Paul Graham: When Startups Make Decisions, Don't Just Ask How to Make More Money—Ask How to Make the Company More Powerful

In startup advising, there is a watershed question. When facing a startup that is lost or plateauing, seasoned mentors often resist the urge to ask "how can you make more money?" and instead pose a more fundamental question: What can you do to make this company more powerful?

These two ways of asking represent entirely different dimensions of thinking. Pondering how to make more money is certainly a useful intuition, but it usually leads to incremental improvements like raising prices, trimming costs, or optimizing customer acquisition. By contrast, thinking about how to grant a company genuine commercial power can directly reshape the architecture of the business, unlocking several orders of magnitude in value at critical moments.

This explainer completely dissects Paul Graham's systematic thinking on "commercial power," clarifying how startups can leap from passive suppliers to rule-making kingpins who control the game.

Levels of Questioning: Why Pursuing "More Power" Far Outweighs "How to Make More Money"

When most startups hit a growth plateau, their instinctive reaction is to hunt for new ways to make money: adding paywalls, building bespoke enterprise features, or raising average contract values. While intuitive, these moves are essentially wringing water from a dry towel within an existing commercial structure.

Real power is not squeezed out by wringing towels. In a business context, power means bargaining leverage, indispensability, and gravitational pull over the entire upstream and downstream ecosystem. A company with power can forfeit short-term profits to the ecosystem and still capture the core dividends of the entire industry over a long horizon. Conversely, a company lacking power—no matter how flattering its current top-line figures—can be wiped out in an instant by an app store policy shift, an upstream component cutoff, or a canceled enterprise contract.

Therefore, evaluating the potential of a startup idea hinges on whether it can undergo a morphological transformation—moving from a vulnerable, subservient link in the value chain into a central hub that sets the rules.

Seizing the Chokepoint: Making Money Flows, Token Flows, and Industry Standards Pass Through You

To make a company more powerful, the most direct path is to alter its physical positioning in the transaction chain.

Many startup ventures begin as specialized technology modules—that is, "component suppliers." The greatest agony of being a supplier is that you must live inside the box that downstream companies draw for you. One way out that Graham repeatedly pursues is finding ways to directly own the relationship with the end customer and having cash flows route through your business. Having money flow through you usually puts you in a more advantageous position; yet this position must still rest on product value that customers are genuinely willing to adopt.

Entering the artificial intelligence era, this logic has evolved a timely new variant: letting tokens flow through you.

Graham extends this logic to the AI era: if tokens flow through your product and users pay for these tokens, money will usually flow through you as well. In this conception, you command the end-customer relationship, while foundation model providers look more like underlying component suppliers. However, this position is not inherently unassailable—foundation model providers can still expand upward, eat the middle layer, or even reach your customers directly.

Beyond financial flows, a higher-order form of ecosystem power is the app store model. If other companies can build extensions on top of your product, their value-creating efforts will in turn augment your own value. If your scale is not yet large enough to erect a full-fledged app store, another strategy is attempting to define industry interaction standards. When an emerging field lacks an established protocol, even a tiny startup can propose a valuable standard; Graham believes that the market's hunger for standards makes the earliest workable proposal much more likely to win.

Inducing Network Effects: Transforming Point Tools into Marketplaces

Network effects dramatically amplify a company's power. Graham notes that when talking with startups, he is almost always thinking about how to introduce network effects, even treating it as a personal challenge: "finding network effects in businesses where you wouldn't expect them." Remarkably, this thought experiment often works, transforming the nature of the idea—what began as a mere service can evolve into a marketplace. In a footnote, he points out that YC itself is an example of an accidentally formed network effect: the team had no such deliberate design intent at the beginning, but quickly realized they had stumbled onto one.

Even the most basic utility software can spin up a network flywheel by introducing information-sharing mechanisms:

Benchmark data sharing: Allowing users to opt into an anonymized data comparison pool. By authorizing the sharing of their operational data, users receive system benchmarks showing their ranking and performance relative to all industry peers. The larger the data pool, the sharper the analytics, and the stronger the incentive for new users to join.

Interactive data feedback training: In AI tools, allowing users to opt into letting the system use their interaction data to train models. Even if some users decline, Graham argues that as long as some opt in, the models they use may outperform standard versions trained without such data, creating a performance edge powered by scale of participation.

If you generalize the concept further, a point tool can even morph into a vibrant marketplace.

Suppose a startup is building a payment gateway for AI agents. The basic conception is enabling agents to buy things for humans. But if you take a step back and ask: Can agents pay each other? This slight adjustment in premise instantly converts a simple API gateway into a two-sided marketplace across agents. In the market's nascent days, even if what agents will transact with one another remains unclear, it is worth the entire team pouring their energy into exploring use cases, or even stepping in as market makers to bootstrap liquidity. The power and value of a marketplace far eclipse those of a one-way payment pipe.

However, Graham offers a vital methodological caveat here: running these hypothetical transformational thought experiments on an initial idea (whether turning it into a platform, a marketplace, or a full-stack business) by no means always yields promising results—in fact, it often doesn't. Yet these mental drills are always worth undertaking; even if they ultimately come up empty, the sheer process of trying to morph the idea forces founders to understand the fundamental nature of their business far more deeply.

Going Full Stack Along the Grain: From Selling Shovels to Digging for Gold Yourself

Programmers often experience a wonderful sensation: when a codebase architecture becomes clean, abstract logic acquires the palpable solidity of a physical entity. Reframing startup ideas feels much the same.

Among all forms of business reinvention, the strategy with the most tangible visceral force is "going full stack": instead of selling software to companies operating in a given trade, you use the efficiency advantages of that very technology to enter the business yourself, competing directly with your original target customers. Your former customers are swallowed whole, and you leap to become the entity directly facing the end consumer.

Full-stack evolution also has a gentler yet stealthier trajectory, which can be described as incremental cannibalization:

Initially, you merely help customers handle their most agonizing, technically demanding steps. As your system's capabilities deepen, you progressively take over all core decision-making and cognitive heavy lifting across the business chain. In the terminal state, the customers who originally bought your software are effectively demoted to offline errand runners, behaving like marionettes dancing on your strings; their original customers, naturally, become your true customers.

Listening to User "Misuse": Letting the Tail Wag the Dog

Throughout business history, the dramatic spectacle of the "tail wagging the dog" plays out repeatedly—a marginal feature hastily thrown together to support a core business ends up usurping the host and growing into a generational giant.

The quintessential archetype of this phenomenon comes from early payment pioneers:

PayPal's Host Takeover: The team originally built handheld device security software, creating PayPal merely as a demo for it. But eBay sellers began using it to collect payments. A few months later, the founding team acknowledged that this peripheral tool was the actual business the company was in, even though it had never been their initial plan.

When great founders discover that users are not following the intended path and are instead "misusing" or "abusing" the product, they never feel annoyed. User abuse is the market's rawest, most urgent cry for help. It means users are confronting a pain point so acute that they are desperate enough to use an ill-fitting, makeshift tool just to survive. These unintended usage trajectories often harbor golden new directions.

Graham attributes PayPal's rapid ascent to a simple mechanism: it helped users make money. Once a product helps customers generate revenue, they adopt it much faster and are willing to pay higher prices, propelling company revenue from both adoption velocity and willingness to pay.

Moving Upstream: Why Selling to Tiny Customers Early on Builds Giant Trees

When shaping sales strategies, many startup teams instinctively fixate on mature enterprises with over 500 employees, assuming large clients yield high contract values and have deep pockets. Yet in commercial chess, the further upstream you position yourself toward the business's origin, the stronger your footing becomes.

When advising startups, Graham consistently poses a diagnostic question of immense practical value: "What is the smallest customer size that would buy your product?" He always hopes that threshold is as low as humanly possible. If the bar is high, he presses the team: is there any way to tweak the product so it can be sold at an earlier stage? The ideal scenario is being able to sign up a fellow cohort founder on the spot, just as Stripe's founders did (what he calls the Collison-install).

Selling early to nascent startups draws its power from a source that founders often find counterintuitive—after all, the earlier a customer is, the tighter their budget. But if you sign them at day one and charge on usage, your revenue will surge in lockstep with the startup's breakneck growth rate.

Behind this lies a broader commercial law: Having customers who make decisions lightning-fast makes a company extraordinarily powerful. This is not just because execution moves faster, but because fast-deciding customers tend to make decisions based purely on product quality. Startup teams generally build the best products (if a company is both tiny and mediocre, it simply cannot survive in the market); and in a space where customers decide rapidly, making the best product leads directly to making the most money. Conversely, if your product can only be sold once a company reaches 500 employees, you are relegated to a far more passive, disadvantaged position—bogged down in the protracted quagmire of enterprise sales, where it is notorious that the best product rarely wins.

Selling to institutions like hospitals or public school districts feels like trudging through knee-deep mud. In normal industries, customer sluggishness toward new technology presents a golden opportunity for startups—the slower customers move, the higher your odds of winning by going full stack and running the business yourself. What makes hospitals and school districts so soul-crushing, however, is that startups usually cannot go full stack there at all. When encountering teams trapped in such mud, Graham's diagnostic approach is: can you at least carve out a faster-deciding subset of the market first? In schools, you might even consider bypassing the institutional hierarchy altogether and serving students directly.

Stripe's Ultra-Early Lock-in Rule: Stripe has always insisted on signing customers at the earliest possible milestone. In payments infrastructure, as long as a system isn't broken, customers will never switch away; since Stripe works beautifully and reliably, companies that install it rarely churn. This fundamentally upends how you view customers: don't ask "what size customer should we target?", but rather "at what point in the customer's lifecycle should we acquire them?" Large enterprises are simply specific companies that have reached a later stage of development. If you are confident customers won't churn, why not lock them in early and grow alongside them? This is the core logic behind YC's emphasis on growth rate over absolute scale: if the growth rate is high enough, the absolute numbers will take care of themselves; and the path to the fastest growth rate is selling to customers who grow the fastest and decide the fastest. Viewed quarterly, prospective customers look like fixed quantities; but over decades, every company is an outward-expanding growth trajectory.

Rippling's Source Data Position: Similar to acquiring customers early, an equally potent strategy is acquiring data at the earliest possible moment. Rippling's vision has always been to become both the operating system for employee-data applications and the provider of most apps running on it. Rather than attempting to blueprint the final system from the outset, they started by building employee onboarding software, because an employee's data lifecycle begins at onboarding. Because they were aiming for a broader operating system ecosystem, they polished their onboarding tool far beyond what a point solution required, firmly locking down employee data at its absolute fountainhead.

The Unified Upstream Principle: Whether it concerns money, user relationships, customer life stages, or underlying data, "moving upstream" stems fundamentally from the same commercial architecture. Graham summarizes: Whether it involves money, user relationships, customer stage, or data, moving upstream is almost always beneficial. The closer you get to the origin of these critical elements, the harder it is for downstream links to constrain you.

Playing the Long Game and the Way of Generosity: Value Created Must Far Exceed Value Captured

Another covert path to formidable strength is making long-term bets that most mediocre competitors are either unwilling or unable to make.

In business competition, many of your direct rivals are founded by opportunists angling for a quick acquisition by a tech giant; meanwhile, the executives at mature incumbents you face often have tenures of only a few years, their eyes glued to quarterly earnings. This means deep architectural bets that take ten years to bear fruit are severely discounted in today's market.

Short-sighted operators fixate on squeezing every last cent out of existing customers—a nickel-and-dime strategy that yields at best a 2x bump in profits. Wholeheartedly exploring and creating entirely new capabilities for users, however, easily reaps exponential returns of 10x or 100x.

In a footnote, Graham notes that Tim O'Reilly's worldview of "creating more value than you capture" is far more common among founders. When a company pivots from building new products to squeezing more profit from existing operations, it is often because control has passed from founders to hired managers. In his view, this is partly because founders possess greater willingness or capability to create new things; more deeply, it stems from the fact that founders have personally tasted what it feels like to be weak. Hired CEOs easily take a company's existing power for granted, whereas founders still remember the days when their company was so frail that pleasing users was the only way to survive.

Open-source software is the classic paradigm of trading generosity for power. A company literally gives its software away for free, but this concession can turn the software into a standard while deepening user trust. This fuels widespread adoption, ultimately allowing the company to claim a smaller slice of a vastly larger pie.

If you don't wish to go fully open source, you can capture similar dividends by making your product "extensible." This extensibility spans a spectrum: on one end are app stores that set strict rules and extract commissions; but if you don't restrict extensions or charge developers, you may nurture an even larger ecosystem. In a footnote, Graham further speculates that this flexibility could become the key to another ecosystem superseding Apple, since he is convinced Apple will continue tightly restricting products that integrate with its hardware and software.

The highest manifestation of extensibility is making a product directly callable via API. Many companies resist opening APIs because they dread the accompanying loss of control, preferring external parties to use their software strictly according to preconceived scripts. Graham argues that, barring a few specialized domains, this defensive posture is usually a mistake. Especially today, as AI agents rapidly replace human operators, no one can foresee how future agents will invoke your capabilities. The earlier and more embryonic a startup is, the less it has to lose, and the more aggressively it should lean into open APIs.

To be clear, playing the long game and offering generous terms does not mean mindless subsidization. Offering compelling terms or even selling at a short-term loss to acquire early users can be sound tactics; but you must never subsidize excessively. If you sell a $10 bill for $5, the resulting top-line growth rate is completely meaningless, because you have sabotaged the critical demand signal that customers provide through cold, hard cash.

Sidestepping "Mafia" Battlegrounds and Overcoming Internal Self-Diminution

Startups that conquer markets through stellar product quality rely naturally on level playing fields where rules are transparent and merit prevails. Once you step into monopolistic sectors dominated by entrenched special interests, sheer product superiority often evaporates instantly.

Graham likens record labels and American PBMs (Pharmacy Benefit Managers) to "mafia"-like industry forces, noting in a footnote that both appear staffed with armies of lawyers. In such markets, having the best product does not guarantee victory; a startup's very existence depends on whether incumbents permit it. They are not invincible, but the viable approach is to enter obliquely—winning the market along another dimension first until the old moats lose relevance, rather than making a frontal assault on giants a prerequisite for success.

In exploring how to make a company more powerful, founders are often searching for how not to be held back by other companies. The text connects several scenarios of being held back with their corresponding escape strategies:

• Held back as a component supplier: If you are confined to being a parts supplier, you must live inside a box (sometimes literally) built by another company; the escape is finding ways to directly own the relationship with the end customer.
• Held back by bureaucratic enterprise customers: Try selling to them when they are still small and make decisions quickly.
• Held back by your original target customers: If conditions permit, leverage your own technology to enter that business directly and compete with them through full-stack execution.

This pattern is so pervasive that Graham suggests turning it into a core generative question for deriving ambitious ideas: "In what ways is the current idea being held back by other companies?"

To help teams systematically diagnose where their idea is being held back and map out an escape route, consult the power audit checklist below, organized by vulnerability type and strategic breakthrough:

Yet just as often, if not more so, startups are not held back by outside companies—they are held back by themselves.

Graham found that a shockingly high percentage of his advice to startups contains the word "Just": "You don't need to x. Just y." Startup ideas tend to tie themselves into knots and bloat with complexity for two main reasons: first, historical baggage from older directions—the idea evolved from prior iterations and still carries vestiges the team no longer needs, without founders even realizing it; but for very early startups, the deeper reason for complexity is usually internal fear. Subconsciously, teams flinch from confronting the grander, more ambitious, yet strikingly clean and simple true direction, retreating instead into a smaller, convoluted corner to comfort themselves.

In Graham's eyes, "Just y" essentially means: "Just stand up straight." When founders summon the courage to dispel fear and stand tall to confront their real ambition, the stature the company projects before the market instantly becomes vastly larger than before.

The Ultimate Iron Law: All Power Derives from Making Customers' Lives Better

No matter which power-reshaping tactic is deployed—whether commanding money flows, introducing network effects, or swallowing customers to go full stack—it must unconditionally obey an immutable iron law: it must objectively make the end customer's life better.

Founders cannot simply graft network effects onto a product or unilaterally impose a toll bridge just because they crave leverage. If the end result of any power restructuring fails to deliver lower costs, higher efficiency, or a superior experience for end users, the market will instantly vote with its feet, and adoption will be zero.

The initial frailty of startups explains why they are, on the whole, a force for good in the world: a newborn company lacks the power to force anyone to adopt its product. The only way it can amass power is by making customers' lives better.

This hard constraint is not merely a litmus test for business models; it serves as an ideation compass for reverse-engineering breakthrough concepts: From the customer's perspective, what does an absolute, ideal world look like? If there is a component of that vision that the company could transform itself into, it is well worth seriously exploring reinventing the business in that direction.