ICONIQ

Sources#

Summary#

ICONIQ Capital is a private investment firm best known for managing money on behalf of a small set of ultra-high-net-worth families; its venture and growth-equity arm publishes under ICONIQ Growth and, from the September 2026 materials onward, ICONIQ Venture & Growth. It reaches this wiki not as an investor but as a publisher of operating benchmarks — the firm's research team runs periodic surveys of software-company executives and periodic benchmarking studies drawn from portfolio and public-company financials, and those publications supply most of the vault's numbers on AI product margins, pricing models, provider mix, org restructuring, forward-deployed engineering, and revenue per FTE.

Three ICONIQ publications are compiled here, and they are different instruments with different trust profiles — the distinction is load-bearing every time an "ICONIQ says" figure is quoted:

State of AI 2026: The Builder's Economy (2026-07-08) — the third bi-annual executive survey, ~305 respondents at software companies building AI products. Self-reported, and its headline trajectories (AI revenue share, gross margin, revenue per FTE) are mostly forward projections for 2026 and 2027, which are closer to prediction-grade than to measurement. Compiled at AI Product Economics Maturation, AI-Native Organization and AI Investment Story, Not Efficiency Story.
The ICONIQ Pacesetter Index (2026-09-17) — not a survey. Quarterly financial and operating data from 2024 to Q2 2026, drawn from "a select dataset of public software companies and our private venture and growth portfolio companies," benchmarked across seven metrics by four ARR bands. This is receipts, not sentiment — but on a sample ICONIQ partly owns, and deliberately selected (see below).
2026 State of Scaling: The Great Sorting (September 2026, 52 pages) — the annual growth-and-efficiency report the Pacesetter Index is an excerpt of, and the one to cite whenever the question is "compared with what?". Same cohort machinery, wider apparatus: quarterly operating and financial data from 137 software companies (portfolio plus 11 public companies chosen on ICONIQ's own IPO-performance criteria), 2022 – Q2 2026, and — the thing the Index page lacks — an explicit "Others" comparator plotted beside the Pacesetter bar on six metrics. Two small proprietary surveys are bolted on where noted: the Annual Growth Operating Trends Survey (Feb 2026), n = 38 (the source of every AI-adoption figure) and the Enterprise Buyers of AI-powered Software Survey (Jun 2026), n = 132 (every buyer-preference figure). Compiled at AI-Native Startup Lifecycle, AI Investment Story, Not Efficiency Story, AI Product Economics Maturation, Seven Powers Applied to AI, AI-Native Organization, Firm AI-Spend Intensity and Headcount Growth, Product Velocity as Moat, Compounding Data Moat and Build Instead of Buy Under Agentic Coding.

A unit trap that applies to all three and is stated only in a footnote: n is company-quarters, not companies. ICONIQ's own methodology page says it outright — "each datapoint (n) represents a single fiscal quarter of data per company included, not the number of individual companies," and "a given company's quarterly datapoints can be included multiple times in aggregated views." So an n of 76 on a headcount chart may be a dozen companies observed over several quarters, and the Pacesetter cells in the 2026 comparison charts run n = 4 to 12 company-quarters — meaning the report's two most quoted contrasts, "284% vs 124% OpEx" and the "sub-1x burn multiple," each rest on four Pacesetter company-quarters. Never report an ICONIQ n as a sample size.

What the Pacesetter Index is, and what its selection does to every number in it#

The Pacesetter Index is the announced successor to ICONIQ's pre-AI Enterprise Five Scorecard, and ICONIQ is explicit about why it replaced it: the Enterprise Five "measured historical and aggregate performance across the software universe," and ICONIQ's position is that "identifying top performing companies means benchmarking against them, not the broader market." The comparator was deliberately moved from the population to the winners.

The cohort definition differs slightly between the chart and the prose, and the chart's is the operative one:

Chart, verbatim: "Top-quartile revenue growth over the past 3 years, and either AI-Native or AI-Driven."
Body text: "AI-forward companies with revenue growth in at least the top quartile for their scale range."

Same substance — a double selection on (a) top-quartile three-year revenue growth within the ARR band and (b) ICONIQ's own AI-Native/AI-Driven classification.

The full report settles the wording split and adds two details the Index page omits (State of Scaling Methodology, p.5, and "So who are these Pacesetters?", p.26). The two wordings are the same rule: top-quartile growth is measured within each ARR scale bucket (<$10M, $10–25M, $25–100M, $100M+) — the chart's terse phrasing and the body's "for their scale range" describe one criterion, and the compile-time decision to treat them as "same substance" is confirmed rather than revised. The AI-Category leg is also spelled out for the first time: AI-Native = "core product or business model is fundamentally created on foundation models"; AI-Driven = "embedded AI across their core product suite while also developing standalone AI offerings." Still ICONIQ's own taxonomy, still with no rubric for adjudicating a borderline case. The second detail is sharper and is a survivorship warning in the report's own words: the Pacesetters are "A LIVING COHORT" — "companies enter and exit this cohort as they meet or fall below the top quartile growth thresholds." Membership is re-decided each period on the outcome being reported, so a Pacesetter series can never show a Pacesetter decelerating; deceleration is how a company leaves the sample. Cohort firmographics as of Q2 2026 (p.26): 60% AI-Native / 40% AI-Driven, 65% horizontal / 35% vertical, 70% on hybrid or consumption-based pricing, ~2–4 years to $100M on average, median age 6 years from founding.

The consequence, and it is the single most important thing to carry from this page: a Pacesetter median is not a population median. Every figure in the index is the middle of a pre-selected top slice. ~~There is no non-Pacesetter control group anywhere in the publication, which makes the index structurally incapable of answering "are AI companies better than non-AI companies."~~ (superseded 2026-09-22 by 2026 State of Scaling: The Great Sorting — see below; the selection point stands unchanged, the missing-comparator point does not.)

The comparator exists — it was missing from the excerpt, not from the work. The State of Scaling report plots a Pacesetter bar beside an "Others" bar for 2026 on six metrics: net dollar retention, gross-new-ARR mix, gross margin, FCF margin, OpEx margin and burn multiple, each split at <$100M and $100M+. So "compared with what?" now has an answer, and the answer is frequently more interesting than the headline (at $100M+ Pacesetters spend less OpEx than Others, 53% vs 75%, the reverse of the sub-$100M contrast ICONIQ leads with). Three limits keep this from being the control group the vault's AI Investment Story, Not Efficiency Story ladder actually wants, and they are not small:

"Others" is within-sample, not population. It means the non-Pacesetter remainder of ICONIQ's own 137 companies — its portfolio plus 11 public companies themselves selected for top-quartile IPO performance. It is a comparison of winners to other winners.
It is not an AI-vs-non-AI split. "Others" is defined by failing the growth-or-AI test jointly; nothing says which leg failed, so no figure here separates an AI effect from a growth effect.
The Pacesetter cells are n = 4–12 company-quarters against 18–86 for Others, before any of the above.

Two further limits of the same kind:

Top Quartile means the 75th percentile for every metric except Burn Multiple, where it means the 25th (lower is better). Mixing the two columns without checking the metric inverts the reading.
The growth column is where the selection bites hardest: 900% median / 2600% top-quartile YoY growth in the sub-$10M band is what selecting top-quartile three-year growers at tiny revenue arithmetically produces, and it is not transferable to any unselected company.
Figures are rounded to the nearest multiple of 5 where applicable, so small band-to-band differences are inside the rounding.

Conflict of interest#

ICONIQ is a growth-stage venture investor, and the Pacesetter sample "partly is" its book: the source note reads "a select dataset of public software companies and our private venture and growth portfolio companies." Three distinct pressures follow, all of which point the same way:

Selection of the sample — which portfolio companies are in the dataset "where data is available" is ICONIQ's call, unaudited.
Selection of the cohort — the AI-Native / AI-Driven label is ICONIQ's own taxonomy, with no published rubric.
Selection of the story — ICONIQ profits from being seen to back the companies setting the pace, and the page carries its own disclaimer that the research is proprietary, subjective, and not to be relied on for investment decisions.

None of this makes the numbers wrong; the data is closer to receipts than anything a survey produces. It means an ICONIQ benchmark should be read as a well-measured description of a hand-picked set, and never as a market fact.

A practical note on access and aliases#

~~Both reports are lead-form gated.~~ (superseded 2026-09-22.) The Builder's Economy PDF was obtained and parsed; the Pacesetter Index's "Download report" button resolves to href="#" (a client-side lead-capture action, no ungated file URL in the page HTML), so what the vault holds for that page is the web article plus its single benchmark image — 1,216×882px, every cell legible, transcribed and re-verified cell-by-cell at compile time.

The gating is per-page, not per-publisher, and the route around it is worth recording. The State of Scaling landing page is also form-fronted, but the full 52-page PDF is served ungated from ICONIQ's Webflow asset CDN — cdn.prod.website-files.com/…/2026_State_of_Scaling_The_Great_Sorting.pdf — and was fetched and docling-parsed in full. The same shape as the openai.com report route: when an ICONIQ report page looks gated, look for the cdn.prod.website-files.com asset URL in the page HTML before settling for the article. The Pacesetter Index page genuinely has no such URL; State of Scaling does. Since the Index is an excerpt of the report, a caveat that reads "ICONIQ publishes no X" should always be re-checked against the full report before being carried.

Domain alias, for future dedup: the Pacesetter page's og:url and its own "Copy page URL" widget resolve to iconiqcapital.com, not the iconiq.com host actually fetched. A future ingest of the same article under the other host would not dedup against the existing row.

Connections#

AI Product Economics Maturation — the home of the Builder's Economy unit economics: AI revenue share, the projected 45%→53%→59% AI-product gross-margin ladder, pricing-model composition, provider mix (Anthropic 51%→81%), internal AI spend, and the FDE-as-revenue-driver motion
AI Investment Story, Not Efficiency Story — where both ICONIQ instruments land in the vault's revenue-per-head instrument ladder: the survey's forward projection ($272K→$496K by 2027) and the Pacesetter Index's measured $655K/$890K at $100M+ — the same publisher supplying both the weakest and the strongest-instrument reading on one page
AI-Native Organization — the Builder's Economy org half: role-mix shift, function-level headcount reallocation, management layers and spans of control
AI-Native Startup Lifecycle — the Pacesetter Index read as a stage ladder: unit economics by ARR band, and the implied headcount that falls out of the revenue-per-FTE row
Seven Powers Applied to AI — ICONIQ added gross retention to its benchmark specifically because "switching tools has become significantly easier," which is the switching-cost-erosion thesis showing up as a change to the measurement apparatus itself
Anthropic — the subject of the Builder's Economy's sharpest single market datapoint: Anthropic moved from #3 to the #1 model provider (51%→81%) among ~305 AI builders in six months
Telemetry vs. Survey Measurement — ICONIQ is unusual in publishing on both sides of that split within one quarter, which is why its reports must never be pooled into a single "ICONIQ says"; State of Scaling does it inside one document, hanging n=38 and n=132 survey figures off a 137-company operating-data spine with only a footnote to tell them apart
Firm AI-Spend Intensity and Headcount Growth — where State of Scaling's headcount-growth-by-growth-cohort series lands: ICONIQ's gate on hiring is revenue growth, the paper's is AI-spend intensity, and the two produce the same "adopters staff up" direction from different instruments
Product Velocity as Moat / Compounding Data Moat — the two defensibility claims ICONIQ's Pacesetter operators make in their own words ("product velocity is the moat that compounds"; "as AI models become interchangeable, the edge is data and workflows rivals cannot copy"), both practitioner-opinion inside an empirical document
Build Instead of Buy Under Agentic Coding — ICONIQ's public-market read on the same thesis: frontier models "making in-house application development more accessible" is named as a cause of multiple compression across all software sub-sectors

Sources#

State of AI 2026: The Builder's Economy — ICONIQ Growth, State of AI 2026: The Builder's Economy, 2026-07-08 (empirical — a real survey of ~305 executives, but self-reported, and its 2026P/2027P figures are projections). PDF obtained and docling-parsed; most chart values were read from the deck's images in a second pass
The ICONIQ Pacesetter Index — ICONIQ Venture & Growth, The ICONIQ Pacesetter Index, 2026-09-17 (empirical, and the stronger of the first two instruments: quarterly financial and operating data 2024–Q2 2026, not self-report). Source for the Enterprise Five succession, the Pacesetter definition and its chart/prose wording split, the seven-metric × four-band benchmark table, and the COI and gating notes above. That page's report PDF is lead-form gated and was not fetched; the seven-metric table exists only as a page image, transcribed and re-verified cell-by-cell. Now known to be an excerpt of the source below, which is where its missing comparator turned out to live
2026 State of Scaling: The Great Sorting — ICONIQ Venture & Growth, 2026 State of Scaling: The Great Sorting, September 2026 (empirical; 52-page PDF, docling-parsed; month-only publication date recorded as 2026-09-01). Source for the three-instrument framing, the company-quarter n trap, the living-cohort and ARR-bucket clarifications of the Pacesetter definition, the AI-Native/AI-Driven definitions, the Pacesetter firmographics, the "Others" comparator and its three limits, the two bolted-on surveys (n=38 Feb 2026 adoption; n=132 Jun 2026 enterprise buyers), the 11-public-company IPO-performance selection criteria, and the ungated CDN route. Parse notes: four table-shift warnings are the Pacesetter Index's two-row "Median / Top Quartile" header (false positives, verified against pdftotext -layout p.27); the one canary-recall miss is a citation footnote. Every figure carried from a chart was read off the page image in a two-pass and cross-checked against pdftotext -layout. One real prose weld: a CFO quote's "~5-10% of payroll" broke across a line and docling welded it to "~510% of payroll" — a plausible-looking nonsense number that no table check can see