PDF Solutions unveils Exensio Aurora for petabyte-scale chip data
PDF Solutions (Nasdaq: PDFS) has unveiled Exensio Aurora, a rebuilt analytics architecture for its Exensio platform, designed to handle semiconductor manufacturing data at petabyte scale and to support agentic AI workflows across fabs, test floors and global supply chains. A beta programme for a small cohort of early adopters begins in September 2026, with full demonstrations planned for the company's CONNECT 2026 event in San Francisco on 15–16 October.
The Santa Clara-based vendor positions the launch as a response to a structural shift in the semiconductor industry: the proliferation of 3D packaging, advanced heterogeneous integration and globally dispersed production has generated exponentially larger and more complex datasets. According to PDF Solutions, fewer than 5% of semiconductor manufacturing data is currently analysed using conventional tools, a figure that underpins the company's argument for a purpose-built, domain-specific analytics platform.
What Aurora introduces
Exensio Aurora departs from its predecessor in design philosophy as much as raw performance. The new distributed analytics engine is said to deliver approximately 25 times faster query performance at comparable hardware cost, using a server-side computation model that sends only the data required for a given visualisation to the client rather than pulling full datasets across the network.
The architecture is built around four principal components. A Manufacturing Data House evolves the platform's semantic model to support 3D and higher-dimensional data structures, custom metadata, and AI-ready data ingestion from virtual or logically generated sources, not just directly measured sensor outputs. A Model Lifecycle capability, built on Kubernetes, is designed to support training, simulation and deployment of models to distributed edge locations, including outsourced assembly and test houses (OSATs). An Agentic AI layer uses customised LLM-enabled agents, grounded in the company's 35-year semiconductor domain knowledge, to assemble and orchestrate multi-step workflows, with on-premise deployment options available. A Workflow engine sits beneath all of these: every analytic, rule, machine-learning pipeline and batch job is expressed as a workflow, which PDF Solutions says acts as an institutional memory and a guardrail against AI hallucination.
Chief executive John Kibarian said the growth of hybrid packaging and complex global supply chains demands a new class of AI-driven collaboration. "Delivering the operational efficiency that the industry needs requires a new kind of AI-driven collaboration, which can only be achieved when AI is applied at scale to align and analyse data from across the ecosystem, built for semiconductor," he said.
Market context and competitive landscape
The semiconductor analytics market sits at the intersection of specialist manufacturing execution systems (MES) and general-purpose cloud data platforms. General cloud providers have made inroads into enterprise analytics, but PDF Solutions argues that hyperscaler platforms lack native integration with fab equipment protocols, domain-specific semantic models, and supply chain orchestration capabilities that chipmakers require.
Competitors in the specialist semiconductor operations and analytics space include Onto Innovation, Synopsys's manufacturing intelligence products, and a number of point-solution vendors covering yield management and equipment monitoring. The agentic AI angle also places Exensio Aurora in indirect competition with broader industrial AI platforms being developed by companies such as Palantir and C3.ai, which serve process-intensive industries but do not carry deep semiconductor-specific ontologies.
PDF Solutions' product is also subject to the broader dynamics of the chip industry cycle. Capex tightening at leading-edge fabs in periods of oversupply can delay analytics platform investments, even when the underlying efficiency case is strong. The company's Gainshare model, where a portion of revenue is tied to customer yield improvements, aligns its incentives with customer outcomes, though it also introduces revenue variability.
The September beta release will be a key indicator of commercial traction. Named customers, disclosed performance benchmarks from production deployments, and pricing clarity will be the metrics by which the market will judge whether Exensio Aurora's claimed 25x performance improvement translates into accelerated adoption.