Announcing our investment in Allium: the system of record for onchain finance

The most boring question in finance is also the most important one: what is the balance?

A balance feels boring because traditional finance spent a century making it boring – by using standards. Lots of them. Bloomberg standardized prices, DTCC standardized settlement, and SWIFT standardized payments. Each one exists so that everyone reads the same number and can say where it came from. You trust your balance because a hidden, complex institutional machinery answers for it.

This was the promise of Blockchains. They provide the same trust, but without the institutions. The ledger is public, so anyone can verify it. If Ethereum says an address holds 3 ETH, it holds 3 ETH. But a public ledger is not a system of record. A system of record is the version a business treats as the truth: every number means the same thing to everyone who reads it, and any figure can be traced back to where it came from. A blockchain does none of that. It records what happened, not what it means.

To understand why, consider a single token transfer, like a customer payment, market-maker rebalance, bridge hop, or internal treasury move. On the chain, all four look identical: tokens moving from one address to another. Spread that across 150-plus blockchains and thousands of protocols, and two systems looking at the same activity will report completely different numbers because each one interprets it for itself.

For years teams have worked around this ambiguity or tried to avoid it altogether. But expectations are rising with institutional adoption. Stablecoins settled hundreds of billions of dollars last year, and tokenized assets are growing fast. As those numbers start backing real payments and financial reports, a sketchy figure that was fine in a research dashboard now has to hold up in a reconciliation or an audit. 

This is what Allium does for their 150+ enterprise customers. They turn raw onchain activity into a system of record.

Allium ingests, decodes, and enriches data from over 150 blockchains. Decoding turns each chain's raw events into one common schema. Enrichment is the hard part: they've decoded over 10,000 schemas, which lets them label a transaction — a staking reward, a fee, or a Binance deposit — and tie it to the entity behind it. That is how you ask not just what happened onchain, but who did it. Customers get the data in their own warehouse, over an API, or as a live stream, and every number traces back to the underlying transactions. Further, the pipelines are SOC 1 and SOC 2 certified. 

Allium's hundreds of customers represent a range of use cases. Visa's public stablecoin dashboard runs on Allium. Grayscale draws on it for the digital-asset research behind its ETF products. At MoonPay, finance teams reconcile activity across chains by joining Allium's data with their own books. Phantom uses its data to serve live wallet balances to more than fifteen million users. These are very different problems, all built on the same record.

For now, Allium's users are businesses and governments. But the bigger opportunity is still ahead, and its users won't be human. As agents start transacting on their own, machine payments will dwarf human ones, and they will settle onchain. In fact, agents already outnumber people on the web: in June 2026, Cloudflare reported that bot traffic — AI agents reading product pages, comparing prices, and shopping on our behalf — passed human traffic for the first time, ahead of its own forecast. Granted, they don't pay at scale yet. But traffic precedes transactions, and when agents do pay, the economics point onchain. 

An agent could be given a credit card, but it won't use one: cards and bank accounts are built for a human who passes a KYC check, approves the charge, and waits days for it to settle. An agent transacts constantly and at machine speed, so it needs rails that are programmatic, instant, and global, which stablecoins on public chains already provide. The emerging agent-payment standards, like Coinbase's x402, settle in stablecoins for that reason.

Before an agent accepts a payment, it has to trust the data behind it: who actually controls the wallet on the other end, and where the money has been. The raw chain won't tell it either one. Allium fills that in. It ties anonymous addresses to real entities and traces funds across chains, so it can look at an incoming transfer and tell that the money came from Coinbase, moved through Uniswap and a bridge, and arrived clean. 

That capability is ahead of the market it serves. Agentic commerce is young, and the standards for how agents identify each other and move money are still being written, but the rails are already going up: Mastercard launched Agent Pay and Visa extended Stripe and Tempo’s Machine Payments Protocol to card rails. And earlier this year, Allium shipped a product agents use without a human in the loop: they sign up and pay per call in stablecoins through x402.

Allium is a machine learning, distributed systems, crypto protocol, and enterprise trust problem rolled into one. The pool of people with real depth in all four is vanishingly small. Ethan Chan and Cheng Han Lee are two of them. At Primer, Ethan led the machine learning and data infrastructure for real-time intelligence systems serving Fortune 500 companies and governments, all whileand lecturinged on applied ML at Stanford. Cheng Han built payments systems that moved billions at Poynt, (acquired by GoDaddy), and ad-data infrastructure at Meta. They started Allium together in late 2021 and raised through the Terra and Luna collapse, when many in crypto headed for the exits. They've assembled a rare team: by their count, fifteen of the fifty or so people who have built production blockchain indexers at scale.

Reading the chain is mostly solved. The next phase of onchain finance, for organizations and eventually agents, turns on agreement: one boring, trusted answer to "what is the balance?"

That's why we're investing in Allium's Series B. We like boring things. Especially boring things growing this fast. We're thrilled to partner with Ethan, Cheng Han, and the Allium team.