Key Takeaways:
- SEC Moves to Modernize the Rules and Infrastructure for Digital Markets
- Coinbase Brings Tokenized Equities Further Onchain
- Stripe Bets on Infrastructure for the AI Economy
SEC Moves to Modernize the Rules and Infrastructure for Digital Markets
The U.S. Securities and Exchange Commission took several steps in August toward modernizing the regulatory framework and market infrastructure for digital assets. On August 18, the SEC proposed Regulation Crypto Assets ("Reg Crypto"), its first securities framework designed specifically around crypto asset issuance rather than adapting rules written for traditional securities. The proposal would create a pathway for certain token issuers to raise up to $5 million under a startup exemption or up to $20 million or $75 million under a larger Regulation A-style framework, while introducing purpose-built disclosures around token supply, governance, smart contracts and the development of the underlying network. Importantly, it would also establish a mechanism through which the investment contract associated with a token could formally cease to exist once the issuer's promised development efforts are completed. Alongside this, the SEC is updating transfer-agent rules that have remained largely unchanged since the 1970s and is preparing for a September roundtable on 24-hour trading, as other pieces of market infrastructure move toward overnight operation.
Taken together, these initiatives point to a broader modernization of U.S. capital-market infrastructure rather than simply a more permissive approach to crypto. The ability for investment contracts to have a defined lifecycle could provide a framework for resolving the legal status of thousands of existing tokens, while purpose-built disclosures recognize that token investors require different information from traditional equity investors. At the infrastructure level, extended market-data hours, 24x5 clearing and the development of overnight trading demonstrate that continuous markets require ownership records, settlement and compliance systems capable of operating continuously as well. The result could be a more coherent foundation for markets that increasingly operate on programmable and always-on rails, although the durability of the framework will ultimately depend on final SEC rules and congressional legislation.
Coinbase Brings Tokenized Equities Further Onchain
Coinbase entered the growing tokenized-equity market in August by launching tokenized stocks on Base. The initial rollout included 13 U.S. equities, with NVIDIA, Meta, Apple and Alphabet among the first stocks to have circulating supply. The tokens are backed by underlying shares held in regulated, bankruptcy-remote custody and can be traded and used across the Base DeFi ecosystem. Coinbase has since positioned the product as part of a broader tokenization infrastructure designed to bring traditional assets onchain and enable 24/7 trading, lending and borrowing.
The more important question is how these tokenized equities are structured and what investors actually own. Coinbase's model is a third-party-issued structure in which tokenholders have an economic interest and a claim on the underlying shares, while legal title and shareholder rights remain within the intermediary structure. This contrasts with issuer-sponsored models, such as Galaxy's tokenized GLXY shares, where the issuer itself recognizes the token as common stock and maintains the shareholder relationship through its transfer agent. Third-party wrappers are easier to scale because they do not require each public company to participate, but they introduce questions around shareholder rights, governance and regulatory treatment. As the SEC considers its anticipated "innovation exemption" for onchain securities trading, the distinction between these models will likely become increasingly important in determining how tokenized equities develop in the U.S.
Stripe Bets on Infrastructure for the AI Economy
Stripe agreed to acquire OpenRouter, an AI model gateway that routes requests across more than 400 models from over 80 providers, in what would be its largest acquisition to date. OpenRouter reportedly processes more than 10 trillion AI tokens per day for over 10 million developers and companies, with token volume growing rapidly as enterprises increasingly use multiple models rather than relying on a single provider. The acquisition reflects Stripe's view that businesses will need an "intelligence pipeline" alongside their existing financial infrastructure: a system that can select models based on cost, performance, latency and reliability. The announcement came as Ramp launched a competing model-routing service, highlighting the emergence of model orchestration as a distinct infrastructure layer within the AI stack.
The strategic significance extends beyond AI model selection. As inference becomes a material operating cost, companies will increasingly need infrastructure that can allocate intelligence as efficiently as they already allocate financial capital. Stripe's broader push into stablecoins, programmable custody and purpose-built payment infrastructure also points toward a future in which autonomous AI agents can discover services, consume intelligence and settle payments with limited human intervention. This creates a natural intersection between AI-native software and crypto-native financial rails: model routing optimizes the flow of intelligence, while stablecoins and programmable assets can facilitate the flow of value. The competition is therefore moving upstream from individual models toward the infrastructure that coordinates, prices and monetizes AI activity.
Nvidia Expands Its Infrastructure Footprint with Hugging Face
Nvidia agreed to acquire Hugging Face for $12.93 billion, bringing one of the largest open AI model and developer platforms into its broader AI infrastructure ecosystem. Hugging Face is often described as the "GitHub for AI," hosting models, datasets and tools used across the open-weight ecosystem. Nvidia has previously invested in the company and has become one of its largest contributors through model families such as Nemotron. The acquisition follows Nvidia's broader expansion beyond GPUs into AI infrastructure, models and applications, positioning the company across multiple layers of its AI stack. Nvidia and Hugging Face have stated that the platform will remain open and compute-agnostic, including continued support for models and hardware from other providers.
The acquisition is both defensive and offensive. As OpenAI, Anthropic, Google, Meta and Microsoft develop custom accelerators, Nvidia's long-term position increasingly depends on maintaining relevance across the broader AI ecosystem rather than relying solely on GPU demand. Hugging Face provides distribution and developer access at the model and application layers, allowing Nvidia to optimize popular open models for its infrastructure while remaining embedded in the open-weight ecosystem. At the same time, the deal highlights an emerging tension around neutrality: even if Hugging Face remains formally open, Nvidia will have greater influence over model discovery, evaluations, runtimes and enterprise defaults. As open-weight models increasingly commoditize inference and reduce the cost of AI deployment, control of the surrounding infrastructure may become as strategically important as control of the underlying models or chips.
Disclaimer: This post has been prepared for general informational purposes only and reflects the current views of its authors. The views expressed do not necessarily represent those of Eterna Capital, its affiliates, or individuals associated with Eterna Capital, and may change without notice. Nothing contained herein constitutes or should be construed as investment, legal, accounting, or tax advice, or as a recommendation, offer, or solicitation to buy or sell any investment. This material should not be relied upon to evaluate the merits of any investment decision. Eterna Capital makes no representation or warranty, express or implied, regarding the accuracy, completeness, or reliability of the information contained herein. All liability in connection with this material and any reliance thereon is expressly disclaimed.
