Disclaimer: As a former PayPalian Gen2, I have a strong emotional bias in the comments below but not so much financial gains to it. Too bad. And it seems PayPal board also agreed on the undervalue valuation.
Are prospective buyers valuing PayPal as the business it has historically been, a mature payments processor with slowing growth, or as an asset that could become significantly more valuable in an AI-mediated economy?
According to reports, Stripe and private-equity firm Advent International have offered $60.50 per share for PayPal, implying an equity value of approximately $53.4 billion. That represents a roughly 28% premium to PayPal’s unaffected share price. The proposal reportedly includes approximately $50 billion in committed bank financing, although the companies have declined to comment publicly. (Reuters reporting, Axios summary)
On conventional measures, the logic of the offer is straightforward. PayPal is a mature fintech platform with an established brand, considerable cash generation and an enormous payment network, but comparatively modest growth. A buyer can estimate future transaction revenue, margins, cost reductions and operational synergies, discount those cash flows and add an acquisition premium.
That framework may be financially orthodox. It may also be incomplete.
PayPal is sometimes discussed as though it were a legacy technology or SaaS platform. Neither description captures its most important characteristics.
SaaS businesses are generally valued through recurring subscription revenue, customer retention, expansion rates and software margins. PayPal is principally a transaction platform and two-sided network. Its economics depend on payment volume, transaction margins, consumer and merchant engagement, fraud losses, credit performance and network reach.
In 2025, PayPal processed approximately $1.79 trillion in total payment volume across 25.4 billion transactions. It ended the year with 439 million active accounts in approximately 200 markets. PayPal itself says that its relationships with both consumers and merchants allow it to use data to reduce friction, drive sales and improve shopping experiences. (PayPal 2025 Form 10-K)
Those figures describe more than payment-processing capacity. They represent a continuously refreshed record of real economic behavior:
What consumers actually purchase, rather than what they search for or “like”
Which merchants successfully convert demand
How price, timing, geography and payment method affect conversion
Which transactions produce disputes, fraud, returns or credit losses
How commercial relationships develop across multiple years
How online identity connects with verified payment behavior
This distinction becomes increasingly important as commerce moves from human-directed websites toward AI agents that search, recommend and eventually transact for consumers.
General-purpose AI models are trained primarily on language, images, code and publicly available information. That can teach a model what people say about products. It does not necessarily reveal what people ultimately buy, whether they keep it, whether the transaction was fraudulent or how the merchant performed.
PayPal possesses something potentially more useful for commerce applications: long-duration, outcome-linked transactional information.
That information could help improve several AI-driven services:
Purchase recommendations based on demonstrated behavior
Merchant and product ranking
Personalized offers and rewards
Fraud and identity detection
Credit underwriting
Advertising attribution
Pricing and conversion optimization
AI-agent authorization and transaction monitoring
PayPal has already started commercializing this advantage. Its advertising business markets access to a transaction graph informed by approximately 25 billion annual transactions. PayPal says AI can analyze this activity to help advertisers identify high-intent consumers and measure incremental sales. It has also launched off-platform advertising products built around its transaction data. (PayPal Advertising, PayPal Offsite Ads announcement)
The strategic opportunity is therefore not simply to train another large language model. Stripe would not need PayPal’s data to teach a model how to speak.
The opportunity is to build a commerce intelligence layer around existing models: a system that can recognize customers, predict intent, assess risk, rank commercial options and safely execute transactions. In this formulation, PayPal’s historical data is most valuable when combined with its live identity, merchant and payments infrastructure.
Accounting and valuation systems remain uncomfortable with internally generated data.
Data is not normally carried on a company’s balance sheet at an independently assessed market value. The cost of collecting and managing it may appear in operating expenses, while its potential contribution to new products remains largely invisible. As a result, a transaction database accumulated over decades can be treated as a byproduct of the operating business rather than as an economically productive asset.
The OECD now describes data as a core input into AI-enabled business models, while noting that its non-rival nature, context dependence and lack of observable market prices make conventional valuation difficult. The new international statistical framework has consequently begun treating “data and databases” more explicitly as productive assets. (OECD Going Digital Measurement Roadmap, OECD Productivity Indicators)
Research also suggests that data’s value depends heavily on who owns it and what complementary capabilities that owner possesses. The same dataset can have dramatically different values for different buyers. (NBER: “Valuing Financial Data”)
This is particularly relevant to Stripe. PayPal’s information could be more valuable to Stripe than to a financial buyer, or even than it is to PayPal independently, because Stripe already has the technical infrastructure, merchant integrations and developer ecosystem needed to activate it.
The answer is not to discard discounted cash-flow analysis or assign an arbitrary “AI multiple” to every large database. A more credible model would separate four sources of value:
1. Stand-alone operating value
The expected cash flows from PayPal’s existing payments, Venmo, Braintree, credit and related services.
2. Conventional synergies
Cost savings, infrastructure consolidation, improved pricing, lower customer-acquisition expense and increased merchant penetration.
3. Data-enabled product value
The probability-weighted cash flow from advertising, personalization, fraud reduction, underwriting, merchant intelligence and agentic-commerce products that PayPal’s data could support.
4. Strategic option value
The value of owning a difficult-to-recreate commerce and identity graph before the ultimate structure of AI commerce becomes clear.
The third and fourth components should be valued through specific use cases, not by estimating a price per transaction record. For each use case, the buyer should estimate:
Addressable revenue or savings × expected performance improvement × attainable adoption × contribution margin × legal usability × probability of execution
The resulting cash flows can then be discounted according to their technical, commercial and regulatory risk.
This produces a more disciplined “AI data premium.” It also prevents buyers from paying twice for benefits already captured in the ordinary operating forecast.
There are substantial reasons to resist an unlimited AI premium.
First, financial data cannot necessarily be repurposed simply because a company possesses it. Privacy obligations, consumer consent, contractual restrictions, data-localization requirements and financial regulation may constrain how individual records can be used. Data that is legally inaccessible to a model has little incremental valuation.
Second, much of PayPal’s volume includes Braintree processing, where PayPal may not have the same consumer relationship or depth of information that it has within a logged-in PayPal or Venmo transaction. Twenty-five billion transactions should not be mistaken for 25 billion equally rich AI-training examples.
Third, historical behavior can become stale. The value lies in the continuing data-generation system, not merely in an archive.
Fourth, data is complementary to product execution. Research cited by the OECD finds that proprietary operational data can help incumbents fine-tune and personalize AI services, but also warns that many datasets are substitutable. Data becomes defensible when it is unique, legally usable, continuously refreshed and paired with the infrastructure needed to turn predictions into actions. (OECD analysis of AI competition)
Finally, combining Stripe and PayPal would create major competition and governance questions. A transaction joining two enormous merchant-payment systems would likely receive close scrutiny, particularly if the strategic justification includes concentrating payment, identity and consumer-behavior data.
A $53 billion offer may look generous compared with PayPal’s recent market price while still being conservative relative to the value Stripe believes it can create.
The difference depends on whether PayPal’s transaction graph is merely useful data inside a mature payments company or the foundation of a new commerce-intelligence platform.
That value is not automatic. It must survive legal review, technical validation and commercial execution. But it should not be ignored simply because current accounting and M&A models have no convenient line item for it.
The right valuation question is therefore not, “How much is PayPal’s data worth?”
It is:
How much additional, defensible cash flow can a particular owner create from PayPal’s continuously refreshed combination of identity, merchant relationships, transaction outcomes and payment execution, and how much of that value should PayPal’s current shareholders receive?
If AI agents become an important interface through which consumers discover and purchase products, payment companies will no longer be only the infrastructure at the end of a sale. They may become the trust, identity and intelligence layer that determines which sales happen at all.
Under that scenario, Stripe and Advent would not merely be bidding for yesterday’s PayPal. They would be bidding for a potential control point in tomorrow’s AI economy.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.