IBM v. LzLabs — Judge Finds Mainframe Emulation Software Infringes Five IBM Patents

Background

IBM’s mainframe platform — centered on its z/OS operating system and z-series hardware — generates billions in annual revenue and powers the transaction processing systems of most of the world’s largest banks, insurance companies, and government agencies. That dominance rests not only on IBM’s hardware but on the proprietary instruction-set architecture (ISA) and software interfaces that customers depend on. To run mainframe workloads, customers traditionally must license IBM software, paying significant fees as long as they remain on the platform.

LzLabs, a Swiss company, developed technology it called the “Software Defined Mainframe” (SDM) — a system that allowed customers to lift and run IBM mainframe applications on commodity x86 hardware without IBM licenses, by translating mainframe instructions into native x86 code on the fly. IBM filed suit in the Western District of Texas in 2022, asserting patent infringement and trade secret misappropriation, alleging that LzLabs reverse-engineered IBM’s proprietary technology to build the SDM. LzLabs disputed the infringement claims and filed its own Federal Circuit appeal on a pre-trial ruling in early 2026 (No. 24-2130).

IBM asserted five patents covering core aspects of mainframe instruction translation and emulation: U.S. Patent No. 9,804,823 (decimal floating-point operations), No. 8,190,664 (instruction sign encoding), No. 9,235,420 (branch target buffer for emulation), No. 8,713,289 (emulating condition code settings), and No. 7,434,209 (native code execution). These patents collectively cover the machinery a system must implement to faithfully replicate IBM mainframe behavior on non-IBM hardware.

The Court’s Holding

Judge David Ezra found that LzLabs’ Software Defined Mainframe infringed IBM’s five asserted patents. The court found that the SDM’s instruction-translation and emulation mechanisms read on the claims of all five patents — meaning that implementing a system capable of faithfully running IBM mainframe code on x86 hardware requires practicing IBM’s patented techniques.

IBM’s trade secrets claims remain pending; Judge Ezra did not rule on the trade-secrets component of the case at this stage, reserving it for further proceedings. The court’s patent-infringement ruling alone represents a substantial win for IBM, potentially exposing LzLabs and its customers to injunctive relief and damages covering the period of infringing use.

Key Takeaways

• The ruling reinforces IBM’s patent moat around its mainframe platform: any system designed to run IBM mainframe applications without IBM’s own software risks infringement of IBM’s instruction-set and emulation patents.
• Software-defined infrastructure companies that reverse-engineer proprietary hardware ISAs face significant patent exposure even if they do not copy source code — the underlying patented architecture can be enough.
• IBM’s five patents span low-level numeric representation, branching, condition codes, and native code execution — suggesting IBM has systematically patented the building blocks of mainframe compatibility, making it difficult to build a non-infringing “mainframe emulator”.
• The trade secrets claims still pending could result in additional liability, particularly if evidence of misappropriation of IBM’s proprietary documentation or APIs is introduced.
• Parallel litigation in the UK (IBM brought a related trade-secrets action against LzLabs in the High Court) adds international dimension to this dispute over mainframe platform rights.

Why It Matters

IBM’s mainframe business is one of the most profitable niches in enterprise computing, with customers paying high licensing fees for z/OS and related software year after year. LzLabs’ SDM technology represented a direct threat to that revenue stream by offering mainframe customers an exit ramp that did not require IBM licenses. The patent ruling neutralizes that threat and signals that IBM will aggressively defend its mainframe ecosystem through its patent portfolio.

For the broader enterprise software market, the decision underscores the risk of building “compatibility layer” products that emulate proprietary platforms. Even when source code is not copied, the architectural techniques required to achieve compatibility with a well-patented platform can themselves constitute infringement. Companies developing migration tools, emulation products, or interoperability layers for proprietary systems should conduct thorough freedom-to-operate analysis before commercializing.