Have you ever opened your banking app, scrolled past a cluster of tiny monthly charges, and thought, wait, am I still paying for that? Most people have. The average adult in the United States still leaks more than two hundred dollars a year on software and memberships that barely get opened. That quiet leak used to be a feature of the digital economy. Now an AI agent designed to hunt those leaks is sitting on the table, and some analysts think it could take a bite out of Apple’s most prized cash engine.
Why Muse Suddenly Matters For Apple’s Services Machine
Apple’s hardware still gets the headlines. Phones, watches, headsets. Fine. The valuation story, though, has leaned harder and harder on services. Recurring billing. App commissions. Subscriptions that renew while nobody is looking. Annuity-like cash is what keeps a multiple fat when unit growth cools. I’ve found that investors forgive a lot in hardware cycles if the services line keeps compounding. That is exactly why a consumer agent with permission to audit, cancel, and reroute payments is not a cute product launch. It is a potential middleman killer.
Call it disintermediation if you want the Wall Street word. In plain English, an agent can sit between a user and a store, recommend a cheaper path, or simply kill a forgotten plan. Meta’s consumer agent, Muse, is being framed as that kind of helper. The connector layer around it is being sold with a zero fee pitch to developers. Zero is a dangerous number when the incumbent takes a sizable cut of digital goods.
If a meaningful slice of store revenue migrates to a zero-fee connector, Apple does not just lose a few points of commission. It loses the narrative that services are a high-quality annuity.
Needham-style math making the rounds puts a rough $10 billion revenue hole on the table under a set of assumptions that are aggressive but not cartoonish. Take services north of $120 billion in a given fiscal year. Assume a large share is store related. Then assume only a fifth of that store slice finds another pipe. You still land on a number big enough to bruise earnings power and, more importantly, the multiple investors pay for that earnings power. Hardware multiples and annuity multiples are not the same animal.
The Forgotten Subscription Problem Is The Real Fuel
People are sloppy with recurring charges. Not because they are reckless. Because life is noisy. A fitness trial becomes a silent debit. A photo vault from three phones ago still bills. A kids’ learning app survives long after the child moved on. Survey work keeps landing near $21 a month of waste for a typical adult. That is not a rounding error across tens of millions of households.
An agent that actively audits those lines changes the psychology. Cancellation used to require a scavenger hunt through settings menus, old emails, and half-remembered passwords. Friction protected vendors. Remove the friction and low-engagement products start dying in clusters. Bank of America style research has already flagged incremental churn as a real headwind if personal agents scale. I tend to agree. The first wave will not be dramatic. It will look like a slightly higher cancel rate in software that nobody loved anyway.
- Unused streaming add-ons that survived a free month
- Cloud storage tiers bought during a storage panic
- Productivity tools opened twice after onboarding
- Niche utilities that felt essential for one weekend
None of those categories is unique to Apple’s store. The store is simply where a huge pile of those relationships still settle. If the agent lives on a phone but routes discovery and payment elsewhere, the store becomes a catalog instead of a cash register. That shift is the part that should keep platform strategists awake.
Developer Economics Can Flip Faster Than Consumers Do
Consumers move slowly. Developers move when the take rate changes. Reports around Muse’s connector launch talked about more than 1,500 developer applications in roughly a week. That is not proof of mass defection. It is proof that the inbox is open. In my experience, developers complain for years and then relocate the instant a credible alternative appears with distribution and a friendlier split.
Apple has improved terms in pieces, especially after regulatory heat. Still, the house cut remains the organizing fact of mobile software. A zero-fee connector does not have to win every category. It only has to win the categories where the product is digital, the user is already logged in, and the purchase is a subscription instead of a one-tap impulse. Games with heavy in-app economies may stay put. Utility subscriptions and mid-tier software are more portable than people admit.
Platforms rarely lose on day one. They lose when the second-best option becomes good enough for the median developer.
Perhaps the most interesting aspect is incentive design. If an agent can recommend an app and complete a purchase without the traditional store tax, the developer can drop price, raise ad spend, or simply keep more margin. Users see a cheaper plan. The agent looks like a hero. The original store looks expensive. That loop does not need 100 percent share. Twenty percent leakage already dents a growth story that investors treat as semi-sacred.
How A $10 Billion Hit Would Actually Show Up
Round numbers help. Suppose services sit near $123 billion. Suppose 40 percent of that is store-ish. That is about $49 billion. Move 20 percent of that slice to a rival connector and you are looking at roughly $10 billion of vanished top line. Apply a rich incremental margin and earnings drop by something like $7.5 billion in a back-of-the-envelope case. Those figures are scenario math, not a forecast carved in stone. Treat them as a sensitivity table, not destiny.
| Assumption | Illustrative Figure | Why It Matters |
| Services base | About $123B | Sets the scale of the annuity |
| Store share of services | 40% | Defines the exposed pool |
| Share that leaves | 20% | The leakage rate investors fear |
| Revenue impact | Near $10B | Large enough to change growth math |
| EBITDA impact | Near $7.5B | Hits the high-quality profit pool |
Would Apple stand still? Of course not. Fee experiments, bundle discounts, tighter agent rules, more first-party subscriptions, louder privacy messaging. The company has spent a decade turning the store into a regulated utility with a luxury brand wrapped around it. Utilities do not die. They get squeezed. The question is whether the squeeze arrives while hardware is already in a slow patch. Stack two soft stories and the multiple compresses even if cash flow remains enormous.
Agents Do Not Need To Replace The Store To Hurt It
This is where a lot of commentary gets sloppy. Nobody serious thinks a new agent deletes the App Store next quarter. Distribution, trust, parental controls, refunds, fraud checks, and habit all still sit with the incumbent. What an agent can do is change the default path for a subset of purchases. Discovery happens in chat. Payment happens on a connector. The icon still lives on the home screen. The money does not.
Think of travel agents in the early web years. Airlines did not vanish. They just stopped capturing the same economics on every ticket. Software platforms can rhyme with that history. I’ve watched enough platform cycles to know the first year is noise and the third year is a line item. If Muse-like agents become the household CFO, low-love subscriptions get purged and new purchases get price-checked. Both actions reduce take-rate volume.
- Audit existing bills and flag low-use plans
- Recommend cancel, pause, or cheaper substitutes
- Complete the new signup outside the old fee pipe
- Repeat monthly until the household stack is lean
That sequence is boring on purpose. Boring processes scale. Flashy demos do not. If the agent is allowed to hold credentials and act, the store’s greatest ally — user laziness — starts working against it.
What Apple Still Controls, And What It Does Not
Apple still owns the glass people touch all day. That is not a small thing. Default browsers, default wallets, default identity, default update cadence. An agent that wants to live on that glass has to play by house rules unless regulators pry the door wider. Privacy positioning remains a real brand asset. Plenty of users will hesitate before handing an outside agent the keys to every subscription.
What Apple does not fully control is developer patience. Or the political weather around store fees. Or the simple fact that consumers like saving $21 a month when a polite assistant offers to do the paperwork. Brand love and switching costs can coexist with gradual leakage. That combination is how high-quality franchises get rerated without a collapse in units.
In my view, the smarter risk frame is not “Apple is doomed.” It is “services growth becomes less automatic.” Less automatic growth is enough to matter when a stock is priced as a compounding machine. Markets punish a change in slope more than they punish a one-off miss.
Small Businesses May Feel This Before Casual Consumers
Households waste money on unused apps. Small firms waste money on unused seats. Project tools, design suites, inbox add-ons, analytics widgets bought during a busy month and ignored after. An agent that speaks operations language could become a part-time controller for shops that cannot afford a real one. That is a different TAM than teenage gamers. It is also a category where fee sensitivity is high and brand loyalty to a store is low.
If agents standardize vendor reviews — usage last 30 days, cost per active seat, cheaper analog — the long tail of software subscriptions gets marked to market. Some vendors will improve product. Some will cut price. Some will just churn. Platforms that taxed that long tail will notice even if flagship games keep printing money.
Valuation Is The Quiet Battlefield
Hardware is cyclical. Services were supposed to be the ballast. Investors pay up for ballast. If ballast starts looking like another competitive software market, the premium compresses. You do not need a collapse in iPhone units for that to happen. You need a credible story that store take rates are no longer a structural given.
That is why the Muse conversation spread beyond product blogs and into equity notes so quickly. It maps onto an existing anxiety: what happens when AI agents become the user interface for commerce? Stores, search boxes, and ad auctions all sit in that blast radius. Apple is not the only name in the room. It is the name with the cleanest high-margin digital tollbooth attached to a luxury device.
Simple investor checklist: Is services growth still outrunning hardware? Is store mix stable inside services? Are developer net fees still attractive versus new pipes? Are cancel rates in subscriptions drifting higher?
Four ordinary questions. If the answers wobble at the same time, the multiple does the talking.
How This Could Play Out Over The Next Few Years
Near term, expect theater. Demos. Policy statements. A few developers testing dual distribution. Lots of language about safety and user control. Medium term, watch cohort cancel rates and the share of digital goods that settle outside official stores. Long term, either the incumbent matches economics closely enough to keep the herd, or agents become a parallel checkout layer the way wallets became a parallel payment layer.
I do not buy the most dramatic version of the story. Ecosystems this large have immune systems. I also do not buy the idea that nothing important changed. A helper that cancels waste and shops fees is aligned with how people actually want to live. Alignment beats nostalgia. If Apple treats this as a messaging problem instead of a pricing and developer-relations problem, the $10 billion thought experiment stops being a slide and starts being a trend line.
One last thought, and it is more personal than the models. Most of us are tired of being the unpaid accountant of our own digital lives. An agent that does that job well will win affection even if it makes a famous platform slightly less profitable. Affection is not a valuation metric. It is how distribution shifts when nobody is watching the settings menu. That is the part worth tracking after the first headlines fade.
