Latest / Deep Values: Value Investing Deep Dives / Stripe and the Dawn of Agentic Commerce Explained
Transcript
- 0:00You know, if you hand a toddler a credit card, you basically
- 0:02expect absolute chaos. Oh yeah, total disaster.
- 0:05Right. Like maybe $1000 in accidentally
- 0:07purchased tablet games or, I don't know, a lifetime supply of
- 0:11gummy bears suddenly showing up on your front porch.
- 0:15Honestly, the gummy bears sound kind of great, but yeah, it's a
- 0:18funny, slightly terrifying image because a toddler has the
- 0:24physical means to spend but absolutely zero understanding
- 0:28of, you know, value or consequence.
- 0:30Exactly. It's spending without reasoning.
- 0:32Yeah, it's an action just totally disconnected from any
- 0:35sort of economic logic. Right.
- 0:36But imagine you hand a credit card not to a toddler, but to a
- 0:40highly intelligent, perfectly rational AI.
- 0:43Now that is a very different scenario.
- 0:45Suddenly it's not chaos anymore, it's an entirely new economic
- 0:49reality, like what actually happens when the entity shopping
- 0:52online isn't human. It is the absolute definition of
- 0:55a paradigm shift. I mean, we are looking at the
- 0:57dawn of a completely automated financial layer.
- 1:00Where machines are just negotiating and buying things.
- 1:02Buying, selling, negotiating with other machines in literal
- 1:06milliseconds. And that is exactly what we are
- 1:09exploring with you today. Welcome to our deep dive.
- 1:13Our mission today is to unpack this massive quiet revolution
- 1:18happening right under our noses, the rise of agentic payments and
- 1:23the machine to machine economy. Which is really starting to
- 1:25accelerate now. Oh, definitely.
- 1:27A lot of this crystallized recently during the massive
- 1:30Stripe Sessions conference in San Francisco.
- 1:32But before we get into the mechanics of algorithms with
- 1:35wallets, we have quick note for you to access more advanced
- 1:39researches, applications and our latest published personal
- 1:42intelligent multi agent Deep value intelligence.
- 1:44Welcome to explore and join deep values dot AI community to track
- 1:48total market temperature. Welcome to Search and download
- 1:51iOS App Buffett Indicator from Apple App Store.
- 1:54Setting the stage here, it's crucial to realize that you know
- 1:56while you hear everyone talking about AI generating text.
- 1:59Painting pictures. Exactly.
- 2:00Or writing code that is really just the surface level.
- 2:03It's the shiny stuff. Yeah, the tectonic shift, the
- 2:07one that completely alters global financial infrastructure,
- 2:11is what happens when those AI agents can independently hold
- 2:15and spend capital. OK, let's unpack this, because
- 2:18to understand why agentic payments matter to you, we first
- 2:21have to understand the sheer unprecedented speed at which new
- 2:25businesses are forming right now.
- 2:27It's moving so fast. It really is.
- 2:29During that recent strike conference, their founder,
- 2:32Patrick Collision, kept referring to a singularity.
- 2:35Right, but not the sci-fi kind. Exactly.
- 2:37He wasn't talking about ADI. You know, artificial general
- 2:40intelligence suddenly waking up, becoming sentient and taking
- 2:43over the world. Which is what everyone usually
- 2:45assumes when they hear that word.
- 2:47Right, he was talking about a purely economic inflection
- 2:49point. He literally noted they were on
- 2:52the 119th day of the singularity meeting, early 2026, where
- 2:56business growth curves have just suddenly gone vertical.
- 2:59Yes, he's describing a state change in the velocity of
- 3:02commerce itself. And they see this in their own
- 3:04data. Oh absolutely.
- 3:05The metrics they are seeing on the back end of the global
- 3:08payment infrastructure are simply staggering.
- 3:11It's a it's unlike anything in modern economic history.
- 3:14The hard day to hear is what really blew my mind.
- 3:17Listen to this In 2020, only 8% of startups using Stripe got
- 3:23their first customer payment within a month of incorporating,
- 3:27which is. Pretty standard, right?
- 3:29It takes time to set up. Exactly.
- 3:31But today, yeah, that number has jumped to 20%.
- 3:35Wow. The time it takes a company to
- 3:37go from a mere idea on a whiteboard to $10 million in
- 3:40annual recurring revenue is being ruthlessly, almost
- 3:43violently, compressed. Yeah, we are talking about
- 3:46compression from years down to month.
- 3:48It's like going from dial up Internet to fiber optic, but for
- 3:51business creation. But wait, here's where I have to
- 3:54push back a bit, or at least look at the mechanics.
- 3:56Sure, go for it. If businesses are launching,
- 3:58scaling and transacting literally overnight, human
- 4:01beings physically cannot click, buy or prove fast enough to
- 4:04drive that kind of vertical growth.
- 4:06Right, we have a biological speed limit.
- 4:08Exactly. So if we can't move that fast,
- 4:11who or what is actually doing the buying and selling at that
- 4:14speed? That is the pivotal question,
- 4:16and the answer is that the buyers and sellers are
- 4:18increasingly algorithms. Not humans.
- 4:22No, not you or me sitting at a laptop clicking add to cart.
- 4:25If we connect this to the bigger picture, this explosion of new
- 4:30tech startups, the severe compression of time, it is
- 4:33directly driven by AI. Meaning the AI is doing the
- 4:36heavy lifting. All of it.
- 4:37AI is writing the initial code. AI is optimizing the server to
- 4:41employment. It's handling the initial
- 4:43customer service. So the whole metabolism of the
- 4:45economy is just speeding. Up way beyond human capacity to
- 4:48manage it manually. Which brings us to a vital
- 4:51distinction we need to make for you.
- 4:52Since the economy is speeding up because of AI, we really need to
- 4:56clearly distinguish between how you use AI as a tool to buy
- 4:59things and how AI buys things for itself.
- 5:02A very necessary distinction. Yeah, we are looking at the
- 5:05difference between agentic commerce and agentic payments.
- 5:08Because they represent 2 totally different areas of the Internet,
- 5:11really. Right, so agentic commerce is
- 5:14essentially you using an AI as an ultra capable assistant.
- 5:18Like imagine you tell your AI order me a medium cappuccino
- 5:21from the place down the street. Right, a simple fetch quest.
- 5:23Yeah, the AI goes to the site, navigates the menu and fills out
- 5:28the forms, but you are making the final choice and your
- 5:31payment info is just being passed along through Apple Pay
- 5:35or your saved visa. You are still the center of that
- 5:38transaction. Exactly.
- 5:40Agentic payment, however, is entirely machine to machine.
- 5:43Right, in agentic payments the human is completely removed from
- 5:46the micro decisions. So it's not just filling out
- 5:48forms. No, not at all.
- 5:50It is evaluating services, deciding what raw data or
- 5:53computing power it needs to accomplish a goal, and
- 5:55independently executing a financial transaction with
- 5:58another machine using its own dedicated wallet.
- 6:00Here's where it gets really interesting.
- 6:02Let's look at a specific live demonstration from that Stripe
- 6:05Sessions conference that perfectly illustrates how this
- 6:08actually works. This was a great demo.
- 6:11It was wild. So a developer asked their AI
- 6:13agent to write a comprehensive research report on energy.
- 6:19That was it. That was the only prompt given.
- 6:21No other instructions. None.
- 6:23And without any human help, the agent realized it needed highly
- 6:26specific proprietary data to write a good report.
- 6:30So it used its own digital wallet to spend a few fractions
- 6:33of a cent to query a specialized gated database.
- 6:36It. Bought its own research
- 6:37material. Exactly.
- 6:38It gathered the data, wrote the report, and then this is the
- 6:41wild part. It independently paid Amazon Web
- 6:44Services to host a brand new website and listed that energy
- 6:48report for sale to other AI agents.
- 6:50It's a fully automated closed loop economic cycle.
- 6:53All on its own. Yeah, the AI acts as the
- 6:55consumer buying data, the producer synthesizing it, and
- 6:58the merchant selling it, all in the span of a few seconds.
- 7:01OK. Let's pause here because I'm
- 7:02thinking and imagine you were listening to this and thinking,
- 7:04Wait a minute, am I just supposed to hand my AIA credit
- 7:07card, tell it to do research and let it run wild?
- 7:10Right. People are terrified of that.
- 7:11How do you ensure you don't wake up completely bankrupt because
- 7:15your AI decided the most efficient way to write a report
- 7:18was to buy an entire server farm in Virginia?
- 7:21This raises an important question regarding trust and
- 7:24guardrails, and it's a completely valid fear, but you
- 7:28aren't giving the AIA blank check.
- 7:31OK, so how does it work? The human still sets the
- 7:34overarching goals and the strict mathematical parameters you tell
- 7:37the agent. Write this report and your total
- 7:41hard coded budget is exactly $5. And it literally can't spend
- 7:45more than that. Exactly.
- 7:46The AI operates entirely within that cryptographic constraint.
- 7:49What the shift does is completely remove human decision
- 7:52fatigue from the transaction process.
- 7:54Decision fatigue. Oh, that makes perfect sense,
- 7:57right? You don't have to get a push
- 7:58notification to manually approve the one cent database query or
- 8:01the three cent hosting. Fee.
- 8:03You'd go crazy if you did. I would throw my phone out the
- 8:05window. The AI manages the micro budget.
- 8:08But wait, if decision fatigue disappears for machines the
- 8:11entire way we currently charge for software like monthly
- 8:14subscriptions suddenly makes zero sense.
- 8:16It completely collapses, right? Oh absolutely.
- 8:19I mean, to understand why, we have to look at why micro
- 8:23payments historically failed for us humans.
- 8:26Tech visionaries have been predicting a pay per click micro
- 8:29payment revolution since the 1990s.
- 8:32Yeah, I remember reading articles about that years ago.
- 8:34But it never caught on. Think about Spotify.
- 8:36Paying $900.99 a month for unlimited music is a pleasant,
- 8:40seamless experience for you. Set it and forget it.
- 8:43Exactly. But imagine if instead of a
- 8:46subscription, you had to consciously decide whether a
- 8:49single song was worth $0.15 every single time you press
- 8:51play. Oh I would absolutely stop
- 8:53listening to music. You're right.
- 8:54The mental friction of asking myself, is this track worth
- 8:57$0.15? Yeah, over and over is just
- 9:01exhausting. Precisely.
- 9:02Humans naturally hate micro decisions.
- 9:05The cognitive load of evaluating a $0.15 purchase is actually
- 9:08higher than the economic value of the $0.15 itself.
- 9:11We crave the simplicity of a flat fee so we don't have to
- 9:13think about it. It reminds me of the old like
- 9:16100 year old general store in a small town.
- 9:18Oh, that's a good comparison. You know you didn't pull out
- 9:21your physical wallet and count out pennies every time you grab
- 9:23an apple or a bag of flour. You kept a running tab and
- 9:27settle up once at the end of the month because paying for apple
- 9:29was just too annoying, right? Subscriptions are essentially
- 9:33the modern digital version of the general store tab.
- 9:37But here is the massive difference that changes
- 9:39everything. AI does not get tired.
- 9:44AI experiences 0 cognitive load. Evaluating a fraction of a cent
- 9:49transaction against a budget takes an AIA millisecond and
- 9:53causes it 0 frustration. It just does the math and moves
- 9:56on. Exactly.
- 9:57Therefore, for an AI, pay as you go is actually the most
- 10:00perfectly logical and efficient economic model.
- 10:03Why on earth would an AI pay a $50 monthly subscription for a
- 10:07data service if it only mathematically needs to query
- 10:09that service once for a single penny?
- 10:11It wouldn't. And if you are running a
- 10:14Business Today, you have to really let the gravity of that
- 10:16land. This signals a massive
- 10:18structural threat to the traditional Sauce software as a
- 10:21service subscription model that has completely dominated the
- 10:24tech industry for the last decade.
- 10:25It's going to upend everything. Think about massive giants like
- 10:28Adobe, Salesforce, or Netflix. Their entire multibillion dollar
- 10:33valuations are built on humans paying a flat monthly fee
- 10:38regardless of how much they actually use the service.
- 10:40Yep, the unused portion is pure. Profit.
- 10:43Exactly. But if your primary customers
- 10:45are becoming AI agents, you cannot force them into monthly
- 10:48subscriptions. They would just route around you
- 10:51and find a competitor that charges per API call.
- 10:53And for the listener, an API call is basically just
- 10:56software's way of knocking on another software's door and
- 10:59asking for a specific piece of data or action.
- 11:03Just a digital request. Yeah.
- 11:05It fundamentally rewrites the revenue models of these
- 11:08companies. They have to shift from selling
- 11:10unlimited access to selling raw metered usage.
- 11:13So if metered, pay as you go is the inevitable future.
- 11:16We need to talk about the actual plumbing here.
- 11:19How do these agents physically transmit fractions of a cent to
- 11:22each other instantly? Because you can't use
- 11:24traditional methods for that. No, you cannot swipe a
- 11:27traditional Visa or MasterCard for 110th of a penny.
- 11:31The flat 30 cent transaction fee that credit card networks charge
- 11:35would be, what, 300 times the cost of the actual item being
- 11:39bought? It's economically impossible.
- 11:41This is exactly where we get into the new protocols being
- 11:44rapidly developed to support this machine economy.
- 11:47The traditional rails simply cannot handle the volume or the
- 11:50tiny denominations required by AI.
- 11:52Right. And two major technical
- 11:54solutions are battling it out right now, Coinbase's X4O2
- 11:57protocol and the Machine Payment Protocol, or MPP, which is
- 12:01backed by players like Stripe and Ripple.
- 12:03Both very interesting approaches.
- 12:05But the history behind X4 O2 is absolutely fascinating.
- 12:08It literally resurrects an old, completely unused Internet error
- 12:12code. I love this part.
- 12:13Like we all know HTTP four O 4 which means page not found.
- 12:17But decades ago, in the late 80's, the original architects of
- 12:20the Internet created HTTP Four O 2, which stands for Payment
- 12:23Required. They knew it was coming.
- 12:24They did. They envisioned a native payment
- 12:27layer for the web, but digital cash didn't exist yet, so it
- 12:30just sat dormant for decades. And now it's the foundation of
- 12:36AI payments. It's a beautiful piece of
- 12:38Internet archaeology. Let's look at how it works
- 12:40mechanically though. Yeah, break it down.
- 12:42When an AI agent tries to scrape a site or hit a ay wall on an
- 12:45API, instead of justice blocking access, the server returns a 402
- 12:50header code. That code essentially contains a
- 12:53digital invoice saying I have the data you want, but you need
- 12:56to send 0.01 cents to this specific cryptographic address.
- 13:01And then the AI just pays it. Yeah, the AI's wallet reads that
- 13:04header, validates it against its budget, automatically finds a
- 13:08transaction, and unlocks the data in roughly 400
- 13:11milliseconds. But hold on, I have to challenge
- 13:12this premise a little bit. Whenever we talk about
- 13:15cryptographic addresses or crypto rails, people immediately
- 13:18think of volatility, FTX collapsing, and massive
- 13:22regulatory Gray area. That's the common perception.
- 13:25If I'm a legitimate, compliant business owner, why on earth
- 13:28would I want my autonomous AI using an unregulated, highly
- 13:32volatile casino chip to buy server space?
- 13:35Why can't we just build a better virtual corporate credit card?
- 13:39That is exactly the pushback most traditional finance
- 13:41executives have, but it misunderstands what is actually
- 13:44being transacted here. We aren't talking about AI
- 13:47speculating on Bitcoin. So what is it?
- 13:49What's fascinating here is the concept of the Shared Payment
- 13:52Token, or SPT. Now, despite the word token,
- 13:56SPTS are not volatile crypto currencies.
- 14:00They are programmable digital authorizations, essentially
- 14:04smart contracts. Ending they have rules attached.
- 14:06Exactly when you authorize your agent, you give it an SPT.
- 14:09That token physically contains code saying you can spend up to
- 14:12$5 only today, and only at this terrific list of verified
- 14:16merchants. Oh wow.
- 14:17OK, so it's not just a blank digital coin.
- 14:19It's a payment that has a contract physically embedded
- 14:22inside the money itself. It cannot be spent outside side
- 14:25of the rules I define. Exactly.
- 14:26The compliance and the security are baked into the asset itself,
- 14:29not enforced after the fact by a bank.
- 14:31That makes a lot more sense. And while the machine payment
- 14:34protocol MTP does support traditional Fiat currency if you
- 14:38really want to use it, the reality is that stable coins are
- 14:41naturally suited for the AI economy.
- 14:43And just for the listener, a stablecoin is a cryptocurrency
- 14:46that is directly pegged to the US dollar, usually backed one to
- 14:50one by actual dollars or treasury bills in a bank so it's
- 14:55value doesn't wildly fluctuate. One stable coin equals $1.00.
- 15:00Always. Precisely because stable coins
- 15:02live on blockchain rails instead of banking rails.
- 15:05They are inherently borderless. They settle instantly, 24 hours
- 15:09a day, seven days a week, and on modern networks they cost
- 15:12fractions of a penny to send. Unlike a credit card.
- 15:15Right, Visa wasn't built for machines.
- 15:17It requires banking hours, intermediaries and high fees.
- 15:20Yeah, and AIH and doesn't care that it's a bank holiday in
- 15:23London or that international national wire transfers take
- 15:25three days to clear. It will automatically route the
- 15:28payment through the most mathematically efficient
- 15:30cheapest channel available always.
- 15:32And nine times out of 10, that most efficient channel is going
- 15:35to be a stable coin on a high speed blockchain.
- 15:37It is the path of least resistance for machine logic.
- 15:41The AI optimizes for speed and cost, completely ignoring the
- 15:44traditional financial borders humans have built.
- 15:46So if the buyers are increasingly these AI agents
- 15:50using stable coins and HTTP4O2 codes, the sellers had to
- 15:54completely change how their storefronts look.
- 15:56Oh, everything has to change. If you are selling data or
- 15:59software today, you are no longer just selling it to a
- 16:02human sitting at a laptop, you are selling it to a piece of
- 16:05code. Which brings us to the massive
- 16:07architectural shift required for merchants to actually capture
- 16:10this new revenue. The venture capital firm a 16
- 16:13users coined a brilliant concept for this.
- 16:15They call it headless commerce. Let's visualize this for a
- 16:19second. Think about a traditional online
- 16:21store. You have beautiful website, high
- 16:23resolution photos, a friendly little sales team chat bubble
- 16:26popping up. Hi, how can I help you today?
- 16:29Exactly. And maybe a complex CRM customer
- 16:31relationship management system to track the customer's birthday
- 16:35and send them a 10% coupon. An AI agent doesn't care about
- 16:38literally any of that. A beautiful user interface is
- 16:41entirely wasted on an algorithm. It's like taking down the flashy
- 16:44neon signs, the mood lighting and the friendly greeter at a
- 16:47grocery store and replacing the whole building with a giant
- 16:50pitch black warehouse where robots just scan barcodes in the
- 16:54dark. That is a perfect analogy.
- 16:56The robots don't need the lighting or the greeter, they
- 16:58just need the barcodes to be hyper accurate and easy to read.
- 17:02The algorithm only cares about the underlying data structure.
- 17:05It wants predictable pricing, incredibly high server uptime,
- 17:09and machine readable documentation.
- 17:11If the barcode is blurry, the robot leaves.
- 17:14And Stripe is actively adapting to this headless future right
- 17:18now. They aren't just building pretty
- 17:19payment portals for human web pages anymore, they are
- 17:22providing their documentation in markdown formats.
- 17:25Which is super important for AI. Right.
- 17:27Markdown is essentially A stripped down plain text format
- 17:31that is incredibly easy and fast for a large language model like
- 17:35ChatGPT or Claude to digest. They're creating specific skills
- 17:39that be fed directly into these AI models.
- 17:41Skipping the website entirely, Yes.
- 17:44They are making it so an AI can read their entire technical
- 17:47manual in a millisecond and instantly know exactly how to
- 17:51route money without ever looking at a web page.
- 17:54We are witnessing the transition from optimizing for UX user
- 17:58experience to optimizing for AX agent experience.
- 18:02Agent experience, I love that. In the next few years, if your
- 18:05business doesn't have a seamless AX, the AI buyer will simply hit
- 18:09an error, bypass you, and instantly spend its stable coins
- 18:12at a competitor that does. Wild is that this connects
- 18:15directly back to Stripes original origin story.
- 18:18It really does. When Stripes started over a
- 18:20decade ago, their whole pitch to the world was payment literally
- 18:24just 7 lines of code. They bypassed the finance
- 18:27executives who were arguing about merchant accounts and went
- 18:30straight to the developers saying just copy and paste this
- 18:32code and your payment problems are solved.
- 18:34And it worked brilliantly. Now they're running the exact
- 18:37same playbook, but instead of making it easy for human
- 18:39developers, they're making it easy for autonomous AI agents.
- 18:43It is a remarkable full circle moment.
- 18:46The core philosophy remains exactly the same.
- 18:49Ruthlessly eliminate friction from the system, but the target
- 18:53audience has evolved from human engineers to autonomous
- 18:57software. So what does this all mean for
- 18:59you? If we step back and look at the
- 19:00entirety of what we've unpacked today, from the speed of
- 19:03business formation to the rebirth of four O and two error
- 19:06codes, to the potential death of the SAS subscription model and
- 19:10the rise of headless commerce, we are looking at a fundamental
- 19:13transition in human history. A.
- 19:15Completely new. Era We are moving from humans
- 19:18using AI merely as a tool, like a glorified calculator or an
- 19:21intern, to AI acting as an independent autonomous economic
- 19:25participant. It means the friction of doing
- 19:28business, The time it takes to negotiate, execute and settle a
- 19:31transaction is dropping to near absolute 0.
- 19:34And when economic friction drops to 0, the volume of transactions
- 19:37will likely skyrocket in ways we can't fully predict yet.
- 19:40But it leads us with a fascinating, almost unsettling
- 19:43question to Mull over. If headless commerce is indeed
- 19:48the future and AI agents are the primary entities shopping around
- 19:52the Internet for the best deals, optimizing every fraction of a
- 19:55cent, what happens to the multi billion dollar advertising and
- 19:59marketing industries? They are in deep trouble.
- 20:02How on earth do you market a product to an algorithm?
- 20:05It doesn't care about your emotional branding, it doesn't
- 20:07watch your Super Bowl commercials, and it doesn't care
- 20:09about your influence or campaigns.
- 20:11It only cares about raw data ricing efficiency and uptime.
- 20:15It fundamentally changes how value is perceived and
- 20:18communicated in the economy. You just can't manipulate a
- 20:20machine with a flashy ad. You really can't.
- 20:24It brings us right back to that image we started with.
- 20:26Handing a toddler a credit card is chaos because they act on
- 20:30pure, unfiltered emotion. Handing an AIA wallet is a
- 20:33revolution because it acts on pure, unfiltered logic.
- 20:37That the question is, is your business ready for a consumer
- 20:39that never sleeps, never gets confused by marketing, and never
- 20:42hesitates? Something everyone needs to be
- 20:44thinking about. Absolutely.
- 20:47Before we officially wrap up, I need to read our required legal
- 20:50note. Disclaimer All analysis is based
- 20:52on publicly available sources and maybe subject to revisions
- 20:55or differing interpretations. All contents from this channel
- 20:58is for information and research purposes only and does not
- 21:01constitute investment advice. Conduct your own due diligence
- 21:04before making investment decisions.
- 21:06If you like to access wider topics, including life and
- 21:08investing philosophy, please subscribe to our sister channel,
- 21:11Deep Values. Don't forget to like this video.
- 21:13Much appreciated. Thanks for joining us on the
- 21:16Deep Dive. We'll see you next time.