Hi all, happy Tuesday. While crypto prices have shown a modest recovery in the short term, negative breadth and relative underperformance to TradFi benchmarks on the monthly suggests a prevailing bearish environment. Zooming out, we look at the secular growth trends within stablecoins and lending as durable sectors for multi-year growth, and recap some of the recent beef/drama between Kamino and Jup Lend within this sector. Enjoy!

Crypto markets remain rangebound after November’s harsh selloff. The past week has shown a modest recovery, with BTC at $90.4K now being 12% off of its recent correction low of $80.7K. Over the trailing 24 hours, the AI and Modular sectors were the top winners, with TAO (+6.4%)  and TIA (+6.2%) as notable contributors to this short-term strength. The Perp Index was the top loser, with DYDX (-3.1%) and HYPE (-0.6%) accounting for the sector’s weakness.

Zooming out to the monthly, the picture remains unfavorable. TradFi benchmarks like Gold, Nasdaq and the S&P 500 are all green over the past month, while every crypto index we track is measuring negative returns.

Breadth is decisively negative on the monthly for all crypto indices, with no safe haven provided. Notably, the Protocol Revenue index is the top performing among crypto sectors, suggesting relative strength in protocols with strong fundamental positioning.

Despite recent strength, it remains to be determined whether this rally is countertrend to a prolonged downtrend, or if the low is in for the rest of 2025.

— Luke

Amid the downtrend, don’t lose sight of the bigger picture, and take a moment to appreciate just how far we’ve come. At the bear market low, lending applications within DeFi accounted for just $5B in deposits, a rounding error within the larger financial system. In the years since, this figure has grown to $71B in deposits, having just previously tagged near $100B.

Similarly, loans outstanding on these applications amounted to $1.6B at the bear market low. Since, active loans have risen to over $27B now, and recently passed $38B. These figures have grown to a scale worth paying attention to, not just for the crypto native. Variance can cut both ways, to the upside and to the downside, for both key metrics and spot prices. But 20x growth in topline metrics in just a few years suggests this sector is a long way off from its terminal growth rate. 

Amid the drawdown in crypto prices, the aggregate stablecoin supply has retraced back up to its all-time high of $310B, after a brief period of -$10B in outflows. The growth in stablecoins is a trend that I would not bet against. Onchain money markets stand as primary beneficiaries of the growing stablecoin supply, purporting to be the primary applications for utilization. Stablecoin lenders seek yield, while borrowers look for leverage. While spot assets will remain volatile, this secular trend should remain a persistent and durable tailwind for DeFi applications.

Notably, continued growth in the stablecoin supply comes amid very modest onchain yields. Benchmark supply rates on lending applications are reading 3.6%, a discount to SOFR at 3.9%. Growing depth in onchain stablecoin liquidity may continue to dampen the upside in supply rates, and absent a material risk premium to legacy rates, growth in stablecoin utilization on money markets may falter.

— Luke

How are DeFi and traditional rails actually converging?

Join this live Roundtable to hear voices from Blockdaemon, Aave, and Circle hash it out!

Kamino and Jupiter Lend

Over the past several days, the exchange between Kamino and Jupiter has escalated from healthy competition to a clear public dispute. The events started on Nov. 27, when Jup Lend introduced a refinancing tool on its frontend to migrate looping positions from Kamino Multiply directly into Jup Lend with a single click. The refinance operation initiated an atomic transaction involving four steps:

1. Repay outstanding debt on Kamino.

2. Withdraw the associated collateral.

3. Transfer these assets to Jupiter Lend.

4. Recreate the position inside Jupiter Lend, maintaining the same loan amount and collateral ratio.

On Dec. 2, Kamino updated its smart contracts to block Jupiter’s program, preventing one-click refinancing. Both the Jupiter and Fluid teams (Jup Lend uses Fluid in the backend) framed the move as anti-competitive and against “open-finance principles.”

On Dec. 6, Kamino’s co-founder publicly explained the rationale for blocking Jup Lend’s migration tool, noting that Jupiter had repeatedly suggested that borrowers’ collateral is isolated, implying it is neither rehypothecated nor exposed to cross-contamination risk. However, this claim was not true, with even Fluid’s co-founder acknowledging rehypothecation within Jup Lend.

​Notably, Kamino never prevented users from repaying their loans manually and withdrawing their capital to Jup Lend. Whether against open-finance principles or not, the move to block the refinancing program was fundamentally a business decision, much like Jup Lend’s decision not to open-source its code (though it has plans to do so). In this regard, it’s interesting to analyze the competitive dynamics between both money markets over the past few months.

Since its launch in late August, Jup Lend has grown to $1.6 billion in deposits and $610 million in borrows. The chart below shows that Kamino’s deposits and borrows have decreased by $1.3 billion (-28%) and $460 million (-26%), respectively, during the same period. 

The top five assets by deposit growth since Jup Lend’s launch are USDC ($485M), JLP ($225M), SOL ($206M), syrupUSDC ($174M), and jupSOL ($85M). During the same period, Kamino has seen sizable outflows for all of these assets, except syrupUSDC. However, even for syrupUSDC, Jup Lend still attracted roughly 3x more inflows. 

Kamino’s growth over the past few months has come from assets not yet supported by Jup Lend. In particular, stablecoin inflows in Q4 have been driven by PYUSD ($42M) and Phantom’s CASH ($125M). Kamino has also been proactive in onboarding DATCO LSTs; most notably dfdvSOL and more recently fwdSOL.

Kamino’s PRIME integration stands out as a catalyst that can bring net new inflows into the money market. PRIME gives users exposure to a regulated credit pool backed by US real estate loans originated and serviced through Figure. This integration effectively gives access to a source of yield uncorrelated from crypto markets that may attract more institutional borrowers.

Wrapping up, Kamino and Jup Lend are obviously competitors, and competition is healthy as it drives innovation and ultimately benefits users. That said, as Solana Foundation’s Lily Liu noted, instead of fighting with each other, Kamino and Jupiter should focus on growing the pie and capturing market share from other chains and TradFi thereafter. Combined, both money markets still account for less than 10% of Aave’s deposits, and without initiatives like the PRIME integration, it will be impossible to close this gap. 

— Carlos

1. Institutional Flows Will Overpower the 4-Year Cycle

CIO of Bitwise Matt Hougan joins the Empire podcast to discuss the crypto market cycle. He argues the classic four-year bitcoin cycle is effectively over because its old drivers (like halvings and blowups) are weaker, while institutional adoption and better regulation are much stronger, so he expects 2026 to be an up year. A big theme is that financial advisors move slowly, care more about not getting fired than maximizing returns, and therefore add crypto gradually via simple, blue-chip assets (BTC, ETH, SOL, LINK, UNI, AAVE) and diversified baskets. He also highlights covered call strategies by long-time holders, dismisses fears that Strategy will be forced to sell, and explains why tokens will increasingly be valued on real revenues and fees as onchain activity and new narratives (stablecoins, tokenization, privacy, prediction markets) grow over time.

2. State of AI: An Empirical 100 Trillion Token Study with OpenRouter

OpenRouter and a16z published a report on the state of AI. The release of the o1 reasoning model in December 2024 marked a turning point for LLMs, shifting the field from pattern generation to multistep deliberation inference, accelerating deployment, experimentation and new classes of applications. Analyzing over 100 trillion tokens of real-world LLM interactions on the OpenRouter platform, this work reveals substantial adoption of open-weight models, the unexpected popularity of creative role-play and coding assistance, and the rise of agentic inference. Furthermore, their retention analysis identifies foundational cohorts: early users whose engagement persists far longer than later cohorts, a phenomenon the authors call the Cinderella "Glass Slipper" effect.  These findings emphasize the complex, multifaceted way developers and users engage with LLMs "in the wild," offering implications for model builders, AI developers and infrastructure providers.

3. Realities, Myths, and Misconceptions of BAM: A Steelman Exploration

Jito published an article to address the most common criticisms of BAM, and in doing so lays out the team’s vision for why they designed BAM in the manner that they did. The article argues that BAM is not a centralized sequencer, but an opt-in, verifiable scheduling layer intended to make Solana blockbuilding transparent, programmable and auditable. The article responds to concerns about centralization and orderflow monopolization by emphasizing plans for independent node operators, geographically distributed deployment, open-source code, and cryptographic attestations so anyone can verify ordering decisions, with validators free to disconnect and fall back to other clients. It also tackles performance and validator economics, with a long-term focus on growing sustainable yield by improving execution quality rather than “scheduling games.”Finally, it frames Plugins and Application-Controlled Execution (ACE) as the key unlock for “internet capital markets” (apps can enforce market-structure rules like liquidation or cancel priority), argues FIFO isn’t sufficient under congestion, and defends the TEE security model with attestations.

Decentralized AI coordination directly targets a structural mismatch in Crypto via applications designed to minimize trust and single points of failure. Within this landscape, Allora represents a differentiated attempt to move beyond static ensemble methods.

Competing approaches like Bittensor’s subnet marketplaces and Numerai’s stake-weighted ensembles highlight that there is no consensus architecture for decentralized intelligence. Instead, the sector is experimenting with how to measure contribution, how to allocate rewards, and how to trade off decentralization, performance, and coordination overhead.