Meeker's 2025 AI Report
The May 2025 report broke Meeker's six-year publishing silence with a singular focus: artificial intelligence. At 340 pages, it is her most concentrated analytical document, applying the comparative framework honed across decades of Internet Trends reports to a phenomenon she explicitly called unprecedented. The report layers AI adoption curves against every prior general-purpose technology — telephone, radio, television, internet, mobile — revealing a step function rather than a continuation. It documents infrastructure investment at $212 billion by the Big Six in 2024, inference cost declines of 99% over two years, and global distribution patterns that mirror prior inequalities with novel structural intensification. Its most analytically significant feature is not what it asserts but where it stops: the repeated acknowledgment, at the junctures that matter most, that only time will tell.
The Velocity Trap — Contrarian ^ Opus
There is a parallel reading of Meeker's unprecedented acceleration: not as evidence of AI's unique transformative power, but as evidence of infrastructure capture creating artificial velocity. The speed she documents—ChatGPT reaching 100 million users in two months—measures distribution through channels already owned by the platforms that built the product. When the same six companies control both the infrastructure and the applications, "adoption" becomes a metric of bundling power rather than genuine user demand. The curve compression she celebrates may be documenting the elimination of competitive adoption cycles, not their acceleration.
The $212 billion infrastructure investment she neutrally presents deserves a different frame: it represents the Big Six building the only on-ramps to the technology while simultaneously positioning themselves as the primary landlords of its economic returns. Her historical framework—comparing AI to the telephone, radio, television—misses that those technologies were subject to common carrier regulations, antitrust enforcement, and public utility frameworks precisely because their concentration threatened democratic governance. The report's silence on regulatory structure is not neutral restraint but a choice to analyze adoption curves while ignoring the power geometries those curves are being drawn upon. What she marks as the honest boundary of her analytical framework—where only time will tell—may be the central question: whether we are witnessing technological adoption or the completion of platform enclosure.
In the AI Story
The report's structural innovation is its sustained focus. Previous Meeker reports ranged across dozens of industries and platforms; this one concentrates on a single phenomenon with the kind of density her earlier work reserved for annual surveys of the entire technology landscape. The concentration is itself a claim: that AI warrants the full weight of her analytical apparatus applied to one subject.
The report's empirical foundation rests on adoption curve compression — the progressive shortening of intervals between introduction and mass adoption across successive technology waves. The charts stack these curves against each other, producing a visual argument that does not require Meeker's editorializing. The telephone took 75 years to reach 50 million users; ChatGPT reached 100 million in two months. The acceleration itself is accelerating.
The report identifies the mechanism behind the acceleration: AI is a compounder on internet infrastructure. Unlike prior technologies that required new distribution channels, AI inherited the global internet, mobile ecosystems, and cloud infrastructure that two decades of prior investment had already built. Each S-curve laid the groundwork for the next. AI inherited all of them simultaneously.
The report's acknowledged boundaries are its most consequential feature. At specific junctures — monetization outcomes, productivity realization timelines, the ultimate role of humans — Meeker marks the edge of what her quantitative framework can resolve. These markers are not hedging. They are analytical statements about the scope of adoption data as a predictor of outcome.
Origin
The report emerged from Meeker's foundation of Bond Capital in 2019, her growth-stage venture firm that succeeded her partnership at Kleiner Perkins. Between 2019 and 2025, Meeker published no major reports — a silence that preceded the most consequential technology transition since the commercial internet.
The decision to break the silence with a single-subject report on AI, rather than a return to the sweeping Internet Trends format, signaled her judgment that the phenomenon warranted concentrated analytical attention. The report was published as a public document, continuing her tradition of making rigorous technology analysis accessible beyond the industry's inner circles.
Key Ideas
Unprecedented is a technical claim. When Meeker uses the word, it carries the weight of three decades of comparative data. She has plotted every adoption curve against every prior curve, and AI does not fit the pattern.
Compounding explains the acceleration. AI achieved its adoption speed not by inventing distribution but by inheriting the infrastructure of every prior technology wave simultaneously.
The boundaries are marked honestly. The report's only time will tell junctures are not hedging but precise acknowledgments of what adoption data can and cannot predict.
Infrastructure investment at unprecedented scale. $212 billion from the Big Six in 2024 alone, a 63% year-over-year increase — capital flowing into existing revenue streams, not speculative ventures.
The map knows where it ends. The report's most valuable feature is not the density of its data but its willingness to mark the territory beyond what data can describe.
Debates & Critiques
Critics argued the report's neutral presentation of infrastructure concentration understated the political implications of power consolidation. Others suggested that Meeker's historical optimism — grounded in prior technology transitions producing more jobs than they destroyed — underweighted the structural differences between AI's automation of judgment-intensive tasks and prior automation of routine work. The report's defenders noted that its restraint is its authority: it presents evidence, identifies trends, and leaves value judgments to the reader.
Appears in the Orange Pill Cycle
The Speed-Structure Synthesis — Arbitrator ^ Opus
On the empirical foundations—adoption curve compression, infrastructure compounding, the $212 billion capital deployment—Meeker's framework operates at 100% accuracy. These are measurable realities requiring no editorial overlay. The data describes what is happening with unusual precision. The question is what explanatory frame that data supports.
On the mechanism of acceleration, the weighting is roughly 70/30 in favor of Meeker's compounding explanation, with the contrarian view capturing a meaningful remainder. AI did inherit global internet infrastructure, cloud distribution, and mobile ecosystems—this explains most of the velocity. But roughly 30% of the speed derives from platform control: the same companies building the models own the distribution channels, creating adoption curves that conflate genuine demand with bundling power. Both mechanisms operate simultaneously. The synthesis recognizes that inheritance and capture are not competing explanations but complementary forces.
On the boundaries Meeker marks—monetization outcomes, productivity realization, the ultimate role of humans—her restraint is methodologically sound (100%), but the contrarian critique identifies the load-bearing omission: power structure is not outside her analytical scope but central to the phenomenon she is measuring. The true synthesis reframes the report not as incomplete analysis but as a specific genre—adoption cartography—that deliberately brackets political economy to maintain empirical discipline. The value lies in recognizing both what the map reveals (distribution dynamics with unmatched clarity) and what it structurally cannot show (the governance questions those dynamics require). Meeker's only time will tell is honest about data's limits; the contrarian view correctly notes that some questions demand frameworks data cannot provide.
Further reading
- Mary Meeker, Trends — Artificial Intelligence (Bond Capital, 2025)
- Mary Meeker, Internet Trends reports (Morgan Stanley / Kleiner Perkins, 1995–2019)
- Carlota Perez, Technological Revolutions and Financial Capital (Edward Elgar, 2002)