Watch the full interview on Bloomberg
Earlier this month, I joined Bloomberg Television’s Paul Allen on Insight with Haslinda Amin, just after we reported our second-quarter results. We covered a lot of ground in a short window — where our growth is really coming from, how a 37-year-old hardware company stays relevant as computing shifts toward AI-driven services, and how we’re managing a memory shortage that’s squeezing the entire industry. A few of the points I made are worth expanding on here.
The growth is real, and it’s broadening
Our server business is in hyper-growth: It grew around 200% YoY and approximately 58% QoQ in the second quarter. At the beginning of the year, we targeted full-year AI server business growth of between 50% and 100%. But, given the strong momentum, I told Paul that I’m now confident we can achieve around 150% growth for the full year. AI server visibility remains strong, supported by continued demand for compute infrastructure.
That demand is also important in the contesxt of the broader PC market, where ASUS continues to outperform. Among the top five Windows PC companies, and amid increased headwinds, ASUS was the only one to continue delivering growth. Our commercial PC shipments are up more than 50%, and AI PCs are taking a larger share of consumer revenue – over 30%.
This isn’t a one-quarter spike. We’re in the first wave of mass production on the NVIDIA Vera Rubin platform, and our order visibility — across Neo Cloud customers, large enterprises, research institutions, and government-led sovereign AI programs — extends well into 2027. I see three engines carrying us through next year: AI servers and infrastructure, premium PCs driven by the AI PC upgrade cycle, and a fast-growing layer of AI applications and services.
AI stopped being a feature. It’s the company.
The question I get most is how a hardware company stays relevant as value moves toward software and AI-driven services. My answer is that ASUS isn’t trying to defend hardware against that shift. We’re building the full stack: AI devices, AI infrastructure, and AI applications, from the cloud to the edge. That’s what “Ubiquitous AI. Incredible Possibilities” means in practice — not a tagline, but the design brief for everything we ship.
Agentic AI needs strong local compute, and we’re seeing that play out as the industry moves back from thin client toward heavy client — which plays directly to our strengths. Our ASUS Zenni Claw is a good example: one-click setup, ready-to-use ASUS skills, flexible local-cloud routing, and built-in safety protection, all aimed at making agentic AI practical rather than theoretical. We’re pairing that with a services layer — the ASUS AI Hub, healthcare and smart-city deployments, and sovereign AI infrastructure through Taiwan AI Cloud — so the hardware and the intelligence running on it are inseparable. That integration is exactly where we intend to compete.
Generative AI will continue to help businesses improve productivity, but the opportunity extends much further. LLMs will extend AI from cyberspace into the physical world. As physical AI becomes increasingly important, every physical AI device will require significant computing power
Competitive differentiation
We compete against companies many times our size, and specifications get matched quickly. But what doesn’t get matched as easily is culture. We’ve run Design Thinking as a formal discipline for more than 15 years, starting from real user needs rather than a component roadmap. We combine that with what our Chairman calls Extreme Engineering — going back to first-principles physics and thermals to solve trade-offs the rest of the industry treats as fixed. That discipline is how we’ve landed in the first wave on every major new platform, from the ASUS AI POD on Vera Rubin to being among the first with NVIDIA’s RTX Spark in our ProArt line.
The honest answer on the memory shortage
I don’t think the memory and chip shortages will ease quickly. When Paul asked me how long I expect the constraints to last, I said they could continue until the end of 2028. The fundamental reason is the strong demand for AI servers, combined with memory manufacturers prioritizing AI and data center requirements over consumer markets. While suppliers are expanding production capacity, building new facilities and installing the equipment required to increase output takes time.
We’ve responded on four fronts: building strategic inventory ahead of the curve, locking in long-term supplier agreements, designing flexibility across configurations, and prioritizing higher-value products. While we’ve adjusted prices on some lines, our approach is to absorb what we reasonably can and be transparent with customers about the rest — prices reflect real input costs assssssssssscross the industry, and our job is protecting the value customers get for what they pay. Wherever we operate, we meet our customers’ requirements and comply with the regulations of the countries and markets wsssssse serve.
Where this goes next
What I hope came through in the conversation is that ASUS is not treating AI as a single product category or short-term trend. Our ambition is much broader: to become the world’s most comprehensive AI company, connecting devices, infrastructure, applications, and services in ways that make AI useful everywhere. I was grateful for the opportunity to discuss that vision with Paul and Bloomberg, and I look forward to continuing the convsersation as this next chapter of AI takes shape.SS
