Hypercubic announced on August 18, 2026 that it has raised $5.3 million in seed funding led by CIV, joining a growing cohort of startups betting they can automate one of the tech industry's oldest headaches: getting companies off ancient mainframe systems. The San Francisco company, which participated in Y Combinator's Fall 2025 batch, uses what it calls agentic AI to migrate COBOL-based mainframe systems to modern infrastructure.
The round drew participation from Y Combinator, Afore Capital, Multimodal Ventures, Pioneer Fund, Epsilon Ventures and Unpopular Ventures, plus a roster of individual investors that reads like a who's who of tech operators. Kaz Nejatian, CEO of Opendoor, joined. So did Gokul Rajaram, a board member at Coinbase and Pinterest, along with Venky Harinarayan, who co-founded Walmart Labs.
It's a market with staying power, if not glamour. Mainframes still process 87% of all transactions worldwide, according to IBM. A July 2025 report from the Government Accountability Office found that federal agencies continue running core Treasury functions on COBOL and Assembly code, languages that predate the internet by decades. IBM Research pegged the volume of COBOL still in production at more than 200 billion lines of code as of January 2024.
Hypercubic's pitch centers on turning what has traditionally been a grueling, bespoke consulting engagement into something more like packaged software. The company built four products that operate entirely within a customer's virtual private cloud or on-premises environment. HyperDocs converts COBOL codebases into searchable documentation. HyperTwin captures workflows from subject-matter experts who actually know how the old systems work.
Then there's Hopper, an agentic development environment for mainframes that the company launched in May, and HyperLoop, a verification-driven modernization system that also debuted in May. Hopper works inside TN3270 terminal sessions, writing Job Control Language scripts, parsing outputs and querying VSAM databases. HyperLoop employs three verification loops, including what the company describes as a verified simulation twin built with Dafny, to ensure the new system behaves identically to the legacy one. Migrations can be reversed at the package level, Hypercubic says.
Co-founder Sai Gurrapu spent time as a machine learning engineer in Apple's AIML organization before launching Hypercubic with a co-founder who also worked at Apple as an engineer. Gurrapu has published research at IEEE and AAAI conferences, according to the company's Y Combinator profile, though the startup has kept other details about its founding team relatively close.
"Our thesis is simple: mainframe modernization should become a repeatable, software-driven process rather than a bespoke human undertaking," Gurrapu wrote in the company's funding announcement. Whether that vision can scale remains an open question. Hypercubic says it's already compressing parts of the modernization lifecycle with customers but declined to name clients or disclose revenue.
The competitive landscape is crowded and varied. IBM offers watsonx Code Assistant for Z. AWS runs a Mainframe Modernization service. Consulting giants like Accenture, DXC and Infosys have built entire practices around this work. On the startup side, CloudFrame raised $7 million in March 2022. Heirloom Computing, Raincode and TSRI are also in the mix, each with their own approach to cracking what has proven a stubborn problem.
Hypercubic targets a broad swath of industries where mainframes remain entrenched: financial services, insurance, government, manufacturing, defense, airlines, telecommunications, healthcare, oil and gas, logistics, retail and utilities. The angel investor list for this round included Henri Stern, co-founder of Privy; Tony Dang, co-founder of Infisical; and Garry Tan, president of Y Combinator. Others who participated were Mike Hinckley, Ryan Bednar, Ben Bryant, Youcef Es-skouri, Gaurav Jain, Brian McCullough, Chris Messina and Jeff Rosenthal.
The company earned a spot in TechCrunch's Startup Battlefield 200 at Disrupt 2025 last December, a nod to traction that extends beyond just the funding itself. Still, migrating enterprises off systems that have run without interruption for decades is less a sprint than a marathon, and one where technical elegance matters far less than trust. Hypercubic will need both.
