Infrastructure Over Everything: The Next Phase of Crypto and AI Capital

This week overview (March 23 - March 27, 2026)

Institutional capital is concentrating around infrastructure that directly runs financial systems and enterprise workflows. BitGo and ZKsync are pushing tokenized deposits into banking rails. Startale is building a vertically integrated stack for onchain capital markets in Asia. Reflection is attracting billions to scale open AI models as strategic infrastructure competing with closed systems. Harvey is turning legal work into agent-driven execution used by global enterprises. Across crypto and AI, capital is moving away from experimentation into systems that control settlement, computation and workflow execution.

BitGo x ZKsync: Turning Bank Deposits Into Onchain Instruments

BitGo and ZKsync are pushing deeper into institutional finance with a new partnership focused on tokenized deposits. The two firms are integrating custody and wallet infrastructure with Prividium, a privacy-preserving blockchain system designed for regulated banks, enabling them to issue and settle onchain representations of traditional deposits without moving funds outside the banking system. The stack is built around compliant, always-on settlement and programmable treasury operations, effectively turning bank liabilities into blockchain-native instruments. The platform is already being tested with multiple regulated institutions, with broader deployment targeted for the end of 2026.

Startale: Building Asia’s Onchain Capital Markets Stack

Startale Group closed a $63 million Series A, anchored by a $50 million second close from SBI Group and a prior $13 million tranche from Sony Innovation Fund. The raise is aimed at accelerating a vertically integrated onchain stack spanning infrastructure, stablecoins, tokenized securities and consumer applications. At the core is Strium, a Layer 1 built for institutional RWA and securities trading, alongside JPYSC and USDSC stablecoins designed for fiat integration and onchain yield distribution. Startale is also pushing its SuperApp strategy, bundling asset management, payments, and mini-app ecosystems into a single interface that abstracts blockchain complexity. The SBI partnership, which already includes joint development of Strium and a trust bank-backed JPY stablecoin, signals a coordinated move to build the settlement and exchange rails for Asia’s onchain capital markets, positioning Startale across both institutional and consumer layers as adoption scales.

Reflection: Open Models as Strategic Infrastructure

Reflection AI is reportedly raising $2.5 billion at a $25 billion valuation, with potential backing from JPMorgan Chase, positioning it as one of the largest emerging players in open AI models. Backed heavily by Nvidia, which previously invested around $800 million, Reflection is part of a broader push to build an open-source AI ecosystem in the U.S. to counter Chinese players like DeepSeek. The core thesis is accessibility: open models that developers can freely use and modify, competing directly with the closed systems from OpenAI and Google. Through initiatives like Nvidia’s Nemotron Coalition, capital is now flowing into open-model infrastructure as a strategic response to China’s push for exportable, lower-cost AI systems. 

Harvey: Legal Work Becomes Agent Infrastructure

Harvey AI raised $200 million at an $11 billion valuation, co-led by GIC and Sequoia Capital, with continued backing from Andreessen Horowitz and Kleiner Perkins. The focus is scaling AI agents as core legal infrastructure: more than 25,000 agents already run on Harvey, automating workflows across M&A, due diligence, contract drafting and compliance. 

“AI isn’t just assisting lawyers. It’s becoming the system through which legal work gets done,” said Winston Weinberg, CEO and co-founder of Harvey. "The law firms and in-house teams leading the way are building agents that execute complex workflows so lawyers can focus on judgment, strategy, and outcomes.”

Legal work is moving from manual execution to fully agent-driven systems, with long-horizon agents handling complex, multi-step processes over time. Harvey now works with the majority of the AmLaw 100, over 500 in-house teams and operates across 60 countries, with more than 100,000 lawyers using the platform. The capital is going toward expanding these agent systems and embedding legal engineering deeper into customer workflows, positioning Harvey as the operating layer where legal work is executed rather than assisted.

The common thread isn’t innovation at the edges, it’s ownership of execution. The layers that move money, run models and process work are where capital is consolidating, and where long-term leverage is being built.