Mundo Secures $24 Million to Advance Perceptual Intelligence for Multimodal AI

The remaining $4 million — a previously undisclosed seed — brings the company's haul to a figure that, in the current AI capital cycle, barely qualifies as table stakes.

The cap table

GreatPoint wrote the Series A check. Y Combinator, E12 Ventures, and Montana-based Next Frontier Capital filled out the syndicate. No lead investor here is writing the kind of check that demands board control — the round is small enough that dilution stays manageable. But Mundo is still a 30-person data and evaluations outfit with no disclosed revenue, and $24 million in the bank buys roughly 18 to 24 months of runway before the next raise, assuming headcount scales as planned across research, engineering, and operations.

The product thesis

Mundo sells what it calls "the data layer for perceptual intelligence" — datasets, evaluations, and applied research aimed at labs building multimodal systems that process audio, video, and other unstructured sensory input. The argument, as reported: today's large models handle structured reasoning well but remain blind to the physical world. CEO Jason Liao, COO Naijide Anwaer, CPO Garreth Lee, and CTO Kenneth Wu founded the company in 2024 around a multilingual data library before reframing the mission around perception.

The reality check

The thesis is plausible — perception is, in fact, the obvious bottleneck — but the competitive set is no longer theoretical. Wonderful closed a $550 million Series C at a $5 billion valuation eighteen months after founding, to coordinate enterprise AI agents. Horizon3.ai is approaching $100 million in ARR after a $250 million Series E that valued it above $2 billion. OLIX, a London-based inference chipmaker, raised $312 million at $3.3 billion for dedicated silicon. Mundo's $24 million buys a seat at a table where the buy-in just went up materially. The next round, if there is one, will price against multiples set by companies that already have product, revenue, or proprietary hardware. Capital flowing into AI is increasingly bifurcated across private equity and venture vehicles, and Mundo's data-layer pitch only holds if customers keep paying for it after the labs internalize the work.