[Las Vegas, United States = Digital Today reporter Jin-ho Lee] Dell Technologies named data management as a differentiator for getting ahead in the competition over artificial intelligence (AI) infrastructure. It said capabilities to manage data scattered across companies in a way suited to AI have become more important.

Travis Beihl (트레비스 비힐), a senior vice president at Dell Technologies, said in an on-site interview at Dell Technologies World 2026 (DTW26) in Las Vegas on May 18 that data management is a major challenge. The ability to connect the right data to the right use cases will become very important, he said.

Beihl, who oversees infrastructure solutions product management, explained storage’s role in AI workloads by linking it to graphics processing units (GPUs). A storage infrastructure that can supply data quickly is needed to make full use of GPUs, he said. He stressed that the Dell AI Data Platform can solve problems customers face in adopting AI.

The Dell AI Data Platform is largely made up of three layers. The bottom layer handles protocol support and performance, including file, object and parallel file system (PFS). The middle layer is the data engine. It converts and queries structured and unstructured data and prepares it for use in generative AI. The top layer is the Dell orchestration engine, which supports data cleansing and processing, metadata enrichment, vectorisation, and matching data with large language models (LLMs).

Beihl said the three layers of the Dell AI Data Platform are directly linked to what customers consider important. Securing the right speed and performance, converting data and managing data are key, he said. He also highlighted the fast speed of the Dell Lightning File System, a PFS product. Beihl introduced the Lightning File System as the fastest parallel file system on the market.

Dell is also helping with AI use in manufacturing. Beihl cited Samsung Electronics as an example of supporting that area. He said electronic design automation (EDA) workloads for semiconductor manufacturing require both computing and storage capacity because designing semiconductors involves handling numerous drawings and image-type data while precisely matching geometric structures and tolerances.

Beihl said EDA ultimately involves handling many images or drawings. That makes it compute-intensive while also storage-intensive, he said.

Beihl again stressed data management as a challenge for the AI infrastructure market going forward. Dell, which acquired Data Loop last year, is building the Dell orchestration engine based on it.

He said larger companies in particular need to focus on data management because they accumulate vast amounts of unstructured data such as knowledge documents, service requests and customer conversations. Beihl advised that properly discovering that data and linking it to work is key to creating AI value.

Large companies have many knowledge base documents, customer conversations and service requests, and all of that exists as unstructured data, Beihl said. The ability to find the right data and provide it to the right target is how to unlock business value, he said.