Hypercubic raises $5.3M to map out and rewrite legacy COBOL apps with AI agents
SiliconANGLE Mike Wheatley
Hypercubic just raised $5.3M to use AI agents that read and rewrite ancient COBOL banking code. Most COBOL programmers are about to retire, and nobody younger knows the language.
Based on reporting by SiliconANGLE, Mike Wheatley — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Somewhere behind the ATM you used last week, there's a decent chance a chunk of COBOL is quietly doing the math. That language dates back to 1959, and by some estimates over 200 billion lines of it still underpin more than 95% of the world's card swipes and cash withdrawals. It works fine. The problem is who's left to maintain it.
That workforce is vanishing. One estimate cited in the space suggests 90% of current COBOL engineers will retire within five to ten years, leaving banks and industrial firms sitting on mission-critical systems nobody can read anymore. Rewriting that code into something like Java sounds like the obvious fix, except COBOL has earned a reputation as the software equivalent of asbestos — deeply woven into mainframe stacks where one undocumented change can crash a bank's ledger or a government payments system, and take days to untangle.
Hypercubic, founded by former Apple engineers Sai Gurrapu and Aayush Narik, is betting AI agents can do the untangling. The company just closed a $5.3 million seed round led by CIV, with Y Combinator, Afore Capital, Pioneer Fund, Multimodal Investors and angels including Opendoor CEO Kaz Nejatian and former Amazon Marketplace engineer Venky Harinarayan joining in. Its agents are trained specifically on COBOL, and instead of jumping straight to translation, they map an entire codebase first, pulling out the buried business logic and generating documentation that, in many cases, never existed for these systems in the first place.
Gurrapu frames the real problem as recovering decades of hidden logic and proving a rewritten system actually behaves the same way the old one did in production — not just converting syntax. Once that documentation exists, generating a modern replacement that preserves the original app's data and behavior gets a lot more tractable. Gurrapu claims the agents can modernize legacy applications in a couple of months rather than the years a manual rewrite typically demands.
Hypercubic already has live work underway with one of Latin America's largest banks and a Caribbean retailer serving more than 50 million customers a year — real stakes, not pilot-project theater. The new funding is earmarked for building out the agents further and taking on more enterprise clients, which suggests the company thinks the market of COBOL-dependent institutions quietly dreading this problem is bigger than five million dollars' worth of seed capital.
My take — AI-written commentary, not fact-checked reporting
Every institution running COBOL has known this reckoning was coming for decades and mostly chose to look away, because rewriting a bank's ledger system is terrifying and expensive. Throwing AI agents at the problem doesn't remove the risk of breaking something that processes real money, it just changes who's making the judgment calls about what
Read more about this at: SiliconANGLE
