WD CEO Irving Tan on the architecture powering the AI era

Compute and memory may dominate the headlines, but the intelligence AI creates begins and ends with data. It has to be generated, moved, accessed, stored, and preserved efficiently and economically at enormous scale.

That makes data one of the most consequential conversations shaping the future of AI infrastructure.

In a recent conversation with theCUBE at the CASPA CEO Summit, WD CEO Irving Tan shared his perspective on how AI is reshaping data infrastructure, why the architecture behind it matters, and how WD is innovating for the opportunity ahead.

“Our view is actually AI is a data system.”

WD CEO Irving Tan joins theCUBE to discuss AI infrastructure, data architecture, storage innovation, and what comes next for AI.

Why is data at the center of AI infrastructure?

AI depends on an entire data ecosystem working together.

Billions of dollars are being invested in power generation, data centers, GPUs, networking, and other infrastructure to create intelligence from data. But creating that intelligence is only part of the equation. The data and intelligence also need to be stored, accessed, and preserved.

As AI workloads scale, those requirements become increasingly complex.

Different workloads require different combinations of performance, capacity, and economics. Data may need to move between hot, warm, and cold storage tiers. Training and inference create different demands. And throughout the process, infrastructure needs to make data available where and when it is needed.

The result is a broader way of thinking about AI infrastructure, one that extends beyond any individual technology.

Why does the complete AI data architecture matter?

The future of AI infrastructure will not be defined by a single device. It will be defined by how the complete architecture works together.

As Irving explained:

“You can’t think of it as a device or system-level solution. You need to think of it as a complete architecture.”

That architecture increasingly stretches across cloud, large-scale data centers, the edge, and devices themselves.

For WD, this broader approach is influencing how the company works with hyperscalers and AI innovators. The conversation extends beyond optimizing an individual drive to considering software, workloads, and the performance of the broader data center system.

As AI scales, capacity, performance, power, reliability, and economics have to be considered together.

How is storage innovation evolving for AI?

More capacity is only part of the answer. What that capacity enables is where things get interesting.

Higher-capacity HDDs can help data center operators improve rack density, power efficiency, and total cost of ownership. But capacity is not the only dimension where storage technology continues to advance.

Irving highlighted WD’s work to increase HDD performance:

“We’ve actually developed capability to increase the throughput from initially 2x all the way up to 8x and double the IOPS.”

For hyperscalers, AI labs, and emerging cloud providers, advances in capacity and performance can create new possibilities for how AI data is architected, stored, and accessed.

But technical advancement is only meaningful if it works economically at scale.

As Irving put it:

“Can we deliver it scalably? Can we deliver it reliably? And can we deliver it economically?”

Those three questions capture an important reality of AI infrastructure: innovation ultimately needs to translate into better outcomes for customers.

What happens as AI moves into the physical world?

Today’s AI infrastructure needs to be ready for applications that extend far beyond today’s workloads.

Physical AI, robotics, autonomous vehicles, and other intelligent systems have the potential to create enormous amounts of data outside traditional data center environments. Much of that data will likely be video-based.

“When we talk about physical AI, when we talk about humanoid robotics, autonomous vehicles, a lot of that data generation is going to be video-based. That’s going to drive a lot more storage.”

Some applications will also introduce new requirements around how long data needs to be retained, potentially creating new approaches to data tiering and archival storage.

The implication is clear: infrastructure decisions being made today need to account not only for current AI workloads, but also for increasingly distributed and data-intensive applications ahead.

Innovating for where AI is going

AI is changing both how intelligence is created and the infrastructure required to support it.

For WD, that creates an opportunity to continue advancing data storage across capacity, performance, energy efficiency, and economics while working with customers at an increasingly architectural level.

Irving summed up his directive to the WD team simply:

“Innovate, move with speed and agility.”

As AI infrastructure continues to evolve, that mindset will be essential to transforming extraordinary amounts of data into intelligence that can be accessed, preserved, and applied at scale.

Because ultimately, the future of AI isn’t only about how much compute we can deploy.

It’s about what we can do with the data behind it.