Elena Burger

1. Key Themes

Consumer AI is a power-user game: usage is broad, payment is razor-thin and concentrated

Half of Americans use AI, but only a sliver pay, and a tiny fraction of those carry the market. Olivia Moore laid out the distribution: "About half of Americans report using AI. Around 4.5% of US consumers are paying a subscription to an AI product." 00:00:14 She went deeper on concentration: "even within that top 4.5% of people who do spend, the top 10% of those are like more than half of the revenue. And the top 1% are 20% of the revenue, whereas the bottom 50% are like 16%" 00:15:31. The top 1% spends "$903 per month personally on their personal credit cards on AI. And the median is spending $25." 00:00:14 Her framing: "the top 10% of users are driving this market." 00:16:00

Spend data reveals a different market than traffic data

Adding card-spend data (via YIPIT) reshuffled the rankings. Moore: "of those 50 that ranked for spend, 29 of them were not on either traffic list." 00:03:18 Traffic is "settling" with only 11 new products, "the least we've had in any of the prior editions" 00:02:51, but spend surfaces a different cohort. The top spenders skew to builder tools: "mostly developer tools, productivity tools, creative tools, things used to like build, make, sell. Things like N8N, Granola, Higgs Field, Manus. Those all were way overrepresented amongst the top spenders." 00:16:25

The shift from AI-as-tool to AI-as-agent: personal assistants are the newest breakout

Josh Ellman described the inflection: "AI went from something that we were using as a tool to enhance our productivity... to something that would actually get things done for us. And I think OpenClaw was an incredible pioneer to do this." 00:05:20 Moore named the leading consumer assistants: "the biggest developments have been Muse and Instinct." 00:06:32 Instinct "had announced like 100,000 users growing 10% day over day... 40% of users connected a credit card in the first three weeks and they were spending over $1,000 on average in their first month." 00:06:32 Muse hit "500,000 downloads and 250,000 active users in the first 12 days" 00:06:56, though versus Threads it "still pales in comparison": roughly 5 million vs. 16 million US/Canada downloads in the first 22 days 00:07:45. The category exploded in the "last three weeks to a month" 00:05:07, so most of it isn't on the ranks yet.

Agent usage is still dominated by coding and technical automation, which explains the cost-to-serve puzzle

When founders quote hundreds or thousands of dollars per month to serve a user, Moore explains the mix is skewed to technicals. David Paulon's assistant benchmark tracks "over 170 agents" and hosts a 1,500+ user community. "The vast majority still the number one use case was coding and technical automation, even for these consumer assistant products." 00:13:06 Mainstream-oriented agents "are seeing costs in the tens of dollars, not the thousands of dollars a month." 00:13:06 Ellman on the mainstreaming gap: "when you just hand this to somebody new, it feels like a blank box." 00:13:41

Business model inversion: subscriptions dominate AI, but history says ads and transactions win consumer scale

Moore: "85% or something of our web list monetized via subscriptions and other 62% monetized via kind of credits or token extra usage payments. Only like 13% had ads." 00:21:06 She argues this is "a pretty unnatural inversion" versus consumer internet history, where big companies earn from ads or transaction fees. Her conclusion: "We have transcended the need for everyone to buy a subscription to AI and we need to see these other business models come back." 00:22:28 High COGS prevents the old playbook: "if a company was making money in the first five years, it was like, whoa, what's happening? The rule was like, let's build density of users... because they cost near zero to serve." 00:23:04 Ellman: lower inference costs are the unlock for ads and transaction fees to "finally start working." 00:22:34

OpenAI's ad business hit scale unusually fast

Moore: "They're already at, you know, a billion, billion in annualized run rate. Like that would previously take a company years and years and years to get to even after launching an ads product." 00:24:38 Two drivers: density ("1.2 billion weekly active users") and richer targeting ("hypothetically, they should be able to charge more and have higher conversion rates because they just know so much more about you"), plus products like login with ChatGPT and a coming wallet 00:25:06. Ellman reports the rollout feels native: "very clearly labeled and it actually feels like a natural add to the conversation." 00:26:17

The frontier labs are diverging in monetization and user base

ChatGPT leads: "ahead of Claude around 6x on the web, around 2x ahead of Gemini on the web" 00:30:15. But Moore flagged a surprise: "Claude has actually passed Gemini in terms of number of paid subscribers, which is kind of crazy given Gemini has a much bigger install base." 00:30:42 ChatGPT has "around three times more paid consumer subscribers in the U.S. than both Gemini and Claude," but Anthropic, with its "no ads" stance, has "around 7.5% of subscribers on the $100 plus per month... max plan. And that's like 1% for both ChatGPT and Gemini." 00:31:34

Value is migrating back to the software/experience layer, not just the model

Ellman: "the value really now is moving back to the software layer. That now that we have these incredible models... you can build really rich products." 00:42:30 He rejects the "wrapper" framing: "it is much more than just a shim on the model or a harness... It's actually a really rich product and an experience that happens to now use a model." 00:00:31 Moore's moat example is accumulated personal context: Town "has built these playbooks of who I am, what I sound like, that no other product has, and that I cannot easily take and migrate over." 00:44:04

Specialized creative tools survive alongside the labs where taste, workflow, and legal friction matter

Audio is where labs have left room: "11 Labs and Suno are now very reliably near the top of our traffic lists and are also ranked very highly in our spend list." 00:32:57 Midjourney fell off the traffic rankings but "if you look at the revenue rankings, MidJourney is back. So people, the power users, still want the model that they feel has taste." 00:35:15 Video is a "wild frontier" where Chinese companies "can train on any data and do have a real advantage." 00:35:36

Whitespace: the "spend time" categories and network products remain nearly empty

Moore: "dating is one we haven't seen... Recruiting is another one... Social AI, we really haven't seen anything take off." 00:46:52 Also open: shopping, home buying, gaming, entertainment. Moore's lens: "Almost everything we've seen in consumer AI is save time. And that spend time, those end up often actually being amongst the biggest companies, if not the biggest ones." 00:49:22

2. Contrarian Perspectives

A consumer AI investor does not want the paying base to expand

Moore: "I actually don't necessarily want to see that 4.5% of people paying for AI products directly expand, which is maybe controversial because I'm like invested in consumer AI products and I'm a consumer AI maximalist." 00:20:44 Her reasoning: "most people don't have the funds or don't want to spend their funds on software." 00:21:34 The real prize is ad- and transaction-funded AI that reaches the other 95%, mirroring how every large consumer internet company made its money.

Most people don't want to save time; they want to spend it

Quoting portfolio CEO Eugenia of Wabi, Moore: "most people aren't looking to save time. They're looking for ways to spend their time. Mm-hmm. Like, this is why social media, entertainment, Netflix, TikTok, all of these, YouTube are so... the most used consumer products." 00:18:32 She adds that "how do I do this thing a little bit faster... is not an incredibly compelling daily or hourly active value proposition for most people." 00:18:54 This implies the productivity-heavy ranking list is structurally capped, and the biggest AI consumer outcomes will look like entertainment and social.

Incumbents with distribution can't or won't reinvent their own surfaces, and even the labs struggle outside their chat box

Moore: "incumbents being unwilling or unable to cannibalize their existing interfaces. The classic example of this is Google. Like we haven't really seen them reinvent docs or Gmail or calendar for the AI age." 00:38:57 Extending to OpenAI and Anthropic: "they have been much less successful in anything that is not a launch within their existing ChatGPT or Codex or Claude or Claude Code interfaces." 00:39:19 The labs-eat-everything thesis is weaker than assumed.

Anthropic gained consumer paid share without an image/video model, and "Claude overtook Gemini" in payers despite Google's distribution

Moore expected distribution to dominate: Gemini "has a much bigger install base and user base overall. And it also has like the natural distribution mechanism of something like a Google," yet Claude has more US paid subscribers 00:30:42. Elena Burger added that Anthropic "kind of famously abstained" from image/video models but "now people are making images and videos because coding is so good at creating like images and videos." 00:32:10 Capability in code can substitute for modality-specific models.

More personal usefulness may limit, not enable, network effects for assistants

Moore: "I still don't feel like any of the assistant products have unlocked person to person network effects... the more useful the product is to you as the consumer, the more it knows about you. And then do you really want to introduce that into other contexts with other people?" 00:09:41 The very thing that makes agents sticky (intimacy) works against social virality.

3. Companies Identified

Instinct

Consumer personal-assistant agent product. Called one of the two biggest developments in consumer assistants. Moore: "Instinct had announced like 100,000 users growing 10% day over day. They'd announced 40% of users connected a credit card in the first three weeks and they were spending over $1,000 on average in their first month." 00:06:32

Muse

Consumer AI assistant (described alongside a Meta comparison, with US/Canada-only availability). Moore: "very successful launch, especially within the tech community. I think the number was like 500,000 downloads and 250,000 active users in the first 12 days or so." 00:06:56 Ellman noted it already has "hundreds of other partnerships" including Shopify 00:09:39, while Amazon declined to let Muse browse and purchase on its site.

OpenClaw

Open-source agent pioneer that sparked the personal-agent wave. Ellman: "OpenClaw was an incredible pioneer to do this... We were seeing people buying Mac minis. We were seeing people run all their own data with their own local OpenClaw." 00:05:20

ChatGPT / OpenAI

Dominant consumer AI product by usage and revenue. Moore: "ChatGPT is the dominant global user by both usage and revenue on the consumer side. That continues to be true." 00:30:15 Ads at "a billion, billion in annualized run rate" 00:24:38 and "1.2 billion weekly active users" 00:25:06. Also credited with strong image work: "Images 2.0 is very, very good, both for image creation and editing." 00:34:45 Moore notes ChatGPT is already among "the top 20 consumer subscription products globally... in like three and a half years of growth." 00:22:03

Anthropic / Claude

Passed Gemini in US paid subscribers and runs a no-ads, high-ARPU model. Moore: "Claude has actually passed Gemini in terms of number of paid subscribers" 00:30:42; "around 7.5% of subscribers on the $100 plus per month... max plan." 00:31:34 Success attributed to launches "like Claude Design" and press.

Google Gemini

Large install base, 2x behind ChatGPT on web, but behind Claude in US paid subs. Image models ("Nano Banana") have "taken away a lot of the high-level consumer traffic that was previously going to these standalone image generators." 00:34:45

Town

Prosumer agent connecting Slack, email, Google Drive. Moore's example of context-driven lock-in: "Town has built these playbooks of who I am, what I sound like, that no other product has, and that I cannot easily take and migrate over." 00:44:04 Also credited with "fantastic agents." 00:04:39

Polk

Early pioneer of consumer assistants. Moore: "Polk was a very early pioneer here." 00:06:32 (Transcription of the name is uncertain.)

Tomo

Consumer assistant with a large audience. Moore: "Tomo has a very big audience and is doing incredibly well." 00:06:32

Suno

AI music generation. Moore: "Suno has been able to kind of run away with a really strong lead" because labs won't deal with "all of the IP headaches" 00:33:26. Ellman: "it speaks to you from the moment you start using it through the end." 00:33:43

ElevenLabs

AI voice/audio. Near the top of traffic and spend lists 00:33:26. Ellman: "if you need voice and audio within the product that you're making, it just has a way of working and interacting with it that feels completely different." 00:34:08

Midjourney

Image generator that dropped off traffic rankings but returned in revenue rankings because power users want "the model that they feel has taste, has an aesthetic sense, and that they're able to tune in a really specific way." 00:35:15

Granola

AI notetaker. Overrepresented among top spenders 00:16:25; cited as an example of PLG-to-enterprise expansion 00:38:30 and as a case of products incumbents haven't reinvented 00:38:57.

Whisper Flow

Voice-dictation product. Cited as a PLG-to-enterprise success: "they're very successful in growing via PLG, whisper flow and granola, another great examples there." 00:38:30

Superhuman

Email product cited as an example of startups winning where incumbents won't cannibalize their interfaces 00:38:57.

N8N

Automation tool, "way overrepresented amongst the top spenders." 00:16:55

Higgsfield

AI creative/video tool, listed among tools overrepresented in top-spender data 00:16:25. (Transcribed as "Higgs Field.")

Manus

Agent product also overrepresented among top spenders 00:16:25.

Plaud

AI note-taker hardware plus subscription ("Plod" in the transcript). Moore: "an AI note taker device that started kind of in China in the East and now has come to the U.S. And you buy the device and you can also buy a subscription." 00:41:31 Cited as an example where OpenAI could build hardware but "everything takes longer" at a big company.

Wabi

Portfolio company run by Eugenia; source of the "spend time not save time" insight 00:18:32.

OpenEvidence

AI product for doctors, referenced as a verticalized ad-supported success (the transcript renders it as "OpenAI"). Moore: "it already has 50, I think the most recent numbers were 50 to 60 percent density of all U.S. physicians on the product. And so when you have that kind of density of a really valuable audience... you can... advertise really effectively." 00:29:17

Assistant Benchmark

David Paulon's tracking site for 170+ agents, with a 1,500+ user community that supplied usage data to the report 00:12:36.

Canva

Reference point for enterprise expansion speed: "Canva took six plus years to have real team and enterprise revenue." 00:38:05

Replit, Gamma, CapCut-type consumer-to-enterprise tools

Named as examples of consumer companies becoming enterprise companies quickly: "things like CREA, you know, Replit, Eleven Labs, Gamma." 00:38:05 (Transcribed "CREA", likely a Krea-style product; included as stated.)

Figma

Cited as a design tool "so designed to use deeply for the thing you're trying to express yourself. And it speaks to designers" 00:34:08.

TypeSafe

Referenced by Ellman for enabling developers "to build things in even new ways": "Jeff from TypeSafe that allows even developers to build things in even new ways." 00:42:30

Shopify

Muse partnership enabling agent purchasing 00:09:39.

Amazon

Blocking agent browsing/purchasing (Muse), and appears on the top-20 consumer subscriptions list 00:22:03.

Meta Threads

Used as the benchmark launch (16M US/Canada downloads in ~22 days) and a critical-mass network-effect example 00:07:45.

Netflix, TikTok, YouTube

The "spend time" benchmarks: "the most used consumer products that we have." 00:18:54

Uber

Listed among top consumer subscription products with frequent purchase behavior 00:22:03.

Tomo/Polk, Astra, and 5.5-class models

Elena Burger and Moore describe agent-driven creative: "I have recently been asking Astra and other really powerful models to generate or create... creative for me... and it will go tool call to elsewhere on the internet." 00:36:06

YIPIT

Data provider whose consumer card panel underlies the spend analysis (transcribed "Abit"/"YIPIT"). Moore: "we worked with the Abit data to get specifically consumer card spend data." 00:02:51

4. People Identified

Olivia Moore

a16z partner and author of the Top 100 Consumer AI Apps series. Her central thesis is that payment is concentrated and ads/transactions must arrive. "We have transcended the need for everyone to buy a subscription to AI and we need to see these other business models come back." 00:22:28

Josh Ellman

Recently joined a16z's consumer team ("a couple months ago"). Thesis: value is returning to the software layer. "the value really now is moving back to the software layer." 00:42:30

a16z host. Sharp framing on developer mindset: "developers and coders are people who are naturally attuned to sort of ask, how do I get leverage over this thing?" 00:17:47

Sarah Wang

a16z colleague who observed the power-law dynamic across enterprise and consumer. Burger: "she said, hey, there is such a power law game happening now. It's happening at the enterprise level and it's happening at the consumer level." 00:03:37

David Paulon

Runs Assistant Benchmark, tracking 170+ agents and hosting a 1,500+ user community. Moore: "who runs assistant benchmark, who's fantastic." 00:12:28

Eugenia (Wabi CEO)

a16z portfolio CEO; source of the "spend time, not save time" quote 00:18:32.

Steven Sinofsky

a16z partner voice in the intro/outro; noted for the summary framing of the report's findings (including the "$900 a month" top 1%) 00:01:25.

5. Operating Insights

Build meters on usage before you grow, or growth can bankrupt you

Ellman: "companies get afraid of growing too fast. Like they actually, if they grow too fast and they don't have meters on how they're charging for it, you can actually have out of control spend that can be very, very hard to manage." 00:23:32 Moore adds the Muse example: slower rollout partly reflects "a much higher cost to serve these users and they don't want it to get too big too fast." 00:08:03 Treat cost-to-serve as a growth throttle and design credits/usage tiers from day one.

Use PLG on work-adjacent personal tools, then add enterprise features after pull

Moore's pattern: Whisper Flow and Granola grow via personal use, "get pulled into enterprises and then the founders are like, oh, we got to add privacy and security and team plan. But it's very effective, very low cost distribution." 00:38:30 Sequence matters: win the individual first, build team/security features reactively.

Compound personal context into switching costs

Moore's quantification of why context matters: "the difference between an email that is 99.9% sounds like me and even an email that's 85% sounds like me is the difference between spending like 10 seconds to fix it and like 10 plus minutes." 00:44:30 Product implication: capture playbooks, voice, and preferences so quality compounds and migration is painful.

Make products agent-accessible

Moore: "creative tools that are agent-friendly and agent-accessible are going to increasingly see really compelling tailwinds." 00:36:06 Expose clean tool-call surfaces (APIs/CLI/install-able skills) because agents are becoming a distribution channel.

Package what tinkerers discover; mainstream users need seeded use cases

Ellman: "this is the tinkerers, the people at the edge are figuring out these capabilities and we're going to build products over the next year... they're going to take all of this complexity of what you can do and package it up." 00:36:51 Related: mainstream users face a "blank box," and personal AI lacks the social-sharing loop that spreads ideas, so productize starter use cases and shareable outputs 00:14:10.

6. Overlooked Insights

Amazon blocking Muse versus Muse's long-tail partnerships hints at a coming agent-access war

Dropped mid-conversation: "Amazon is like, we're not going to allow Muse to kind of browse and purchase products on our site. But also Muse has already hundreds of other partnerships with different products." 00:09:12 The largest retailer refusing agent access, while Shopify's long tail embraces it, suggests agent commerce may route around Amazon and strengthen the open Shopify ecosystem.

The "opposite of mainstream" cost structure: assistants costing $1,000+/month per user is a signal of who is actually using them

The $1,000+ first-month spend at Instinct 00:06:56 and thousands in cost-to-serve 00:12:28 were explained as a coding/automation user mix, not a mainstream use pattern. The overlooked implication: current assistant growth metrics (users, card connections) are a power-user cohort proxy, so mainstream unit economics ("tens of dollars") are a different, unproven business 00:13:32.