Mary Meeker. If you’ve spent more than five minutes in a Silicon Valley boardroom or a tech-focused Slack channel over the last two decades, that name carries a specific kind of weight. For years, she was the "Queen of the Internet," the one who dropped the massive, multi-hundred-slide deck every spring that basically told the world where the money was going. Then, she stopped.
The last "classic" Mary Meeker internet report hit the web in 2019. After that? Silence. People started wondering if the era of the mega-trend deck was over, or if Meeker had finally decided that the internet had become so ubiquitous that "trends" were just... life.
Honestly, the world changed. We had a pandemic. We had the crypto boom and bust. We had the rise of TikTok. And through it all, the famous annual report was missing from its usual spot at the Code Conference. But in 2025, she came back. Only it wasn't exactly an internet report anymore. It was something much more intense.
The 2025 Pivot: Why the Internet Report is Now an AI Report
You've probably seen the headlines. In May 2025, Meeker and her team at BOND Capital released a 340-page behemoth titled Trends – Artificial Intelligence. It’s her first real "Trends" report since 2019.
The big takeaway? Meeker argues that AI isn't just a "feature" of the internet anymore. It has officially swallowed the internet.
✨ Don't miss: iPhone 16 Pro Max: What Most People Get Wrong
In her view, we’ve moved past the era of "internet trends" and into a phase of "unprecedented" acceleration—a word she uses over 50 times in the new document. She’s basically telling us that the old metrics we used to track—like how many hours people spend on mobile vs. desktop—are becoming irrelevant because the very nature of how we interact with machines has shifted from clicking and typing to "conversational interfaces."
The "Sputnik Moment" for Modern Tech
Meeker frames this shift as a "Sputnik moment." It's a heavy comparison. She isn't just talking about cool new apps; she’s talking about a global arms race.
According to the data in the report, the six big tech giants (Apple, Nvidia, Microsoft, Alphabet, Amazon, and Meta) poured $212 billion into AI infrastructure in 2024 alone. That is a 63% jump in just one year. To put that in perspective, these companies are effectively burning through $580 million every single day to build the "highways" for AI.
What the Data Actually Says (and Why It’s Scary)
A lot of people think AI is just another hype cycle like 3D TVs or the Metaverse. Meeker’s data suggests otherwise. The speed of adoption is, frankly, nuts.
- The 60-Day Sprint: It took the internet seven years to reach 100 million users. It took smartphones two years. ChatGPT did it in 60 days.
- The Global Launch: Unlike the original internet, which started in the U.S. and slowly trickled out to the rest of the world, AI went global on Day 1. By April 2025, ChatGPT hit 800 million weekly users.
- The Cost Collapse: This is the part that most people miss. Meeker points out that while it's getting more expensive to train models (think $10 billion per model soon), the cost to use them (inference) has dropped by 99.7% in two years.
What does that mean for you? It means that using AI is becoming "too cheap to meter." It's following a path similar to the lightbulb, but instead of taking 80 years to become affordable, it took about 18 months.
💡 You might also like: Finding an mp4 converter mac os x That Actually Works Without Killing Your Battery
The Rise of the "Computational Labor Unit"
One of the most fascinating (and slightly terrifying) concepts in the 2025 report is what Meeker calls the "computational labor unit."
Basically, she’s saying we shouldn't measure company size by headcount anymore. In the old world, if you wanted to double your output, you hired more people. In the AI era, team output equals Humans × Agents.
She cites Klarna as the poster child for this. Their AI assistant handled two-thirds of customer chats in its first month, doing the work of roughly 700 full-time agents. It’s not just tech, either. Kaiser Permanente has over 10,000 doctors using AI medical scribes. The "internet" part of this—the website or the app—is just the plumbing. The value is now in the autonomous execution.
The China Factor: A Different Game
You can't talk about a Mary Meeker internet report without looking at China. Meeker has always been one of the first to point out when China starts outperforming the West in specific niches (like mobile payments back in the day).
In 2025, she’s sounding the alarm on two fronts:
✨ Don't miss: PSCC Anti Theft Tracking: Why Asset Protection Just Got Way More Complicated
- Industrial Robots: China now has more industrial robots installed than the rest of the world combined. They are automating the physical world while the West is still mostly focused on automating white-collar emails.
- Open Source Power: While we talk about OpenAI and Google, Chinese open-source models (like DeepSeek) are closing the gap. In some math and coding benchmarks, the difference between a "closed" U.S. model and an "open" Chinese model has shrunk to almost zero.
Is the "Queen of the Internet" Still Right?
Look, Meeker has had her critics. Back in the dot-com burst of 2000, she was vilified for being too bullish on companies that eventually went to zero. People say her reports can be "rambling" or "overambitious."
But honestly? If you look at her track record of spotting the big shifts—the move to mobile, the rise of the "freemium" model, the explosion of interactive gaming—she’s usually ahead of the curve.
The 2025 report acknowledges a massive problem, though: The Monetization Paradox. OpenAI is projected to bring in $3.7 billion in revenue but spend over $5 billion on compute. It’s "investment inversion." We are currently in a "shovels vs. gold" economy. Nvidia and the cloud providers are getting rich (the shovels), while the companies actually trying to build the AI "gold mines" are still bleeding cash.
Practical Steps to Navigate the Post-Report World
If you’re a business owner, a developer, or just someone trying not to get left behind, Meeker’s latest findings offer a few survival tactics.
- Move to the Application Layer: Don't try to build your own "frontier" model unless you have billions in the bank. The value is now in "Vertical AI"—tools that do one specific job (like legal discovery or medical coding) better than a general-purpose bot.
- Focus on Workflow, Not Features: AI "moats" are shallow. If your only advantage is a "cool AI feature," a bigger company will copy it in a week. Your real moat is how deeply you integrate into a user's daily mess of a workflow.
- Prepare for "Agentic" Expectations: Customers no longer want to wait for a "lead capture form" to be processed by a human. They want "Time-to-Value" in under five minutes. If your business doesn't have an agentic layer that can solve problems autonomously, you're going to feel very slow, very fast.
- Watch the Energy: Meeker highlights that data centers now consume 1.5% of global electricity. As we move into 2026 and beyond, energy efficiency and "on-device" AI (running things locally on your phone or laptop) will become a huge competitive advantage to avoid those massive cloud bills.
The Mary Meeker internet report might have changed its name, but the lesson is the same as it was in 1995: things are moving faster than you think, and the "unprecedented" is the new normal.
To stay ahead of these shifts, the most effective move right now is to audit your internal processes for "agentic readiness." Identify the three highest-volume, lowest-complexity tasks in your workflow and test a specialized vertical AI agent against them. This isn't about replacing people; it's about increasing the "output per head" before your competitors do the same. Also, keep an eye on the decline of inference costs—if a task was too expensive to automate six months ago, it’s probably affordable now.