Digital Asset's $300M raise at $2B valuation—led by a16z crypto, backed by Visa and Goldman Sachs—marks the first major capital deployment into permissioned blockchain infrastructure since the crypto winter thawed.
Digital Asset closed $300 million at a $2 billion valuation with a16z crypto leading and Visa and Goldman Sachs participating. The round funds Canton Network—a permissioned interoperability protocol designed to let banks, asset managers, and payment processors synchronize positions across fragmented ledgers without reconciliation lag. Visa is already running its Visa Tokenized Asset Platform on Canton; Goldman has been testing repo settlement on it since 2023. This is not speculative infrastructure. It's production capital flowing into permissioned rails that incumbents believe will replace correspondent banking plumbing and cut multi-day settlement windows to minutes. What matters: the timing and the composition. a16z crypto is writing the lead check into a *permissioned* blockchain stack at the same moment Stripe is scaling stablecoin orchestration and JPMorgan Chase is moving JPM Coin to public rails via Kinexys. The battleground is interoperability—how do you move tokenized dollars, securities, and FX positions between institutions that don't trust each other's ledgers? Canton's answer is privacy-preserving atomic settlement with cryptographic proofs instead of nostro accounts. Visa's continued backing suggests it views this as the path to disintermediate correspondent layers it doesn't control—and capture margin that today leaks to SWIFT, clearinghouses, and custodians. The network effects are inverted: instead of racing to onboard consumers, Canton is racing to onboard the institutions that *are* the network. We're watching whether Mastercard follows with a similar commitment or doubles down on its own tokenization experiments. The other tell: does JPMorgan Chase integrate Kinexys with Canton, or does it remain a walled garden? If permissioned and permissionless rails start to interoperate—Canton handling institutional settlement, stablecoins handling consumer-to-merchant flows—then the category converges and the race becomes who controls the orchestration layer. That's where Stripe's Bridge acquisition and Visa's Canton bet collide.
Including Our Take, the Tailwinds & headwinds framing, Connections across the FOBI roster, and What should you do.
Already subscribed? Sign in →
Anthropic eliminated rate limits and doubled coding capacity by securing access to SpaceX's Colossus GPU cluster—but the move has exposed a widening adoption risk: a coordinated developer migration toward open-source alternatives that bypass API costs entirely.
Anthropic eliminated rate limits on Claude Code and doubled capacity for paid tiers by negotiating access to SpaceX's Colossus 1 supercomputer—a rare infrastructure win that signals confidence in agentic-coding demand. But the timing exposes a critical vulnerability: the same week, OpenCode crossed 157,000 developer sign-ups, establishing an open-source competitor that trades frontier model quality for cost elimination and data residency. This is not a threat to Anthropic's model quality; it's a distribution moat erosion. What's actually happening is a bifurcation in the devtools market. For enterprises and professional developers willing to pay for frontier coding agents, Anthropic's move (combined with integration into Microsoft 365 via Word, Excel, and Outlook) is embedding Claude deeper into the IDE and productivity workflow—the most defensible layer in devtools. But for cost-conscious teams, open-source shops, and developers building on-premise infrastructure, OpenCode's ability to run locally and avoid per-request API billing erodes the API-access moat that Anthropic, OpenAI, and rely on. The Llama and Mistral open-weight models created the supply side; OpenCode is proving the demand exists. We're watching whether this 157,000-developer cohort stabilizes or accelerates. If the attrition is real (and the numbers suggest it is), will likely respond by bundling Claude Code deeper into enterprise workflows and pricing it as infrastructure (the Microsoft integration play) rather than a discrete agent API. The true competitive signal will be if OpenCode matures enough to close the capability gap with Claude on the hardest coding tasks (multi-file refactoring, test generation, architectural decisions)—that's the moment the price/quality frontier truly shifts.
Including Our Take, the Tailwinds & headwinds framing, Connections across the FOBI roster, and What should you do.
Already subscribed? Sign in →
AI coding adoption is now bifurcating by pricing model and data-residency constraint, not just capability. For years, the narrative was "frontier models win because they're best." That's still true for performance-critical use cases, but it misses the larger developer economics: if open-source models are 80% as capable and cost zero per request, the API economics collapse for commodity tasks. Anthropic, OpenAI, and GitHub must now compete not on pure capability but on embedding (IDE/workflow integration), curation (curated model fine-tuning), and enterprise bundling. The rate-limit removal and GPU partnership buys time on the first dimension but doesn't solve the third.
The asymmetric bet here is on open-source moat formation in agentic coding. If 157,000 developers defecting to OpenCode is a leading indicator of a larger cost-arbitrage driven shift, then the real positioning question is not "will Anthropic win coding?", but "who captures enterprise distribution once open-source models get good enough?" GitHub's Copilot integration advantage and JetBrains' IDE dominance matter more than raw model access in this scenario. Anthropic's Microsoft partnership is a hedge against this outcome, but watch whether the open-source cohort can build distribution faster than Anthropic can bundle.
Strategic-positioning commentary · not investment advice
Anthropic's current devtools revenue model is per-request API pricing for Claude Code, supplemented by enterprise contracts for dedicated inference. The SpaceX deal solves compute cost, but not the pricing leverage problem: as open-source models improve, developers optimize for total cost of ownership (TCO), not model quality alone. Anthropic's path forward is enterprise bundling (e.g., via Microsoft 365, security scanning, governance features) where they charge for curation and compliance, not raw inference. This is a lower-margin, higher-friction path than API access, but it's where the moat actually exists.
