Aave Founder: What is the Secret of the DeFi Lending Market?

marsbitPubblicato 2026-02-10Pubblicato ultima volta 2026-02-10

Introduzione

Chain-based lending, which began as an experimental concept around 2017, has evolved into a market exceeding $100 billion, primarily driven by stablecoin borrowing backed by crypto-native collateral like Ethereum and Bitcoin. This system enables liquidity release, leveraged strategies, and yield arbitrage. The key advantage of on-chain lending lies not in technological novelty but in its elimination of financial inefficiencies, offering lower costs (around 5% for stablecoins) compared to centralized crypto lenders (7-12%) due to open capital aggregation, transparency, and automation. On-chain lending is structurally due to permissionless markets that excel in capital pooling and risk pricing, fostering competition and innovation without intermediaries. This model reduces operational costs, replacing manual processes with code, and benefits both capital providers and borrowers. However, the current limitation is not a lack of capital but a shortage of diverse, borrowable collateral. The future of on-chain lending depends on integrating real-world economic value with crypto-native assets, moving beyond abstract financial strategies to serve broader adoption. Traditional lending remains expensive due to inefficiencies in loan origination, risk assessment, and servicing, where misaligned incentives and manual processes inflate costs. Decentralized finance can disrupt this by automating end-to-end operations, ensuring transparency, and reducing expenses. When on-chain lending ...

Author: Stani.eth

Compiled by: Ken, Chaincatcher

On-chain lending began around 2017 as a fringe experiment related to crypto assets. Today, it has grown into a market exceeding $100 billion, primarily driven by stablecoin lending, largely collateralized by crypto-native assets like Ethereum, Bitcoin, and their derivatives. Borrowers use it to release liquidity through long positions, execute leverage loops, and engage in yield arbitrage. The key is not creativity, but validation. The behavior over the past few years has shown that automated lending based on smart contracts had genuine demand and true product-market fit long before institutions began to take notice.

The crypto market remains volatile. Building a lending system on top of the most dynamic existing assets forces on-chain lending to immediately address risk management, liquidation, and capital efficiency issues, rather than hiding them behind policies or human discretion. Without crypto-native collateral, it would be impossible to see just how powerful fully automated on-chain lending can be. The key is not cryptocurrency as an asset class, but the cost structure transformation brought about by decentralized finance.

Why On-Chain Lending is Cheaper

On-chain lending is cheaper not because it's new technology, but because it eliminates layers of financial waste. Today, borrowers can access stablecoins on-chain at a cost of around 5%, while centralized crypto lending institutions charge interest rates of 7% to 12%, plus fees, service charges, and various surcharges. When conditions favor the borrower, choosing centralized lending is not only not conservative, it's irrational.

This cost advantage does not come from subsidies, but from capital aggregation in an open system.Permissionless markets are structurally superior to closed markets in pooling capital and pricing risk, because transparency, composability, and automation drive competition. Capital moves faster, idle liquidity is penalized, and inefficiencies are exposed in real-time. Innovation spreads immediately.

When new financial primitives like Ethena's USDe or Pendle emerge, they absorb liquidity from the entire ecosystem and expand the use of existing primitives like Aave, all without sales teams, reconciliation processes, or back-office departments. Code replaces management costs. This is not just an incremental improvement; it is a fundamentally different operating model. All cost structure advantages are passed on to capital allocators and, more importantly, to borrowers.

Every major shift in modern history has followed the same pattern. Heavy-asset systems become light-asset systems. Fixed costs become variable costs. Labor becomes software. Centralized scale replaces local duplication. Excess capacity is converted into dynamic utilization. Changes initially look bad. They serve non-core users (e.g., lending for cryptocurrency, not mainstream use cases), compete on price before quality improves, and don't look serious until they scale to a point where incumbents can't cope.

On-chain lending fits this pattern perfectly. Early users were mostly niche cryptocurrency holders. The user experience was poor. Wallets felt alien. Stablecoins didn't touch bank accounts. None of that mattered because the cost was lower, execution was faster, and access was global. As everything else improved, it became more accessible.

What Happens Next

During bear markets, demand falls, yields compress, revealing a more important dynamic. Capital in on-chain lending is always in competition. Liquidity does not stagnate due to quarterly committee decisions or balance sheet assumptions. It is constantly repriced in a transparent environment. Few financial systems are as relentless.

On-chain lending does not lack capital, it lacks collateral available for lending. Most on-chain lending today just recycles the same collateral for the same strategies. This is not a structural limitation, but a temporary one.

Cryptocurrency will continue to generate native assets, productive primitives, and on-chain economic activity, thereby expanding the scope of lending. Ethereum is maturing into a programmable economic resource. Bitcoin is solidifying its role as an economic energy store. Neither is a final state.

If on-chain lending is to reach billions of users, it must absorb real economic value, not just abstract financial concepts. The future lies in combining autonomous crypto-native assets with tokenized real-world rights and obligations, not to replicate traditional finance, but to operate it at an extremely low cost. This will be the catalyst for replacing the backend of old finance with decentralized finance.

What's Wrong with Lending

Lending is expensive today not because capital is scarce. Capital is abundant. Quality capital clears at 5% to 7%. Risk capital clears at 8% to 12%. Borrowers still pay high interest rates because everything surrounding capital is inefficient.

The loan origination process is bloated with customer acquisition costs and lagging credit models. Binary approvals cause quality borrowers to overpay, while subprime borrowers receive subsidies until they default. Servicing remains manual, compliance-heavy, and slow. Incentives are misaligned at every layer. Those who price risk rarely actually bear it. Brokers don't bear default risk. Loan originators sell exposure immediately. Everyone gets paid regardless of the outcome. The flaw in the feedback mechanism is the real cost of the loan.

Lending has not been disrupted because trust trumps user experience, regulation stifles innovation, and losses mask inefficiencies until they explode. When lending systems fail, the consequences are often catastrophic, reinforcing conservatism over progress. As a result, lending still looks like an industrial-era product clumsily grafted onto digital capital markets.

Breaking the Cost Structure

Unless loan origination, risk assessment, servicing, and capital allocation become fully software-native and on-chain, borrowers will continue to overpay, and lenders will continue to rationalize these costs. The solution is not more regulation or marginal UX improvements. It is breaking the cost structure. Automation replaces processes. Transparency replaces discretion. Certainty replaces reconciliation. This is the disruption decentralized finance can bring to lending.

When on-chain lending becomes demonstrably cheaper end-to-end than traditional lending, adoption is not a question, it is inevitable. Aave exists in this context, poised to serve as the foundational capital layer for a new financial backend, serving the entire lending landscape from fintech companies to institutional lenders to consumers.

Lending will become the most empowering financial product, simply because the cost structure of DeFi will allow fast-moving capital to flow into the applications that need it most. Abundant capital will create abundant opportunity.

Domande pertinenti

QWhat is the core reason why on-chain lending is cheaper than traditional lending according to the article?

AOn-chain lending is cheaper not because it is new technology, but because it eliminates layers of financial waste. Its cost advantage comes from capital aggregation in an open system, where transparency, composability, and automation drive competition.

QWhat does the article identify as the main limitation for the growth of on-chain lending?

AThe main limitation is not a lack of capital, but a lack of borrowable collateral. Most current on-chain lending recycles the same collateral for the same strategies, which is a temporary constraint.

QHow does the article describe the fundamental operational difference that DeFi's cost advantage is built upon?

AThe advantage is built on a fundamentally different operating model where code replaces management costs. Automation replaces processes, transparency replaces discretion, and determinism replaces reconciliation, breaking the traditional cost structure.

QWhat future development does the article suggest is necessary for on-chain lending to reach billions of users?

ATo reach billions of users, on-chain lending must absorb real economic value, not just abstract financial concepts, by combining autonomous crypto-native assets with tokenized real-world rights and obligations.

QAccording to the article, why is traditional lending expensive despite capital being abundant?

ATraditional lending is expensive because everything surrounding the capital is inefficient. The processes of loan origination, risk assessment, and servicing are bloated with costs, lagging models, manual work, compliance burdens, and misaligned incentives.

Letture associate

Rubin Ultra Makes Major Cuts, Even Nvidia Can't Handle Memory Price Hikes?

NVIDIA's Rubin Ultra, the top-tier variant of the newly announced Rubin AI accelerators, has reportedly seen significant specification downgrades, according to an industry report from SemiAnalysis. Initially designed with four compute dies (4-die), the Rubin Ultra is now said to be reduced to a 2-die design. Key changes highlighted in the report include: * **No increase in peak theoretical compute performance**, remaining at 35 PFLOPs like the standard Rubin. * **Severe reduction in memory capacity** to 192GB using 8-Hi HBM stacks, which is less than the standard Rubin's 288GB using 12-Hi stacks. * **Negligible memory bandwidth improvement** of only 1 TB/s. * **Slightly higher chip-level power consumption**. * The **primary upgrade is a massive increase in scale-up interconnect capacity**, supporting connections for up to 576 GPUs via NVLink, compared to 72 for the standard Rubin. The report suggests the redesign is primarily a cost-optimization move driven by the sharp rise in HBM (High-Bandwidth Memory) prices. By reducing the expensive HBM content and shifting investment towards enhanced system-scale networking, NVIDIA aims to maintain the platform's value for large-scale AI training clusters while managing soaring material costs. The news reportedly triggered a sell-off in South Korean memory stocks, with SK Hynix and Samsung shares falling around 8%, as markets grew concerned that NVIDIA—a major HBM buyer—might be reducing its reliance on high-capacity memory, potentially capping future pricing power for memory makers.

Odaily星球日报12 min fa

Rubin Ultra Makes Major Cuts, Even Nvidia Can't Handle Memory Price Hikes?

Odaily星球日报12 min fa

Can Generative Models Finally Be Trained End-to-End? The Core Is a For Loop

This article introduces a novel training paradigm for generative models called Explorative Modeling (XM), which enables true end-to-end training. Traditionally, powerful generative models like autoregressive and diffusion models are not trained end-to-end. They are trained to predict a single small step but require iterative multi-step sampling for inference. This "exposure bias" leads to error accumulation and limits performance. The core challenge XM addresses is "mode blurring." In generative tasks, a single input (e.g., "generate a dog") corresponds to many valid outputs (multiple modes). Standard training objectives like reconstruction loss force the model to average these modes, producing unrealistic, blurry outputs. To avoid this, existing models break generation into many small, almost deterministic steps, sacrificing end-to-end training. XM tackles this by restructuring the training loop itself. Its key insight is to amplify "generative expressivity." For each training input, instead of generating one sample, the model generates K candidate outputs. Only the candidate closest to the real data is used for computing the loss and updating the model via backpropagation. This simple "best-of-K" mechanism is implemented as a short for-loop. By exploring multiple possibilities, the model learns to distribute its guesses across different modes rather than collapsing to their uninformative average. The paper demonstrates that "exploration" acts as a new, powerful scaling axis. Gains from XM increase with model size, data scale, and compute. Experiments show improvements in FID scores for image generation and significant efficiency gains, sometimes outperforming larger models without exploration. When pushed to the limit, XM enables fully single-step, end-to-end generative models. In robotics tasks, an "Explorative Policy" matched the performance of a 100-step Diffusion Policy with a single forward pass, drastically improving inference speed. While the best-of-K concept is not entirely new, the authors' contribution lies in formally understanding it as a direct method to boost generative expressivity without fragmenting the generation process. This work suggests that as models scale, enhancing exploration during training may become crucial for overcoming fundamental performance bottlenecks.

marsbit1 h fa

Can Generative Models Finally Be Trained End-to-End? The Core Is a For Loop

marsbit1 h fa

Trading

Spot
活动图片