Mary Meeker

Who she is

Affiliation: Founder & General Partner, BOND (founded 2018, first close 2019 –) — role check run 2026-09-21: unchanged. Forbes: "Founder and General Partner, Bond Capital," a "$5.9 billion global portfolio," board seats at Block, Genies and OpenEvidence (Forbes, Midas List 2026). Ex-partner, Kleiner Perkins (2010–2018; co-led the $2.8B Digital Growth Funds); ex-Managing Director and lead internet analyst, Morgan Stanley (1991–2010). Author of the annual Internet Trends Report (1995–2019), then six years of silence, then BOND's 340-page Trends – Artificial Intelligence (May '25). Correction carried from the --investing twin (2026-07-19), re-confirmed today: no 2023 or 2024 editions existed; v1's "AI Trends 2023–present" was wrong. bondcap.com/reports/tai still shows only "May '25" and no 2026 edition was found in three searches (Forbes' "revived in 2024" conflicts with BOND's own dating; BOND wins). "Queen of the Net" from the dot-com era.
One-line position: Technology adoption follows measurable S-curves, and AI is now the steepest one — the macro data says this platform shift will restructure enterprise productivity, labor markets, and economic output faster than mobile did, and the charts already show it. Her 2025 sharpening: AI is "not simply another technology cycle but a compounding force" — "this time it's happening faster" (The Batch, 2025-06).

Discipline & technical bet

A sell-side analyst's method (count what is used, overlay the curve on prior eras, follow the capex) run inside a growth fund. She makes no model-architecture wager; hers is a layer bet, one above the frontier labs: inference serving (Baseten), modality APIs (ElevenLabs — BOND joined the $500M Series D at $11B as a new investor, ElevenLabs, 2026-02-04), vertical workflow (OpenEvidence, AlphaSense), consumer creative (Canva, $42B as of 2025-08 per Forbes), distribution (Substack). No frontier-lab position is public. The deck's numerical spine: inference cost (the price of one model task) down 99.7% in two years; ChatGPT to 100M users in five days and "90 percent adoption in just three years" versus the internet's 23; big-tech capex compounding above 21% a year since 2014; AI job postings +448% in seven years while non-AI IT postings fell 9% (Ken Yeung's chart digest, 2025-06-06). Her one new unit is the "computational labor unit" — one human plus several agents — and the deck's boldest line is "AI will shrink org size and increase output" (SaaStr digest, 2025-06).

Key claims (Says)

• AI adoption is the fastest technology S-curve in history: internet ~7 years to 100M users, mobile ~4, ChatGPT ~2 months; enterprise adoption follows a similarly compressed curve. Empirical. Two halves. Held (consumer) — 900M weekly actives on 2026-02-27, double a year earlier (Search Engine Land). Drifted (enterprise) — on the strict production-use measure, 17–20% of US firms used AI to make goods or services, Dec 2025–May 2026 (Census BTOS, 2026-05); the Fed reconciles ~18% of firms with 41% of workers using generative AI at work (FEDS Note, 2026-04-03). Workers ride the fast curve; firms-as-producers ride a normal one.
• Enterprise productivity gains are real and measurable now: 20–60% lifts in narrow domains, visible in revenue-per-employee. Empirical. Drifted — task-level lifts are real; "visible in company metrics" is not. 11% of S&P 500 companies quantified an AI productivity gain on earnings calls, 2% put a number on earnings, and the earnings-growth gap between those firms and the rest was "small and statistically insignificant" (Goldman via Yahoo Finance, 2026-08-22). Atlanta Fed survey work: 2025 AI-attributable productivity ~0.8% in high-skill services, ~0.4% in manufacturing (TechTimes, 2026-09-09).
• AI is a platform shift, not a feature cycle: comparable to internet (1995) and mobile (2007). Empirical / interpretive. Held — the four hyperscalers guided 2026 capex to ~$700B, up ~60% on 2025's record $388B (CNBC, 2026-02-06); two new incumbents run at $40B–$100B annualized; a new ad channel lives inside chat. New platforms, business models, incumbents, distribution — all four.
• The economic restructuring will be massive: AI-driven productivity gains show in GDP this decade, breaking the 2010s stagnation. Empirical, contested. Drifted — on attribution, not level. Productivity's contribution roughly doubled (1.2 → ~2.5 points annualized, late-2022 to mid-2025) but is concentrated in a few industries and "AI adoption explains little of the aggregate gain" (KC Fed, 2026-02-11). Q2 2026: productivity +1.4%, output +1.7%, hours +0.3% (BLS). The number moved her way; the cause is not yet hers.
• Data is the new moat: proprietary data plus workflow integration capture disproportionate value — not necessarily the model builders. Analytic. Wrong (2026 snapshot) on its distinctive half. Anthropic's run-rate passed $65B (Bloomberg, 2026-08-17) and is pacing above $100B (Axios / NYT, 2026-09-18); OpenAI is reported near $40B (aggregator, lower confidence); about half of 2026 S&P 500 earnings growth goes to AI-infrastructure suppliers, not users (TechTimes, above). Mechanism vs valence: the open-weights gap did narrow and app-layer names with workflow lock-in are winning (ElevenLabs $330M+ ARR at end-2025) — mechanism holds, "not the model builders" is inverted. Run-rate is not profit; the check below covers that.
• The labor transition is real but manageable: economies absorb technology-driven labor shifts, painfully. Empirical / normative blend. Drifted — aggregate holds, distribution doesn't. Payroll data: "no widespread, economy-wide job displacement," but a 19% employment gap for 22–25-year-olds in AI-exposed occupations, up from 15% in July 2025, "primarily through reduced hiring" (Stanford Digital Economy Lab, 2026-08-12). Labor's share of output 52.9%, lowest since 1947; real hourly compensation −3.1% in Q2 2026 (Indeed Hiring Lab, 2026-08-06). Manageable for whom is the live question.
• Usage data > capability benchmarks: measure what people do, not what a model could do on a leaderboard. Methodological. Held — and vindicated: the scaling camp's founder conceded in 2025-11 that models crush benchmarks while economic impact stays thin (Sutskever, Dwarkesh, 2025-11); Goldman's spend-up/earnings-flat finding is the same gap from the buyer side.
• Advertising and attention economics will be restructured by AI: discovery, recommendation and creation change at once. Empirical / interpretive. Held (mechanism) — OpenAI announced ads for ChatGPT's Free and Go tiers 2026-01-16, switched them on in the US 2026-02-09, opened self-serve buying 2026-05-05 (Axios), reached the UK, Mexico, Brazil, Japan and Korea by 2026-08-11, and targets $2.5B of 2026 ad revenue (OpenAI). Plumbing rebuilt; market share is the check.

Notable predictions — with falsifiable checks

• (~1995–1999) Internet becomes the dominant commercial platform. Held (historical) — among the earliest institutional calls on internet scale.
• (~2004–2010) Mobile internet eclipses desktop. Held (historical) — tracked in real time in the decks.
• (2017–2019) Privacy regulation (GDPR, CCPA) reshapes the internet economy. Held (historical, largely) — friction real, enforcement uneven.
• (2023–2024) AI adoption steeper than mobile or internet. Graded under claim 1 (consumer Held, enterprise Drifted). Check: OpenAI officially confirms ≥1B weekly actives by 2027-06 (June 2026 reports put monthly actives at 1B; weeklies unconfirmed); Census BTOS firm production-use ≥30% by 2027-12. Miss both and the enterprise half is Wrong.
• (2024–2025) Enterprise AI produces productivity gains visible in company financials by 2026–2027. Live — not yet visible (Goldman 11% / 2%). Check: by 2027-12, more than 25% of S&P 500 companies quantify AI gains on calls AND Goldman's reporter-vs-non-reporter earnings gap turns statistically significant. Unmet at 2027-12 = Wrong.
• (2025) AI restructures the advertising and content-discovery stack. Graded under claim 8 (Held on mechanism). v1's "[TBD: confirm specific claims from Bond presentations]": this pass found no explicit ads prediction in the 2025 deck — the item is v1's synthesis, graded on the observable regardless. Check: OpenAI reports ≥$2.5B 2026 ad revenue (its own target) by 2027-03; Google search-ad growth stays positive through 2027.
• NEW (2025-05) "AI will shrink org size and increase output." Live — early evidence her way: two years of output up with hours flat (Indeed, above); Canaries' hiring-margin effect is the same pattern at entry level. Check: BLS nonfarm output growth exceeds hours growth by ≥1 point in at least 6 of the 8 quarters through 2027-12.
• NEW (2025-05) Hyperscaler capex is a rational buildout — she printed the bear chart herself (2024 US LLM revenue ≈$3.7B vs ≈$5B compute) and kept the bullish read. Held on trajectory — 2026 guidance ~$700B, +60% (CNBC, above); investors turned skeptical in July (CNBC, 2026-07-28) but no guide was cut. Check: any of the four guides 2027 capex below 2026 (Jan–Feb 2027 earnings) — the first observable that tests "rational."
• NEW (2025-05) Inference cost collapse (−99.7% in two years) — "Will we achieve a ZIRP for inference?" Held on direction — the open-to-closed gap on the Artificial Analysis index narrowed to 6 points from 13; Chinese open-weight models carried ~61% of OpenRouter tokens by 2026-05 (Data Gravity, 2026; secondary). Check: frontier-class price per million tokens falls another ≥50% by 2026-12 on the Artificial Analysis price index.
• NEW (2025-05) China leads open-source; the AI race is "our space race" (PitchBook, 2025; not fetched, title and framing only). Held (open-source half) — Llama fell off the usage rankings; four of the five most-used open models are Chinese (Data Gravity, above). Check: a US lab's open-weight model holds the #1 OpenRouter slot for a full quarter by 2027-06.
• NEW (2025-05-30, Axios) OpenAI shows "no signs" of falling behind. Drifted — OpenAI still leads consumer (900M WAU) but Anthropic passed it on annualized revenue in 2026 (Bloomberg, Axios, above); caveat: Anthropic books cloud-reseller spend gross, OpenAI nearer net. Check: OpenAI's annualized revenue exceeds Anthropic's on like-for-like accounting at 2026-12; if not, Wrong.
• Unfalsifiable — noted, not graded: "in the next decade or two, imagining a world without AI will likely feel the same" as life without the internet; "a period for lots of wealth creation and wealth destruction" (Axios Q&A, 2025-05-30). Rewrite or retire.

Revealed behavior (Does)

• Money — the layer bet, confirmed in 2026: new-investor check into ElevenLabs' $500M Series D (2026-02-04, primary); Baseten — led the $150M round (2025-09), listed among cumulative investors at the $1.5B Series F (2026-06-22) though not a named participant in that round; Substack $100M (2025-07); OpenEvidence (2025-10); AlphaSense and KoBold Metals in the portfolio (Forbes). Zero frontier-lab checks found. She spends where her deck says value accrues — the exact position the "data is the moat" grade now scores against.
• Time — the deck, once: 340 pages in May 2025, no sequel in 16 months. In February 2026 she was playing the AT&T Pebble Beach Pro-Am with Keegan Bradley — leading the pro-am after round one (Golf Channel, 2026-02) and reported as its winners by Golf News Net — a data point, not a slight: the 1995–2019 annual cadence has not returned, so do not expect her to update through a drawdown in real time.
• Seats and funds: boards of Block, Genies, OpenEvidence (Forbes). No new BOND fund found; Fund III ($2.5B, 2022) is four years old — a raise is the normal next event and would make the next deck its collateral.
• Silence on the tell: no Meeker statement on capex sustainability, valuations, or the July 2026 capex sell-off was found. The 2001 test (does she downgrade this time?) has not yet been administered.

Feels

Wants to be right about the big curve again — and to be seen as the one who charted it, not the one who cheered through the crash. Fortune's 2001 line ("couldn't downgrade stocks she fundamentally believed in") is what the 2025 deck's bear charts are quietly answering.

Hears

Hyperscaler IR decks, portfolio dashboards (WAU, ARR, developer counts), NVIDIA and OpenAI usage disclosures, sell-side capex tallies, a small BOND research team. Little academic economics, almost no labor or safety literature — the input diet is corporate telemetry.

Sees

Private usage and revenue curves for ~130 companies and 51 unicorns before they are public, plus growth-stage deal flow in two AI-infra names (Baseten, ElevenLabs) at the moment demand shows up. She sees monetization before the market does; she does not see the worker removed from the denominator.

Incentive map

Re-derived this cycle (diffed against the --investing twin, 2026-07-19): who pays — LPs of three BOND funds (~$5.9B) long the AI application and tooling layer; a Fund IV raise is overdue and unannounced. What she's selling — the deck is the fund's free flagship marketing and deal-flow magnet, so its optimism is structurally load-bearing. Board seats — Block, Genies, OpenEvidence (new to this file; the twin did not carry them) — tie her to specific AI-application valuations. What she can't say — "AI capex is unsustainable" (her portfolio is downstream of it), "valuations are stretched" (mid-raise), "OpenAI is losing" (the consumer distribution her portfolio rides). Independence — no bench co-investment or co-authorship found. Rule-25 discount: high on forward adoption or monetization claims; moderate on her historical S-curve data, which is auditable. The twin's read holds: confirming indicator, never risk manager. 1999's research bent toward banking fees; 2025's is published by a fund long AI — same gravity, different fee.

Theories aligned with

• Exponential curves — her entire methodology; S-curves and adoption data are the backbone
• Augmentation thesis — loosely; her enterprise data shows AI augmenting workers, though she doesn't commit to augmentation-over-substitution as a normative stance. 2026: Canaries finds employment falling where AI substitutes, flat-to-rising where it complements — the split she never makes is now measurable
• Adjacent to great stagnation — her data could be read as showing the stagnation breaking; she implies it without engaging the academic debate. The KC Fed's "AI explains little" is that debate arriving
• Jevons' paradox — consistent with her "cheaper capability expands the market" framing; the −99.7% inference-cost chart is her Jevons exhibit
• Partially adjacent to techno-optimism — data-grounded variant; she lets the charts carry the optimism rather than leading with ideology

What she's reacting against

• Narrative-driven tech analysis that substitutes anecdotes for data — she leads with charts, not vibes
• Dismissal of AI as "just hype" by people who haven't looked at the adoption curves (faster than mobile, faster than internet)
• Doomer framing that ignores the measurable productivity gains already appearing in enterprise deployments
• Slow-diffusion assumptions carried over from prior GPT waves — her data suggests AI adoption velocity is genuinely different
• Pure venture/startup framing that misses the macro-economic restructuring underway — she reads the economy, not just the cap table
• NEW (2025): US complacency on China and immigration — "America has to be a place where the brightest people in the world want to come" (Axios, 2025-05-30)

Where she overlaps / splits (with Rich)

• Overlaps with Benedict Evans on "AI is a real platform shift" and the market-structure lens — both chart-heavy ex-analysts. Splits: Evans is agnostic on timing ("nobody knows yet"); Meeker commits to faster timelines. Evans maps markets; Meeker maps macro-economics. 2026: Evans' "exposure scoring is fooling yourself" was contradicted by the Canaries margin; her usage-first method was not.
• Overlaps with Erik Brynjolfsson on enterprise productivity evidence. Splits: he theorizes the J-curve lag and augmentation-vs-substitution design; she stays closer to the data and is less prescriptive. 2026: his Canaries payroll data is the distributional layer her deck lacks — complements, not rivals.
• Overlaps with Peter Diamandis on exponential framing and directional optimism. Splits: Diamandis extrapolates from curves; Meeker reports them.
• Overlaps with Tyler Cowen on taking the macro view seriously. Splits: Cowen emphasizes the high-agency dividend and gradual diffusion; BTOS's 18% firm adoption currently favors Cowen's speed at the firm level and hers at the worker level.
• Splits with Daron Acemoglu on net impact — he says measured AI productivity gains are small; her enterprise data disagrees. 2026: KC Fed attribution and the Atlanta Fed's 0.4–0.8% sit closer to his number, so far.
• Splits with Marc Andreessen on register — both bullish; she is chart-driven and measured where he is ideological and combative.
• Partial overlap with Sam Altman on transformation magnitude. Splits: Altman forecasts from capability roadmaps; Meeker measures from adoption. She's the trailing indicator to his leading one. 2026: her "OpenAI shows no signs of falling behind" drifted as Anthropic passed OpenAI on run-rate — the trailing indicator was, for once, late.
• Converges unexpectedly with Ilya Sutskever — his 2025-11 "benchmarks without economic impact" is her usage-over-benchmarks method restated by the scaling camp's founder.
• Investing-side cross-refs (from the twin): Aschenbrenner — leading indicator to her confirming one; her usage data is the falsifier for his demand assumptions. Grantham — his 2-sigma discipline is the module her method lacks; her 2001 record is his Exhibit A. Zitron — same revenue-vs-burn numbers, read as terminal where she reads early-curve; 2026 model-builder revenue cuts against him more than her.

Track record

• Internet Trends Report (1995–2019): the most-cited annual technology deck in the industry — framing and data selection consistently strong, if directionally bullish
• Morgan Stanley internet coverage (1990s): correctly identified Amazon, eBay, Google and others as platform winners — but bullish through the dot-com bubble. Fortune's "Where Mary Meeker Went Wrong" (Peter Elkind, 2001-05-14): never downgraded Priceline, Amazon, Yahoo or FreeMarkets through 85–97% declines; kept a buy on Priceline from $134 to under $3; named among analysts paid partly by the companies they covered. Directionally right on the internet, catastrophically wrong on valuation and timing
• Mobile transition calls (2008–2014): timing and magnitude largely right; her "mobile is eating the world" data predated Benedict Evans' similar framing
• Kleiner Perkins (2010–2018): Uber Series B at a ~$330M valuation (~100x, per the twin), Facebook, Spotify, Ring
• BOND (2019–): Fund I $1.25B (2019), Fund II $2.0B (2021), Fund III $2.5B (2022); ~$5.9B AUM; ~176 investments, 51 unicorns; Canva from first check (2019) to $42B (2025-08). Returns not public; the analyst-to-investor shift means her public output carries an asset-manager conflict
• 2025 deck, first-year scorecard: consumer adoption, capex trajectory, inference-cost collapse and China open-source Held; enterprise-adoption speed, productivity-in-financials and "OpenAI not falling behind" Drifted; "data is the moat, not the model builders" Wrong on 2026 revenue
• Pattern: directionally right on big platform shifts; strong at mapping adoption; weaker at timing corrections and downside — and, new this cycle, weaker at calling which layer captures the value

Empirical vs normative

• Empirical: adoption curves, usage data, revenue-per-employee metrics, GDP productivity data — her core register. She lets charts make the argument, with minimal editorializing.
• Normative: light — she implies that faster adoption and productivity gains are positive, but rarely prescribes policy. Does not engage deeply with redistribution, safety, or labor-protection policy. 2025 exceptions: immigration and the China "space race."
• Bracketed: x-risk, alignment, distributional justice, permanent underclass dynamics — almost entirely absent from her analysis. Not denied, just outside her lens.
• Conflict-of-interest note (CONVENTIONS rule 25): BOND is an active AI investor; her public analysis trends bullish on AI adoption, which aligns with her portfolio thesis. Discount accordingly — see the incentive map for this cycle's re-derivation.

Weak spots / open questions

• The analyst-to-investor transition creates a structural bullish bias — she now has financial exposure to the narrative her data supports
• Historically weak on calling downside risks and corrections — right on the internet, wrong on timing and valuations. Live test: the July 2026 capex sell-off passed without a word from her
• Enterprise productivity data is cherry-picked by definition — she presents the best case studies, not the median deployment. Goldman's 11%/2% is the median
• Almost zero engagement with distributional questions: who captures the gains? Her data shows company-level efficiency but says little about whether workers or shareholders benefit — labor's share at a post-1947 low answers it for now
• No engagement with x-risk, alignment, or safety — a blind spot relative to the roster's Bengio, Russell, Amodei cluster
• The S-curve framing assumes AI follows prior platform shifts — if AI is qualitatively different (agentic, recursive), the pattern-match could be Evans' "this time is the same" trap in reverse
• Open: does her macro data predict labor outcomes, or only firm outcomes? Revenue-per-employee rising doesn't tell you what happens to the employees who aren't there anymore. Partly answered 2026: Canaries shows the hiring margin, not the firm, is where it lands
• Open: how does "data is the moat" survive model-builder run-rates of $65B–$100B? v1's "[TBD: watch Bond portfolio positioning]" resolved — BOND stayed at the tooling/app layer through 2026 with no frontier-lab check; she is betting her thesis against the data. The next deck's treatment of model-builder economics is the tell
• Open: "unprecedented" appears 51 times in the 2025 deck — when everything is unprecedented, base rates get discarded (twin's observation; stands)
• Open: is she still publishing? A one-off 2025 deck is a different instrument from the 1995–2019 series; without a second edition the file grades a snapshot

Rich's take

• Meeker is the macro-data backbone for the adoption side of the Keys thesis — her S-curves are the empirical evidence that humanity is handing over the keys faster than any prior technology transition. When Rich says "the handover is accelerating," Meeker's charts are the receipts.
• But her blind spot is Rich's central concern: she maps the aggregate curve without asking what happens to the people who fall off it. The permanent underclass question is invisible in her data because her unit of analysis is the firm, not the worker. Revenue-per-employee rising is a great chart — unless you're the employee who was removed from the denominator.
• Her lens is useful precisely because it's limited: she tells you what's happening (adoption velocity, enterprise productivity, capital flows) without telling you what it means for human agency. That's Rich's job — to take her data and ask the livelihood-sovereignty question she never will.
• Pair her with Acemoglu and Autor for the distributional counter-read. Meeker shows the speed; they show who it hits.

converts-from: people/mary-meeker.md · schema v1 · AI & Society domain — the four bullets above are Rich's, carried verbatim from v1.

2026-09-21 validation note (not Rich's words): the "removed from the denominator" line now has market-level receipts — output up, hours flat, labor share at a post-1947 low (Indeed, 2026-08-06) and a −19% entry-level hiring gap (Stanford, 2026-08-12). Her chart and Rich's question are the same dataset read from opposite ends.

Delta log

2026-09-21 — v1→v2 migration + validation (bulk run)

• Grades: 10 Held / 5 Drifted / 1 Wrong (+2 live predictions with dated checks, 2 unfalsifiable items noted). Window 2026-06-24 → 2026-09-21; per the batch rule the June pass is treated as never having run.
• Wrong: "data is the new moat — not necessarily the model builders" — Anthropic run-rate $65B (Bloomberg, 2026-08-17), pacing $100B+ (Axios, 2026-09-18); ~half of S&P 500 earnings growth to AI-infra suppliers (Goldman via TechTimes, 2026-09-09, linked in claim 5). Mechanism Held; valence inverted.
• Drifted (each cited inline above): enterprise adoption "similarly compressed" (Census BTOS, 2026-05); productivity "visible in company metrics" (Goldman, 2026-08-22); GDP restructuring, on attribution (KC Fed, 2026-02-11); labor transition "manageable" (Stanford Canaries, 2026-08-12; Indeed, 2026-08-06); "OpenAI no signs of falling behind" (Bloomberg / Axios).
• Held (cited inline): consumer S-curve (900M WAU, 2026-02-27); platform shift (CNBC capex, 2026-02-06); usage > benchmarks (Sutskever / Dwarkesh, 2025-11); ads restructuring (Axios, 2026-05-05); capex trajectory (CNBC, 2026-07-28); inference-cost direction and China open-source (Data Gravity, secondary); three historical calls.
• Most surprising delta: the value-capture claim inverted while her portfolio kept betting on it — BOND added ElevenLabs (2026-02-04) and stayed clear of frontier labs in the year the model builders became the fastest-growing revenue lines in software. Says and Does agree with each other and disagree with the data.
• Employer/role check (mandatory, dated search run 2026-09-21): no change — Founder & GP, BOND (Forbes, Midas 2026). No Fund IV Form D or close found (two searches). No 2026 Trends deck found (three searches; bondcap.com still "May '25"); the twin's "[UNVERIFIED] 2026 deck" item resolves to none as of today.
• Incentive re-check vs the --investing twin (2026-07-19): agrees on fund facts, the Fund IV gap and no frontier exposure; adds board seats (Block, Genies, OpenEvidence) and the 2026 ElevenLabs check. Discount raised to high on forward monetization claims.
• Post-review dated sources: no Meeker statement (interview, post, deck) after 2025-06 was found. Third-party in window: Bloomberg newsletter on her reports (2026-02-01; not fetched, title only), Forbes Midas 2026, Golf Channel (2026-02), ElevenLabs (2026-02-04), Baseten (2026-06-22).
• v1 TBDs resolved: "AI Trends 2023–present" → no such editions, first deck since 2019 was May 2025; ads-prediction attribution → not found in the deck, graded on the observable; "watch Bond portfolio positioning" → stayed at the tooling/app layer; New Yorker (Cassidy, 1999) date still unverified.
• Tier: semiannual — one artifact a year at most, no 2026 deck, fund events rare; wake triggers cover the deck, the raise, a capex cut, a drawdown (the 2001 test) and a frontier-lab check. Confidence unchanged at medium: strong market-level data, almost no fresh words from her.

Sources

• Primary (her own): BOND — Trends – Artificial Intelligence (May '25) · 340-page PDF · Axios Q&A (2025-05-30; quotes via index, page not fetched) · BOND partner page
• Deck digests used for numbers: Ken Yeung — 10 charts (2025-06-06) · The Batch (2025-06) · SaaStr (2025-06) · PitchBook — "space race" (2025; title only)
• Role / fund / portfolio: Forbes profile (Midas 2026) · ElevenLabs Series D (2026-02-04) · Baseten Series F (2026-06-22) · Golf Channel (2026-02) · AltAssets — Fund II close (2021)
• Market-level graders (each linked inline at the claim it grades): FEDS Note 2026-04-03 · Census BTOS 2026-05 · KC Fed 2026-02-11 · BLS Productivity Q2 2026 · Indeed Hiring Lab 2026-08-06 · Stanford Canaries 2026-08-12 · Goldman via Yahoo 2026-08-22 · TechTimes 2026-09-09 · CNBC capex 2026-02-06 and 2026-07-28 · Bloomberg Anthropic $65B 2026-08-17 · Axios $100B pace 2026-09-18 · Search Engine Land 900M WAU 2026-02-27 · OpenAI "Testing ads in ChatGPT" · Axios self-serve ads 2026-05-05 · Data Gravity open-weight takeover 2026 (secondary) · Dwarkesh–Sutskever 2025-11
• Historical: Fortune — "Where Mary Meeker Went Wrong" (2001-05-14) · archived Internet Trends decks 1995–2019 (Morgan Stanley / KPCB; Code Conference 2013–2019) · The New Yorker (John Cassidy, 1999; date not verified) · Books: none — her medium is the deck
• Source tier: analyst-report / investor-presentation level — data-rich, not peer-reviewed; weight as high-quality empirical mapping with commercial incentives attached. Two 403-blocked pages (Axios 2025, PitchBook 2025) cited for title and indexed quotes only.

migrated v1→v2 2026-09-21 · backup: _archives/mary-meeker.html.bak-20260921 · converts-from: people/mary-meeker.md · AI & Society domain · investing twin: mary-meeker--investing.html (read, not modified)