5 PearX Startups Winning VC Buzz at Demo Day

PearX, the elite 12-week accelerator run by Pear VC, revealed its latest cohort in San Francisco last week. With just 16 companies selected out of an already-limited batch, the demo day once again drew the sharp focus of investors. Here are five standout startups that captured serious VC attention.

Speridlabs: 3D Spatial Modeling Without Limits

Speridlabs is building what it calls “spatial foundational models” to bring more intelligent 3D environments into robotics, gaming, and visual effects. Whereas firms like Runway, Odyssey, and Google’s Genie offer static or hard-coded models, Speridlabs’ “Mundus” platform lets users alter geometry dynamically—maintaining persistent structure even after changes. It’s a bold move in a space crowded with competitors making model queries rigid and uneditable.

Saia: Redefining On-Device AI Hardware

Saia is tackling one of the biggest bottlenecks in local AI: memory demands. Its chip bypasses traditional memory types by running inference directly from flash storage. In tests, it delivers significantly higher speeds, achieves eight times the capacity, and consumes one-quarter the power of Nvidia’s Jetson—a leading edge AI board. The startup is already in talks with Samsung for memory integration, plans small-scale silicon runs next year, and aims for mass production by 2028.

Ren: A Privacy-First AI Assistant

Ren aims to compete with tools like Muse and Instinct by emphasizing user privacy. It stores as much data on the device as possible, tunnels what it must into private clouds, and enforces guardrails defined by the user. Unlike some rivals, Ren automates voice actions entirely—no human operators involved—offering a hands-off personal assistant with strict privacy boundaries.

Veros: Bringing Trust Planning into the AI Age

Veros sees big inefficiencies in traditional trust and estate management. Using AI, it aims to simplify what lawyers, wealth managers, and administrators now do manually. The company already oversees $250 million in assets under management and is working to acquire a trust charter so it can operate as a regulated trust entity itself—moving beyond pure tech to take on legal and financial responsibilities.

Datum: Accelerating Industrial Design via 3D Intelligence

Datum is solving for slowdowns in physical product development by helping engineering teams locate old designs quickly. Its proprietary Geometric Fingerprint tech indexes libraries of 3D parts and surfaces, recognizing pieces by shape to speed up reuse and reduce reinvention. Especially for companies with vast engineering backlogs, this could slash costs and timelines.

PearX continues to set itself apart from accelerators like YC by keeping cohort size small, keeping startups confidential until demo day, and occasionally investing up to $2 million in select companies. In this recent batch, Saia and Veros impressed with hardware and financial infrastructure, while Speridlabs, Ren, and Datum leaned into privacy, 3D modeling, and AI-powered tools.

Analysis: These five companies show where early-stage VCs are placing their bets—on bringing AI closer to the edge, enforcing user privacy, and automating longstanding inefficiencies. Keep an eye on which of them can execute at scale: securing hardware manufacturing, regulatory approvals for trust operations, and adoption among engineering teams will be key. If they pull it off, PearX’s next demo day may look like an inflection point for infrastructure, privacy, and industrial AI.