprise traction. But open models are gaining ground in key areas like local-language adaptation, grassroots tooling, and sovereign AI initiatives.
Open source is fueling AI nationalism, Meeker wrote, while closed models remain the standard in mainstream software stacks.
Beyond software, Meeker's report charts AI's growing role in the physical world. From robotics and autonomous vehicles to healthcare assistants, AI is increasingly embedded in daily life.
Rather than eliminating jobs, Meeker argues, AI is augmenting them. She likens LLMs to co-pilots for programmers, analysts, and writers — accelerating productivity rather than replacing human work.
Since 2018, job postings related to AI have surged 448%, underlining the sector's insatiable demand for talent.
The software business model is also undergoing a transformation. For decades, enterprise software scaled through vertical SaaS — narrow tools sold to niche customers. But generative AI is shifting the focus toward horizontal platforms that integrate productivity, communication, and search into one seamless interface.
Companies like Microsoft are embedding Copilot across their stack. Zoom and Canva are integrating AI into user workflows. And developer-focused firms like Copula are injecting generative AI into data and dev stacks.
As the global AI race intensifies, so too does the geopolitical competition over chips, data centers, and technical leadership. Meeker likens today's AI arms race to the Space Race of the Cold War, with the stakes no less existential.
But Meeker also cautions that AI's breakneck growth comes with risks: bias, misinformation, unpredictability. She calls for honest leadership, clear rules, and smarter systems to manage the challenges ahead.
AI is both a milestone in tech history — and an unknown variable in the future of business, Meeker concludes. Buckle up for an unprecedented ride.
