Morph Labs Unveils Liquid Metal, Powering the Agentic Future

Jesse Han, Founder and CEO of Morph Labs, took the stage at the AI Engineer World's Fair in San Francisco to articulate a profound shift in artificial intelligence: the transition of AI agents from experimental tools to practical co-workers. This evolution, Han argued, demands an entirely new infrastructural paradigm, one that current cloud computing environments are ill-equipped to provide. His presentation centered on Morph's groundbreaking solution: Infinibranch, and its latest iteration, Morph Liquid Metal.

Han posited that the current technological landscape leaves advanced AI in a state of isolation, stating, "Thinking at the speed of light must be just as lonely as moving at the speed of light." He elaborated on the machine's inherent longing for "infinite possibility," a world capable of reacting to its thoughts with zero latency. This vision underscores the fundamental bottleneck in scaling intelligent agents: the underlying infrastructure.

Morph Liquid Metal, building on the previously released Infinibranch, directly addresses this constraint by reimagining virtualization, storage, and networking. The new offering boasts startup times under 100ms, snapshotting memory and process states in under 200ms, and the ability to branch running virtual machines to dozens of instances in milliseconds. This unprecedented speed and flexibility also extend to indefinite persistence and autoscaling to zero and infinity, with GPU support slated for Q4 2025.

The core innovation behind this capability is what Han termed "Git for compute." He demonstrated how Infinibranch makes "all mistakes become reversible," allowing agents to explore diverse computational paths without irreversible consequences. This "reasoning-time branching" enables AI to efficiently simulate and test multiple solutions in parallel, scaling against verification processes to find optimal answers.

This foundational shift in infrastructure is not merely about speed; it's about enabling a new era of AI reasoning. Han envisions a future where multi-agent systems can branch and reverse their operations, utilizing software during reasoning time, and scaling against rigorous verification. To lead this ambitious charge into "verified superintelligence," Morph Labs announced Christian Szegedy, former xAI co-founder and Grok-3 code reasoning lead, as their Chief Scientist. Szegedy’s expertise in formal verification and large language models is pivotal to developing Magi-1, an Infinibranch-native verified reasoning agent slated for Q1 2026. This collaboration signals a concerted effort to build the infrastructure necessary for AI to interact with complex software environments at unprecedented scale and with algorithmic guarantees.