This week’s Subscribed Weekly column comes courtesy of Amy Konary, Senior Vice President of the Subscribed Institute at Zuora.
Everything old is new again.
A fascinating new paper from Dan McCarthy and colleagues tracks mobile app data for seven AI assistants from 2023–2025. Its headline finding: every major model launch expands the entire market. When OpenAI releases a new model, Claude grows. When Anthropic ships, ChatGPT benefits.
The pie is getting bigger for everyone. What the aggregate numbers don’t show: three companies, the same rising tide, and a roughly 40x gap in revenue per user.
What is going on here? McCarthy’s data reveals three distinct bets, each echoing patterns from the ghosts of SaaS plays past.
ChatGPT is Evernote.
ChatGPT is running a scale play. With roughly 79% market share and 390 million daily users, it dominates on reach. But over 90% of its users don’t pay. Unsurprisingly, OpenAI has introduced lower-priced tiers and is testing ads — classic moves for converting a free base. Scale works, until it trains customers not to pay. If you’ve watched any SaaS company struggle to convert a massive free tier into paying customers, this should feel very familiar.
Gemini (consumer) is Alexa.
Gemini is running an ecosystem play with consumers, designed to keep users inside Google’s ecosystem, where advertising carries the economics. Ecosystem plays have an Achilles heel: they never learn to stand on their own. Amazon spent a decade and tens of billions of dollars learning that the hard way with Alexa.
The enterprise play is different. Google has embedded basic Gemini features directly into every Workspace plan. Above that floor sits Gemini Enterprise, with 8 million paid seats across more than 2,800 companies. The revenue flows through Google Cloud rather than being labeled “Gemini revenue,” which is why this looks like no monetization from the outside. It’s a classic land-and-expand play: bundle a lite version into the platform to create ubiquity, then upsell the tier where real capability, and margin, lives.
Claude is Adobe.
Claude is running the premium play. It currently has less than 1% market share, but has a ridiculously higher revenue per user—three times ChatGPT, and forty times Gemini. This is the Adobe Creative Cloud lesson. Adobe succeeded in going premium because they had genuine switching costs: proprietary file formats, deeply embedded workflows, years of accumulated skills. Claude’s 40x revenue-per-user advantage tells you something real about the value professional users are getting. The question is whether that value is truly durable, and those users are building the kind of deep integration and workflow dependency that creates prohibitive switching costs.
What Each Company Needs to Get Right
Right now, all three strategies look viable because the market is still expanding. This is Phase 1 of any subscription category: land, acquire, grow. But in this phase, rampant user growth can mask profound structural weaknesses. Companies might be walking dead but not realize it.
Phase 2 is different. Growth slows. Renewal economics take over. And revenue models (not user counts) determine outcomes. WeWork and Blue Apron scaled rapidly on adoption, only to falter when business model fundamentals couldn’t sustain them.
Simple adoption is not validation. Revenue quality is. AI is heading toward that same transition. So here’s the advice I’d give each of these companies:
OpenAI: Your biggest risk isn’t a competitor. It’s your own free tier.
You have 390 million daily users and over 90% of them have never given you a dollar. That’s not a growth story, that’s a monetization gap. Lower-priced tiers can help, but be careful about advertising. You can’t afford to lose trust. Once users start questioning your intentions around privacy, data, and purpose, the conversion math gets worse, not better.
Gemini: Turn distribution into a real customer relationship.
Gemini’s enterprise traction is real, but bundled adoption isn’t the same as earned loyalty. When customers pay for Workspace and get Gemini, they’re not necessarily forming a relationship with Gemini, who is their assigned freshman room-mate! They’re tolerating it, or occasionally using it, inside a platform they already depend on. Today’s low engagement outside enterprise contexts suggests that relationship hasn’t fully formed. Consumer trust will determine whether Gemini ever builds an identity beyond “the AI that comes with Google.”
Anthropic: Your revenue-per-user advantage is real. Now build the moat around it.
Claude’s revenue-per-user advantage signals real product-market fit. The challenge now is defensibility. Premium models endure when value compounds—through workflows, integrations, and accumulated context. Each interaction should deepen the relationship, not deliver a one-off result.
The 40x Gap: We’re All Dating Now
That 40x revenue gap between the three leading AI companies reflects three fundamentally different potential customer relationships, in a time when we’re all dating now.
Claude’s users understand the value and pay for it. Gemini’s base is split—enterprise customers who pay, and consumers who may never engage in a value exchange. ChatGPT sits in the middle: enormous demand, but an unresolved question of what that demand is worth. This is the monetization gap: the distance between the value a company creates and the value it captures.
Companies under pressure often default to extractive tactics—ads, opaque pricing, experiences optimized for capture over value. This is the grim process of platform degradation that the writer Cory Doctorow memorably describes as “enshittification.” It can boost short-term revenue but erode long-term trust. Lots of SaaS companies made this mistake.
AI is approaching the same fork in the road. The companies that succeed will make the value exchange clear, fair, and durable—where customers understand what they’re paying for, and believe it’s worth it.
Because when the pie stops growing, that’s what remains: not scale, not hype, but relationships that compound.
