On-chain behemoth war: Who will control the flow of value and become the new oligopoly in the Web3 industry?

Written by: Saurabh Deshpande, Decentralised.co

In November 2023, Blackstone Group acquired a pet care app called Rover. Initially, Rover was just for finding people to walk dogs or care for cats. The pet care industry typically consists of tens of thousands of small, mostly localized, and offline service providers. Rover integrated these suppliers into a searchable marketplace, adding review and payment features, making it the default platform for pet care services. By the time Blackstone privatized it in 2024, Rover had become the hub of demand in this field. Pet owners think of Rover first, and service providers have no choice but to list on this platform.

ZipRecruiter has done something similar in the recruitment field. It collects job information from employers, job boards, and applicant tracking systems and distributes it across multiple channels. ZipRecruiter posts job openings on social networks like Facebook. For employers, ZipRecruiter becomes a one-stop distribution channel; for job seekers, it is a unified entry point to the market. ZipRecruiter does not own companies or positions, but rather has relationships with both parties. Once this relationship is solidified, it can charge for visibility and job matching, which is an introductory lesson in aggregation economics.

Aswath Damodaran refers to this model as "shelf ownership": consolidating chaotic and fragmented supply, controlling its presentation, and charging for access. Ben Thompson calls it the "aggregation theory": establishing a direct relationship with end users, allowing suppliers to compete to serve them, and extracting value from each transaction. The core characteristics across different fields are consistent: Google with web pages, Airbnb with listings, Amazon with products.

The Amazon flywheel is a classic interpretation of this concept. During the downturn after the internet bubble burst, Jeff Bezos and his team drew inspiration from Jim Collins' "flywheel" concept to create a cycle that every MBA can now recite: more choices lead to a better customer experience, attracting more traffic, which in turn attracts more sellers, lowering unit cost structures, thereby providing lower prices, ultimately leading to more choices. The effect of turning the flywheel once is limited, but after turning it a thousand times, the machine begins to roar. Bezos' motto during this period was: "Your profit is my opportunity." The core of this is self-reinforcement: more users, more suppliers, lower costs, ultimately achieving higher profits.

Once this model proves effective, it can be considered perfect. The growth rate of costs is far lower than that of revenue, and the product will be optimized as the number of users increases. However, it only holds under two conditions: the aggregated content must have value, and the suppliers must find it difficult to exit easily; both are indispensable, otherwise, the moat will become shallow. Taking eBay as an example, it aggregated millions of unique niche sellers and buyers in the early 21st century. This aggregation was once highly valuable, but when sellers realized they could build their own stores on Shopify or shift to Amazon, they began to leave. The flywheel doesn't stop turning overnight, but if the suppliers are no longer controlled, it starts to wobble and ultimately becomes ordinary.

Damodaran explains the power of platforms and aggregators in a concrete way. He mentions "controlling the shelf", which is not literally about supermarket shelves, but rather the space that customers first encounter when demand arises. Controlling this space means deciding what content to display, how to display it, and the cost of entry. You don’t need to own the products themselves, just the relationship with the buyers, and others must go through you to reach the buyers. In analyzing Instacart, Uber, Airbnb, or Zomato, Damodaran repeatedly emphasizes that the aggregator's task is to consolidate a chaotic, fragmented market into a single glass window and make that window the only one worth paying attention to. Once that is achieved, you can charge for the "viewing rights".

Ben Thompson believes that aggregators are businesses that establish direct relationships with end users at an internet scale, providing standardized and reliable experiences, while allowing suppliers to compete to serve them. At an internet scale, you are not the biggest store in town, but rather a store that covers all towns at once.

The marginal cost of serving the next customer is nearly zero, but their marginal value is immense. Each customer strengthens your brand, data, and network effects. As aggregators control demand, suppliers become interchangeable. This does not mean there is no difference in quality, but rather that suppliers cannot take customer relationships with them when they leave. Hotels on Expedia, drivers on Uber, sellers on Amazon, they all need each other more than the aggregator.

Damodaran's research reminds us that the flywheel does not operate the same way in all markets. For example, Uber aggregates local driver liquidity, but drivers can open three apps simultaneously and choose the first order they receive. This creates vulnerabilities in the moat. In contrast, Airbnb hosts offer unique listings with limited alternative channels, making their commissions more sustainable.

In low-margin areas, shelves may have value, but the commission space is limited, and suppliers can easily push back. This is why Instacart must venture into advertising and white-label logistics for growth.

The economic structure of supply is just as important as the number of users on the platform. If the goods on the platform are readily available everywhere, you are just a convenience store with a better view; but if the content is scarce, differentiated, and hard to substitute, people will continue to visit, even if you charge higher fees. Think of the high-end listings on Airbnb.

Why did the aggregator fail?

When conditions are missing, the aggregator is no longer a flywheel, but just a costly carousel.

Quibi is a typical case of failing to control the shelf. The platform has expensive Hollywood content and a beautiful app but lacks direct channels to users. Potential users have already gathered on YouTube, Instagram, and TikTok. These platforms control attention, while Quibi locks its content in a standalone app, away from users, causing it to rely solely on ads and promotions to attract them.

Excellent aggregators start with zero marginal cost user reach methods, such as built-in distribution, installation volume, or daily habits. Quibi had nothing and ultimately exhausted time and funds before building these.

Facebook's Instant Articles also faces similar issues. Its concept is to aggregate content from publishers, load it faster within Facebook, and monetize traffic. However, publishers can easily distribute content to their open networks, apps, or other social platforms. Instant Articles has never become the default reading platform, just an option within the information stream.

Both examples violate the same rule: the enterprise fails to have a user relationship in a way that creates default behavior, and the supplier will not be significantly harmed after exiting.

The list of excellent aggregators is simple:

Directly connect and have user relationships;

Suppliers must either be unique or replaceable so that they are not held hostage by a single supplier.

The marginal cost of increasing supply is close to zero or low enough to optimize the business model with scale.

If these conditions are not met, you are just another easily replaceable intermediary.

How liquidity becomes a moat

In the cryptocurrency industry, projects can build moats in various ways. Some establish trust through licenses and regulation (such as USDC), some rely on technology (such as Starkware's proof system or Solana's parallel execution), and others depend on community and network effects (such as Farcaster's user graph). However, the hardest to shake is liquidity.

"Proper execution" is crucial. However, if the incentives are strong enough, liquidity can shift rapidly. In 2020, Sushiswap siphoned over $1 billion in funds from Uniswap in just a few days through liquidity mining rewards. The lesson is simple: liquidity only stabilizes when leaving is more painful than staying.

Hyperliquid is well-versed in this area. It not only builds the deepest order book for perpetual contract exchanges but also allows other applications and wallets to directly access its liquidity. For example, Phantom can access Hyperliquid's order flow, providing users with narrow spreads without needing to build their own markets. In this model, aggregators need suppliers even more. When traders and applications default to your routing, you are no longer an ordinary aggregator, but a core channel they cannot avoid.

In addition to its own platform, Hyperliquid processed over $13 billion in transaction volume through other builders last month. Phantom processed $3 billion in transaction volume through its routing, earning over $1.5 million. This demonstrates Hyperliquid's current strong network effects.

Liquidity allows you to convert assets without affecting prices. In the financial and DeFi sectors, deep liquidity makes trading cheaper, lending safer, and derivatives possible. A lack of liquidity can turn even the most perfect protocol into a ghost town. Once successfully established, liquidity tends to persist. Traders and applications will flow into deep pools, further increasing liquidity, narrowing spreads, and attracting more trades.

This is the reason why protocols like Aave remain strong and enduring. Aave has a large-scale lending pool of various assets, making it the preferred choice for borrowers and lenders pursuing scale and security. As of August 6th, Aave's total cross-chain locked value exceeds $24 billion. Over the past 12 months, borrowers have paid $640 million in fees, with platform revenue around $110 million.

The aggregator Jupiter, also based on Solana, has evolved from a routing tool to the default entry point for transactions on the network. On Ethereum, Uniswap has centralized most of the spot liquidity, so aggregators like 1inch can only provide marginal improvements. In contrast, on Solana, liquidity is dispersed across platforms such as Orca, Raydium, and Serum. Jupiter integrates this into a single routing layer, always providing the best prices. At one point, its trading volume accounted for nearly half of all computing usage on Solana, and any delays or interruptions would immediately affect the execution quality across the entire network.

Viewing liquidity as an aggregated object makes Jupiter's product decisions easier to understand. Acquisitions, mobile applications, and expansions into new trading and lending products are all aimed at capturing more order flow, maintaining liquidity through the Jupiter routing, and solidifying its position.

Jupiter is worth paying attention to, as it is a clear case of evolving from a niche tool to a liquidity platform in DeFi. It started with the search for the best spot prices and gradually became the default route for Solana liquidity, subsequently expanding to attract new liquidity products. Observe how it progresses through these stages and reinforces each other, providing a vivid example for aggregating dynamics.

Aggregated hierarchy

The three questions are a quick checklist for identifying potential aggregators:

What are the key differentiating factors of existing enterprises? Can they be digitized? In DeFi, the differentiating factor is liquidity. Deep pools can provide narrower spreads and safer loans. Liquidity has been digitized, making it easy to read and compare.

If differentiated factors are digitized, does competition shift towards user experience? When liquidity can be accessed at will, competition revolves around execution quality: faster settlement, better routing, and fewer failed transactions. Products like BasedApp and Lootbase are born from this. The former encapsulates DeFi primitives into a smooth mobile experience, while the latter brings Hyperliquid's deep perpetual liquidity to mobile.

If winning the user experience can create a virtuous cycle? Traders come for better prices, attracting more liquidity, which in turn provides better prices. When liquidity becomes habitual and integrated, it becomes sticky.

Become the default entry point of the market. If the suppliers cannot bear your absence, you can collect display fees or determine the order flow in DeFi.

Note: The boundaries between different levels are often blurred. Classification is not precise, but rather provides a thinking model for aggregation levels.

First Level: Price Discovery

This is the most basic work: telling people where the best trades are. Kayak is for flights, Trivago is for hotels. In the crypto space, early DEX aggregators like 1inch or Matcha fall into this category. They check available pools, display the best rates, and provide jump-in access. Price discovery is useful but fragile, and DeFiLlama's exchange function is the same.

If the underlying market is already concentrated (such as Ethereum spot trading on Uniswap), the effect of routing improvements is minimal, and users can go directly to the trading venue; the assistance you provide is not essential.

Second Level: Execution

At this point, you no longer direct users elsewhere, but operate on their behalf. Amazon's "one-click purchase" belongs to this level. In DeFi, Aave's lending function is located at this layer. The liquidity already exists in its contract when borrowing. The execution increases stickiness because the outcome is directly related to you: quick settlement and a good experience with no failed transactions.

Third Level: Distribution Control

You become the entry point. Just as Google Search is to web pages, the app store is to mobile applications, both belong to this category. In the crypto space, the built-in exchange tag in wallets can serve as the starting and ending point for ordinary users.

On Solana, Jupiter has reached this level. It started as a price discovery tool, transitioned into the execution layer through smart order routing, and then embedded into front-ends like Phantom and Drift. A large number of Solana trades are actually Jupiter trades, even if users have never entered "jup.ag". This is distribution control, where suppliers cannot bypass you to reach users.

Climbing the tiers in DeFi

The challenge of DeFi lies in the rapid transfer of liquidity that can occur. Incentives can drain liquidity pools overnight. Therefore, moving from the first tier to the third tier is not only about becoming a leading aggregator, but also about creating sufficient reasons for liquidity and order flow to continuously pass through your routing.

On Ethereum, 1inch mainly operates on the second layer because Uniswap has completed aggregation work through centralized liquidity. Routing for edge cases still holds value, but improvements are limited, and many traders choose to skip it. Additionally, aggregators like CowSwap and KyberSwap also hold a significant share. Aave belongs to the second layer as it controls execution in the niche, but it is infrastructure, not a starting point.

The advantage of Jupiter on Solana lies in its ability to climb three tiers sequentially. Liquidity is decentralized, with significant value in the first tier; the routing engine surpasses manual exchanges, naturally transitioning to the second tier; by directly integrating wallets and dApps, it reaches the third tier, fully controlling the distribution of Solana's liquidity. At one point, nearly half of Solana's computing was driven by Jupiter transactions, as both the demand side from traders and the supply side from liquidity pools rely on Jupiter.

After reaching the third level, the question becomes "What else can be run through this distribution?" Amazon started with books and ended with everything; Google started with search and ultimately controlled maps, email, and cloud computing. For Jupiter, distribution is the order flow. The obvious next step is to add products such as perpetual contracts, lending, and portfolio tracking, leveraging the same liquidity relationships.

A larger move is Jupnet. Solana has yet to match the throughput and execution characteristics of venues like Hyperliquid, which are specifically designed for financial-grade latency and determinism. These characteristics are crucial for scaling the entire financial stack to real-world sizes. A simpler option would be to launch products on chains that already possess these features, but Jupiter has chosen the more difficult path, building Jupnet as an application-controlled low-latency execution layer that runs in parallel with Solana.

Jupnet aims to become a shared infrastructure within the Solana ecosystem, supporting latency-sensitive transactions such as perpetual contracts, quote request systems, and batch auctions, with final native settlement on Solana. If successful, it will provide the expected speed and certainty of vertically integrated venues while maintaining user and asset retention. This is an attempt to bridge the gap between general blockchain throughput and the global financial micro-latency demand, without the need to split liquidity across chains.

However, it should be noted that although Jupiter dominates within Solana, the industry still faces fierce competition. In the cross-chain field, 1inch, CoWSwap, and OKX Swap maintain significant positions. As of 2025, Jupiter's average share among the top five DEX aggregators is about 55%, but this share fluctuates with on-chain activity and integrations. The chart below shows the degree of decentralization of the aggregation layer outside of Solana.

Clearly, Jupiter has become an aggregator within the Solana ecosystem. The flywheel has started: more traders bring more liquidity, more liquidity optimizes execution, and better execution attracts more traders. At this point, you are not only a liquidity aggregator but also a shelf, a habit, and a gateway to the market. So, when liquidity is no longer sufficient, how do you continue to grow? Jupiter's answer is to acquire projects that have taken control of the new user flow.

Mergers and acquisitions as a growth engine

Previously, I wrote an article discussing two major themes of enterprise scaling: the essence of composite innovation and how companies can accelerate this process through mergers and acquisitions. The former relates to building new products, functionalities, or capabilities based on existing advantages, while the latter concerns identifying when "buying" is faster than "building" to establish an advantage.

The evolution of Jupiter combines both aspects. Its acquisition strategy is rooted in seeking founder teams with tangible appeal and integrating them into a distribution network that amplifies impact. The company looks for expert teams in vertical fields to expand coverage without compromising its core roadmap.

This is not just an enhancement of the purchasing function, but an acquisition of teams that have dominated the target market segments of Jupiter. Once these teams integrate into Jupiter's distribution wallet interface, API, and routing, their product growth accelerates, and the generated traffic feeds back into Jupiter's core.

Moonshot brings a token launchpad, transforming the creation of new tokens into direct exchange and trading activities within the Jupiter ecosystem; DRiP adds a community-driven NFT minting and distribution platform, attracting audiences away from trading interfaces and converting them into on-chain behaviors; Portfolio acquisition provides position management tools for active traders. Jupiter could have built these features internally at a lower cost, but its goal is to acquire founders, not just functionality.

However, the growth of some indicators has not yet been evident. For example, in the launchpad sector, market leaders Pumpdotfun and LetsBonk control over 80% of daily token issuance, while Jup Studio and Moonshot together account for less than 10%. The chart below shows the dominance of the incumbents. In this case, the default pattern may have already solidified, and Jupiter may need a completely different approach to break through.

Power Amplifier: Founder-led M&A Acquisition

To expand the shelf, it is necessary to introduce operators who have mastered the target market segment. Jupiter's selection criteria are: does the team bring new liquidity or users that strengthen the flywheel? This logic echoes Amazon's early flywheel: each new category or supplier expands "selection", optimizes customer experience, drives more traffic, and in turn attracts more suppliers.

For Jupiter, each acquisition is like adding new shelves to a store, broadening options and deepening the relationship between traders and liquidity providers.

Acquire creative founders to help Jupiter penetrate unfamiliar territories (such as the NFT culture of DRiP or mass retail token issuance) while not diluting its core competitiveness. These founders understand niche markets, have communities that trust them, and can act quickly. Tapping into Jupiter's distribution channels can exponentially increase its reach overnight, while Jupiter gains new user flows and liquidity.

Acquisition cases reflect this: Moonshot is a minting and trading platform aimed at mainstream behavior, whose issued tokens can be seamlessly transferred into the exchange, capital market, and perpetual contracts within the Jupiter ecosystem; DRiP is a creator-first collectibles distribution channel that attracts communities that would otherwise not engage with trading interfaces.

Moonshot gained over 250,000 users within three days of the launch of the TRUMP token, processing over 1.5 billion USD in transaction volume; DRiP attracted over 2 million collectors, minting over 200 million collectibles, with over 6 million secondary sales.

Integration follows a clear pattern: the founders retain control over product direction; the product goes live and immediately connects to the Jupiter interface and backend, benefiting from its user base, while Jupiter gains new traffic; each acquisition adds unique liquidity primitives (such as issuance, culture, leverage), rather than duplicating existing features. Core competencies remain unchanged, and all paths still lead back to Jupiter.

In DeFi, code can fork overnight, but market share is hard to replicate. Founder-led mergers and acquisitions allow Jupiter to gain market share without losing its core path, making its flywheel harder to replicate. As application-controlled execution and low-latency infrastructure mature, Jupiter may target teams such as risk engines, matching layers, and specialized venues, integrating them into Jupnet.

Aggregator vs Supplier

Looking at the big picture, two dominant models are emerging in DeFi: Jupiter and Hyperliquid. Both are powerful, but their strategies are completely different.

Hyperliquid aims to control liquidity rather than directly own end-user relationships. It provides liquidity as a service. If a better user experience can be constructed, feel free to use Hyperliquid's order book and execution engine. Builder Codes are based on this concept, allowing others to own the front-end experience while Hyperliquid quietly supports the back-end, which is the supplier-first model.

Jupiter focuses on distribution, aiming to have interfaces, shelves, and market entry points, aggregating decentralized liquidity by becoming the default interface and directing it where needed. This means controlling user relationships, not just executing transactions. From perpetual contracts to portfolios, Jupiter strives to make all financial interfaces begin and end within its ecosystem.

However, perpetual contracts may most expose the current limitations of this strategy. Jupiter has made progress on Solana, but Hyperliquid still dominates the market with about 75% of the perpetual DEX market share. The following figure shows the leading margin of Hyperliquid in original trading volume:

Both models bet on the scale, but start from opposite ends. Jupiter believes that liquidity follows the user interface; Hyperliquid believes that liquidity is the interface. Jupiter builds the entry, while Hyperliquid builds the endpoint.

In practice, we witness differentiation: if you need a broad interface and user aggregation, choose Jupiter; if you need depth, certainty, and composability, choose Hyperliquid. One side transforms liquidity into a network of dependency, while the other becomes the underlying structure built by the community.

The winner is not only the one who arrives first, but also the one who others cannot abandon.

This is exactly what makes DeFi so exciting right now. We are witnessing a philosophical showdown for the first time: one side believes distribution is the moat, while the other firmly believes it is liquidity.

The application is the new platform

When Ethereum Layer 2 first emerged, people hoped it would become a new platform: a neutral ground where applications could be composable, competitive, and scalable. However, it turned out that L2 did not become the platform as imagined, but rather remained at the infrastructure level: providing the technical foundation for speed, security, and scalability, yet failing to control user relationships.

The platform is the interface at the starting point of the user journey, where demands are aggregated, habits are formed, and distribution survives. Few L2s cross this line, most are pipelines rather than shelves, and there is rarely meaningful distribution, even less so becoming the default entry point for users.

In contrast, applications like Jupiter and Hyperliquid are gradually revealing platform characteristics. They have user relationships, are embedded in daily habits, and strengthen their positions through acquisitions or integrations with other applications. In fact, they are starting to resemble Web2.

Google has gone beyond search engines by acquiring YouTube, transforming its search advantages into video dominance; Facebook expanded its control over attention by acquiring Instagram and WhatsApp. They target adjacent fields where they were absent but users have already congregated, and the key is acquiring the core players in these areas. Once acquired, these applications can immediately tap into Google's and Facebook's existing distribution flywheel, resulting in the capture of multi-channel user attention.

Jupiter is currently implementing a similar strategy. Launchpad, NFT minting tools, portfolio managers, and now Jupnet all serve the same purpose: to expand coverage, capture more user behavior, and route more liquidity to itself. Its strategy is to become the shelf, the default choice, and the starting point for financial interactions.

But aggregation is not a surefire strategy. History is full of failed platform acquisitions and aggregation attempts, either due to a lack of user relationships or a misunderstanding of how habits are formed.

Taking Microsoft's acquisition of Nokia as an example. This is a bet on controlling mobile distribution, but users have shifted to the iOS and Android ecosystems. Microsoft has hardware and software, but its mobile devices and operating systems are either too similar to existing products or insufficient to motivate users to switch. It did not control the application layer, did not win developer loyalty, nor provided a reason to change behavior. The lack of control over supply or clear differentiation left the shelves untouched.

These cases reflect the core truth: an acquisition itself does not create a flywheel. Without a starting point, habit, or interface, users will not follow, no matter how many features are bundled.

This makes DeFi particularly interesting at the current moment. Jupiter is acquiring front-end, distribution channels, and liquidity primitives in an attempt to become the default entry point for the Solana financial stack; Hyperliquid, on the other hand, takes the opposite approach: building depth rather than breadth, allowing others to build around its portfolio.

In a sense, we are witnessing a real platform war unfolding between applications, rather than between public chains as many anticipated. This raises a bigger question: if L2 does not control distribution, where will the value flow when the applications on it do? What about the fat protocol?

We conclude with unresolved issues, as there are still no definitive conclusions. In the future, we will bring sharper perspectives, new data points, and more stories and analogies to clarify the direction of all this.