Sept. 19, 2025 · New? Subscribe Here

Gm Lucas,

On the latest episode of the Galaxy Brains podcast, I interviewed Kyle Samani, co-founder of Multicoin Capital and now chairman of Forward Industries, a new Solana-focused digital asset treasury company. We talked about why Solana matters, how Forward stands out from the DAT pack, and why Kyle believes the future of DeFi will be built on Solana. 

From the Galaxy Research desk this week: Lucas Tcheyan published a comprehensive overview of decentralized AI training, explaining blockchains’ critical role in that process; and Jack Immanuel wrote a handy “cheat sheet” for readers who don't have time to digest the seminal but tome-length President’s Working Group on Digital Asset Markets report. 

In the newsletter this week, Zack Pokorny contextualizes the Ink rollup’s move to decentralize itself; Thad Pinakiewicz unpacks Tether’s new, U.S.-only stablecoin; and Christopher Rosa explains Google’s new standard for AI agent-initiated payments. 

Have a great weekend,

Alex 

Market Update

Data via Messari as of 7:20 AM 9/19/25

Stories of the Week

001

Kraken to Test 'Based' Rollup Architecture

Ink, the Kraken exchange’s OP Stack-based optimistic rollup (a type of layer-2 blockchain on Ethereum), announced plans to test a so-called based sequencing mechanism. The move comes in the wake of discourse and uncertainty around the regulatory viability of centrally sequenced rollups to handle tokenized securities and other real-world asset (RWA) activities.

In a based rollup, the sequencing of transactions is handled by a subset of the layer-1 (L1) chain's block proposers. This differs from single-sequenced rollups (e.g. Arbitrum, Base, and OP Mainnet) which have a single, central entity that orders transactions.

Ink’s announcement is significant because centrally sequenced rollups have been the go-to solution for Ethereum scalability, yet they have faced criticism over the years. Recently, Alex Thorn of Galaxy Research has criticized optimistic rollups for their “lack of alignment” with ETH holders, and he has argued that they must either decentralize or be regulated to be fit for tokenized securities. Despite years of roadmaps calling for it and some of these criticisms, few established L2s have taken steps to decentralize sequencers. There has been little incentive for them to do so because decentralization would mean relinquishing direct control over the rollup, putting user experience (UX) at risk, and would threaten the economic moat of the sequencing design.

Our Take

Ethereum rollups have long faced criticism for being "parasitic" (siphoning value capture away from the L1) and "lacking alignment" (operating mostly independently of the L1). Now, with regulatory conversations homing in on sequencer decentralization, rollups may face a new and powerful vector of decentralization pressure. Based sequencing, though still in its early stages of development and implementation, increasingly looks like the most viable long-term option for rollups that decentralize, whether voluntarily or under external pressure.

Rollups are presented with four options for decentralizing their sequencers that are reasonably achievable today:

• Shared sequencing networks: These rely on an external and independent validator set and consensus protocol to sequence transactions for multiple rollups. The pros of this approach include immediate decentralization and cross-rollup interoperability among participating rollups. Cons include the complexity and overhead of adding a new consensus protocol and the possibility of a fractured ecosystem, made up of multiple, non-interoperable shared sequencing networks, each with its own independent validator set. This moves the fragmentation problem from individual rollups to the network layer. Additionally, the rollup only becomes as decentralized as the sequencing network it relies on.
• Rotating/committee-centric sequencing: The rollup forms its own decentralized sequencer set, either through rotating leaders or a Byzantine Fault Tolerance (BFT) committee, typically secured with proof-of-stake and a staking token. Pros include decentralizing control of the rollup beyond a single operator; keeping economics internal to the rollup; and allowing a choice between a lightweight leader-rotation scheme that simply enforces existing sequencing rules or a full BFT consensus protocol where a committee reaches agreement on ordering. Cons include the possibility of weaker guarantees of liveness or censorship resistance if only a leader-rotation scheme is implemented; the additional complexity, latency, and governance overhead that come with adding another consensus layer if a full BFT committee is implemented; and in some cases, the need for a staking token (which can be difficult to spin up).
• Dual/federated sequencing: Multiple trusted entities (e.g. exchanges and infrastructure providers) run sequencers together; a threshold number of them (e.g., two of three signatures, or three of five) must agree for a block to be accepted. Pros include easier bootstrapping and fewer points of failure. Cons include a permissioned and cartel-like structure with possibly limited censorship resistance and weaker liveness guarantees.
• Based sequencing: In its truest form, a subset of L1 proposers is recruited to sequence rollup transactions. In practice, they delegate to gateways (preconfirmers), which handle transaction ordering and user experience (UX), while only L1 proposers finalize rollup activity on Ethereum. The pros include inheriting all qualities of the L1 (decentralization, censorship resistance, and liveness), providing native interoperability with Ethereum and other based rollups, aligning rollup economics more closely with the L1’s, and requiring no new consensus protocols. The cons include a UX bounded by L1 performance (unless preconfirmation infrastructure matures), and the fact that the implementation of true based sequencing is still in its infancy.

Based sequencing stands out as the better option because it directly adopts the decentralization of Ethereum’s validator set, avoiding the need to create or depend on a parallel network of sequencers. It provides native interoperability not only with the Ethereum L1 but also with other rollups that adopt the same model, whereas shared sequencing networks limit interoperability to rollups within their own system. By keeping Ethereum proposers as the sequencing layer, based rollups mitigate the “too many cooks” problem and sidestep the need for novel and unproven shared sequencing networks or the complexity of stacking additional consensus protocols. It also inherits Ethereum’s liveness guarantees and censorship resistance, ensuring that as long as Ethereum continues to produce blocks, the rollup can continue to operate, and transactions cannot be indefinitely suppressed. Crucially, based sequencing still allows rollup operators to define how fees and maximum extractable value (MEV) are distributed across gateways, validators, and the rollup treasury. This optionality gives them some degree of control over economic design while aligning incentives with Ethereum validators through blob fees and inclusion revenue. Combined with the ability for gateways to provide fast and robust preconfirmations for a smoother user experience, based sequencing can offer a cleaner, simpler, and more Ethereum-aligned path to decentralization than the alternatives. Based, indeed.

While based sequencing may provide a strong path forward, it also directly challenges the incumbent incentive model: single-sequencer rollups are money-making machines because they internalize all fees and MEV at high margins to a single entity. Any decentralization of the sequencer, whether through shared sequencing, rotating committees, federations, or going based, necessarily fragments that revenue stream and adds costs. This economic reality is a contributing factor to the hesitancy of rollups to decentralize, even in the face of mounting social and now (possibly) regulatory pressure. The tradeoffs in cost vectors between centrally sequenced rollups and based rollups are shown below. Note: it is still unclear what share of based rollups’ revenue they would have to pay to gateways and possibly L1 validators to properly incentivize operation (this is true of all moves to decentralize the sequencer). Presumably, it will vary from rollup to rollup.

That said, authentic based sequencing remains nascent and is being implemented in phases, with current deployments (e.g. Gattaca pilots through Ink testnet and Taiko) still in Phase 1 where preconfirmers are permissioned and much of the infrastructure is experimental. Among the primary blockers: building robust preconfirmation systems to offset L1 latency, because without them user experience is limited by Ethereum’s block time; ensuring consistent L1 proposer opt-in, so validators reliably include rollup batches; and designing sustainable economic splits across gateways, validators, and the rollup treasury. Other open challenges include creating monitoring and slashing mechanisms for misbehaving gateways; balancing decentralization with performance; and transitioning from permissioned to permissionless gateways. These hurdles highlight that while the endgame for based sequencing is clear, the path to a fully mature and decentralized implementation is still unblazed. – Zack Pokorny

002

Stars, Stripes, and Stablecoins: Tether Lands in America

Tether this week unveiled USAT, a U.S.-regulated, dollar-backed stablecoin to be formally issued by Anchorage Digital Bank, N.A., with Cantor Fitzgerald as reserve custodian, and former Executive Director of the President’s Working Group on Digital Assets Bo Hines as CEO of Tether USA. Rather than retrofitting its $170bn+ flagship USDT into a stablecoin that complies with the U.S. GENIUS Act, Tether is pursuing a dual-track approach: a purpose-built, regulated product for domestic markets alongside its offshore behemoth.

The clean-slate approach is telling. USDT’s $170+ billion reserve mix, which is ~80% (~$130bn) in low-risk Treasuries and cash equivalents and ~20% (~$40bn) allocated to bitcoin, gold, loans, and “other” investments, does not comply with the collateral requirements in the landmark U.S. stablecoin law enacted this summer. Rebalancing its holdings would not only be operationally complex and likely involve material slippage; it would also compress Tether’s net interest spreads and reduce the outsize profits that have fueled ~$20bn in dividends to its venture arm.

USAT is architected from Day One around short-term Treasuries, cash, and repurchase agreements (repos), precisely the model legislators outlined in the GENIUS Act. Naming Hines as CEO of the U.S. entity adds an overt political dimension to the launch. Hines' selection counterbalances Tether’s reputational risk with proximity to policymakers and the executive branch, a strategy successfully applied by the Trump family’s World Liberty Financial in the lead-up to its token launch.

Our Take

Tether is building a regulated vehicle for U.S. distribution while keeping its offshore model intact and profitable. The firm is arbitraging regulatory regimes: offering U.S. investors a safe, narrow box while continuing to compound returns globally. The real test will be whether USAT can chip away at Circle’s lead in U.S. trading and gain traction in that market. If USAT flounders, it may portend trouble for GENIUS-compliant stablecoins more broadly. Such stablecoins offer investors only two advantages: enforceable investment mandates on the stablecoin reserves and regulatory approval from the U.S. government. The former is only an advantage when comparing GENIUS-compliant stablecoins with other non-yielding stablecoins; one would prefer the stablecoin backing be stable and readily redeemable. Regulatory approval is only an advantage inasmuch as regulatory uncertainty hamstrung traditional institutional players from utilizing and accepting stablecoins. While institutional uptake of crypto this year has been meteoric, we haven’t seen a rush to mint compliant stablecoins like USDC. By contrast, the fastest growing stablecoin this year is Ethena’s USDE, thanks mainly to the yield it pays to holders, something the GENIUS Act explicitly seeks to prohibit.

Whether GENIUS Act-compliant stablecoins are the endgame is up for debate, but the value in Tether throwing its hat into the arena is not. This is a cheap venture for Tether; it already has the relationships with custodians and banks for traditional asset rails, it has the tech in-house for tokenization, and it has deep partner distribution channels in place. If nothing comes from the foray, Tether loses little; it will continue to compound returns offshore with perhaps the world’s most profitable group of employees. But if GENIUS compliant stablecoins become the norm and dominate the market, this could be the unlock to bring Tether from hundreds of billions of stablecoins in circulation to its first trillion. – Thad Pinakiewicz

003

Google Gives AI Agents the Power to Pay

This week Google announced the Agent Payment protocol (AP2), a payment standard that lets AI agents transact for users and among themselves. Agent communication is not new; Google already offered the Agent2Agent protocol (A2A), which enables agent interoperability. The breakthrough here is payments, with blockchains and stablecoins used to settle online transactions. This opens the door to agent-to-agent settlement and supports a long-term vision of microtasks powered by micropayments.

Earlier this year, Coinbase introduced the x402 protocol to let agents monetize services, pay other agents, and automate micropayments for users. Google has integrated x402 into AP2, signaling that agents are moving beyond simple prompt responses into real economic activity.

Our Take

Traditional payment rails were not built for the modern online economy. Settlement is slow, fees are high, and scalability is limited. Credit card transactions can take days to reach finality and often cost a few percentage points when you add network and processing fees. PayPal improves the user experience but still rides the same rails and charges a meaningful markup. Blockchains introduced global, programmable money with faster settlement and cash-like censorship resistance, but some of them come with tradeoffs such as limited throughput and variable fees. The most promising path today is stablecoins running on high-performance L1s and L2s, which can deliver fast, low-cost online payments.

Stablecoins can become a staple ingredient for onchain payments, but until Coinbase introduced the x402 protocol there was no recipe. To see why a protocol is necessary, consider how services are built online and onchain. Most applications and blockchains live in different ecosystems, use different programming languages, and run on different infrastructure. What they share is the internet, which relies on protocols such as HTTP and TCP IP. A protocol is a common language that tells each side how to format, send, and verify messages so they can interoperate. (If a protocol says dates should be formatted MM/DD/YY, for example, everyone understands that “04/05/25” means April 5, not May 4.) Think of it as a shared recipe that tells everyone what each ingredient is and how to combine them. Without it you get cilantro in one kitchen and coriander in the other. x402 is a payments protocol that standardizes the recipe for agents, so payments work the same way everywhere.

If stablecoins are the main ingredient and x402 and AP2 are the recipes, AI agents are the chefs. Agents are often mentioned but seldom explained. Many people equate AI with ChatGPT, which converts questions into text responses. Agents build on that foundation to plan workflows, call tools, and complete tasks on the internet. The earliest impact is in payments, especially small, automated transfers known as micropayments.

Micropayments show up in three patterns. First, there are real-time purchases with a human present. Without visiting any website, you tell an agent to find a refrigerator or new white running shoes; it assembles an itemized cart; you review and sign an approval that locks the exact items and price; and stablecoins settle onchain. This end-to-end flow has been demonstrated in Coinbase’s refrigerator purchase demo.

Second, micropayments can power delegated tasks with no human present. With this workflow, you can have your agent book a vacation for you from start to finish using Google’s A2A communication to coordinate with airline, hotel, and tour guide agents. They can negotiate inventory and price and use x402 to stream small payments to each party from the first deposit to booking completion. Sounds like sci-fi, but this an imminent scenario. The messaging and payment pieces work today, and wider adoption by travel providers is now the gating factor.

Third, micropayments power network rewards. x402 can pay users in small stablecoin streams for sharing GPUs to support decentralized AI training. As shown in Galaxy's recent report on decentralized training, such reward schemes are coming online and x402 aims to make those payments interoperable across agents.

In all three micropayment scenarios, the agent can send and receive payment for completing the task and the company receives payment for the service, with routing and splits handled automatically over A2A and x402.

As these capabilities converge, the rails beneath them run on blockchains. Every instruction and payment can be verified onchain, which raises the bar on auditability and trust. Concretely, x402 is chain-agnostic. Whether it settles on high-throughput L2s or alternative L1s, the protocol can cost as little as $0.0001 per transaction, confirm transactions in about two seconds, and theoretically can process hundreds to thousands of them per second.

Just as important, the architecture is composable not just across blockchains but also the open web, allowing agents, merchants, and services to plug in without rebuilding their technology stacks. Google has brought in major partners from Web2 and Web3, while the Ethereum Foundation is working to advance agent identity and trustworthiness. At the same time, teams across competing L1s and L2s are racing to become the agent settlement layer. In combination, AP2 and x402 deliver payments, and the EF’s work will give agents a verifiable presence, so software can both move value and have a trusted identity. The direction is clear even if the winners are not.

What stands out most is the opportunity ahead as agents, payments, identity, and open protocols converge. The internet had read and write, blockchains added own, now it gets pay. – Christopher Rosa

From Our Desk

Decentralized AI Training: Architectures, Opportunities, and Challenges

Decentralized training has gone from theory to reality. Projects including Nous Research, Prime Intellect, Pluralis, Templar, and Gensyn are running real training at global scale.

Cheat Sheet: The President’s Working Group on Digital Asset Markets Report 

For readers who were on summer vacation when this landmark document dropped (or just could never read the whole thing), Galaxy Research presents a summary of the key points.

A Near-Miss: How the NPM Breach Almost Wreaked Havoc for Crypto 

The digital assets industry dodged a bullet. Here's how developers and users can protect their assets from such exploits.

The State of Onchain Yield: From Stablecoins to DeFi and Beyond

Read every headline yield number through two lenses: what you earn for the risk, and what remains after the work involved.

Introducing Tokenized GLXY

In the long term, we believe that onchain equity capital markets will become as featureful if not more so than traditional markets.

GALAXY BRAINS PODCAST · EPISODE 189

Solana’s Big Treasury Company with Kyle Samani

Alex Thorn talks with Kyle Samani, co-founder of Multicoin Capital and now Chairman of Forward Industries, a new Solana-focused digital asset treasury company. Kyle explains why Solana matters, how a Solana-focused DAT is differentiated, and why the future of DeFi will be built on Solana. Alex also talks with Beimnet Abebe (Galaxy Trading) about the Federal Reserve’s latest rate cut and what it means for markets.

Spotify  |  Apple  |  Google  |  YouTube

Charts of the Week

Token buybacks by Solana applications have occupied one third of all token buybacks by chain year-to-date (YTD) in 2025. This, in part, underscores the Solana apps’ abilities to generate a relatively elevated amount of fees. Over the past year, Solana applications have increasingly dominated application fees, leading all chains for 11 consecutive months and generating over $4 billion in fees year-to-date. Now applications are increasingly returning those fees to token holders in the form of buybacks.

Application fees reflect what users are willing to pay for services rather than just access to blockspace. Strong, recurring app revenues not only validate businesses built onchain but also benefit the base layer as sustained application demand leads to persistent blockspace usage, which drives network fees, validator income, and, in Solana’s case, token burns. Solana’s top applications primarily consist of memecoin launchpads, trading bots, and decentralized exchanges (DEXs), a stark contrast to Ethereum’s DeFi lend/borrow applications.

Hyperliquid has also established itself as a prominent fee generating application. So far this year, Hyperliquid has bought back $517.3 million worth of HYPE (representing 60.2% of all tokens bought back) as of September 14, 2025 YTD.

Notably, a total of $859.1 million worth of token buybacks have been executed as of September 14, 2025 YTD across all chains. In our predictions for the year we estimated a nominal amount of $1 billion would be used by applications to buyback their tokens throughout 2025. With 16 weeks left in the year, total token buybacks are poised to clear $1.2 billion using the YTD weekly average amount of buybacks.

In Other News

• 🏃SEC Approves New Standards Fast-Tracking Crypto ETF listings
• 🪙Coinbase adds USDC lending with Morpho on Base
• 🔍NYDFS Issues Guidance to Banks on Blockchain Analytics
• ⛓️‍💥Binance Negotiates to End DOJ Compliance Monitor
• 🗽Bullish Exchange Secures a New York BitLicense
• ⏩Forward Industries Launches $4b ATM Offering for Solana
• 📁Bitwise Files for Avalanche ETF Amid New Fund Launches
• 💰Fellowship PAC Launches With $100m for Crypto Policy...
• ♊ ... While Winklevoss Twins Start PAC With Trump-Linked WLFI

Thank you for reading.

Please feel free to contact us with any questions or comments.

This newsletter provides links to other websites that we think might be of interest to you. Please note that when you click on one of these links, you may be moving to a provider's website that is not associated with Galaxy. These linked sites and their providers are not controlled by us, and we are not responsible for the contents or the proper operation of any linked site. The inclusion of any link does not imply our endorsement or our adoption of the statements therein. We encourage you to read the terms of use and privacy statements of these linked sites as their policies may differ from ours. This document, and the information contained herein, has been provided to you by Galaxy Digital Holdings LP and its affiliates (“Galaxy”) solely for informational purposes. This document may not be reproduced or redistributed in whole or in part, in any format, without the express written approval of Galaxy. Neither the information, nor any opinion contained in this document, constitutes an offer to buy or sell, or a solicitation of an offer to buy or sell, any advisory services, securities, futures, options or other financial instruments or to participate in any advisory services or trading strategy. Nothing contained in this document constitutes investment, legal or tax advice. You should make your own investigations and evaluations of the information herein. Any decisions based on information contained in this document are the sole responsibility of the reader. Certain statements in this document reflect Galaxy’s views, estimates, opinions or predictions (which may be based on proprietary models and assumptions, including, in particular, Galaxy’s views on the current and future market for certain digital assets), and there is no guarantee that these views, estimates, opinions or predictions are currently accurate or that they will be ultimately realized. To the extent these assumptions or models are not correct or circumstances change, the actual performance may vary substantially from, and be less than, the estimates included herein. None of Galaxy nor any of its affiliates, shareholders, partners, members, directors, officers, management, employees or representatives makes any representation or warranty, express or implied, as to the accuracy or completeness of any of the information or any other information (whether communicated in written or oral form) transmitted or made available to you. Each of the aforementioned parties expressly disclaims any and all liability relating to or resulting from the use of this information. Certain information contained herein (including financial information) has been obtained from published and non-published sources. Such information has not been independently verified by Galaxy and Galaxy does not assume responsibility for the accuracy of such information. Affiliates of Galaxy's own investments in some of the digital assets and protocols discussed in this document. Except where otherwise indicated, the information in this document is based on matters as they exist as of the date of preparation and not as of any future date, and will not be updated or otherwise revised to reflect information that subsequently becomes available, or circumstances existing or changes occurring after the date hereof. The foregoing does not constitute a "research report" as defined by FINRA Rule 2241 or a "debt research report" as defined by FINRA Rule 2242 and was not prepared by GalaxyDigital Partners LLC. For all inquiries, please email contact@galaxy.com. ©Copyright Galaxy Digital Holdings LP 2025. All rights reserved.