Bloomberg Tech at CES

Welcome to a special Bloomberg Tech from CES in Las Vegas, where we’re bringing you big conversations from the floor. We kick off with AMD’s Lisa Su, then dig into EVs and robotaxis with Lucid, and wrap with Razer’s AI push.

Helios and that new two-nanometer accelerator drew a lot of attention on stage—why is this such a leap?

Helios is a massive rack-scale system and MI455 is a major step up because AI demand is exploding from roughly a billion users toward several billion. It’s built on advanced two and three nanometer tech with hundreds of billions of transistors to deliver far more performance per system.

When will it land in the real world, and who’s first in line?

You’ll see it in the second half of 2026 with strong partners like OpenAI and Oracle ready to deploy.

You also announced MI440 for enterprises; what problem are you solving and what do customers want to do with it?

AI is heterogeneous, so MI440 reuses our core building blocks for on-prem and private cloud so companies can upgrade existing data centers. We’re seeing active demand across workflows, finance, and healthcare where control, privacy, and predictable operations matter.

Why pick your accelerators over a rival’s for on-prem or edge?

We combine leading performance with strong total cost of ownership and an open ecosystem that lets customers stand up systems quickly with partners.

How big is the compute shortfall right now?

Usage is around a billion people and racing toward many billions, and we estimate roughly a hundred zettaflops globally today yet need on the order of a hundred times more in a few years—pushing toward yotta-scale. Early gains, like big boosts in software development, hint at what that extra compute will unlock.

What’s actually constraining deployment—memory, energy, or something else?

It’s a full-stack challenge where power, high-bandwidth memory, silicon, and manufacturing all need to scale together, which is why tight coordination across the supply chain is essential.

Do you aim to own more of what’s inside the server box?

We deliver turnkey systems while staying open—CPUs, GPUs, and some networking from us, integrated with industry standards and partners so customers can be productive on day one.

MI500 is touted as a massive jump—how did you get there?

By driving an aggressive cadence and co-design across hardware, software, and systems—MI455 is roughly ten times MI355, and MI500 stacks another order-of-magnitude through advances in process, packaging, and optimization.

Where do things stand on selling into China?

China is important; we secured licenses for prior-gen data center parts, are applying for newer ones, and remain optimistic while working with U.S. authorities and Chinese customers.

Investors want evidence of share gains, and a view on PCs.

MI455 is our inflection point backed by deep partnerships and a path to tens of billions in AI revenue by 2027; in PCs we grew sharply in 2025 on AI laptops and the Windows 11 refresh, see demand ahead but near-term memory tightness, and we’re leaning into enterprise notebooks and new AI development systems.

Can you already see a macro benefit from AI?

We’re seeing faster product cycles and more efficient operations internally, and I expect those productivity gains to spread broadly across enterprises.

You also unveiled a humanoid partner—what’s your physical AI strategy and business model?

We’ve long powered robots via FPGAs and embedded compute, and with new humanoid collaborations we’ll provide the on-device brain plus the training and inference behind it through end-to-end partnerships.

Has U.S. AI policy moved the needle, and what defines 2026?

A coordinated national approach and public–private efforts like Genesis are helping balance export rules with rapid U.S. buildout, and 2026 should bring faster model progress and tangible everyday impact—we’re still far from the ceiling.

Quick market check: AMD slips despite a bold compute roadmap, while Nvidia firms on Rubin details, self-driving and robotics momentum, and signs of China demand. Now, let’s turn to EVs and robotaxis with Lucid.

What makes your robotaxi different?

We’re pairing our luxury EV platform with Nuro’s driver and moving to paid service in under eighteen months, with a cleaner production-intent design that aims for a seamless rider experience.

How do Nvidia’s platforms and your consumer roadmap fit together?

For retail, we’re integrating Nvidia Drive this year and targeting highway Level 3 around 2028 and Level 4 by 2029. That track is separate from the robotaxi program, which advances in parallel.

In a robotaxi world, do people still own cars?

Both models will coexist—shared rides suit short urban trips, but families and enthusiasts still want their own car and the choice to drive or be driven. Great EV dynamics help tip that choice toward ownership.

After a tough 2025, how is the ramp and supply chain now?

We posted eight straight quarterly delivery records, more than doubled output, and cleared key bottlenecks; we’re localizing components like batteries to the U.S. to mitigate tariffs, while continuing to manage risk.

Where do you fit globally, including China?

We’re focused on luxury and moving into premium around fifty thousand dollars, not the low-cost segments where Chinese players dominate, and we have no plans to enter China’s mass market.

Razer is known for gaming gear; what does an AI ecosystem mean for your users?

Beyond hardware, our platform spans over one hundred fifty million users and tens of thousands of developers, and we’re shipping AI tools and experiences for creators and gamers across software, services, and devices.

Walk us through the new headphones—why this form factor over smart glasses?

Project Motoko adds cameras, microphones, and on-device intelligence to a familiar headset, enabling features like live translation and richer mixed experiences. Gamers will adopt first, and we expect broader audiences to follow.

You’re investing six hundred million dollars—where is it going, and how do you address fears of AI slop?

We’re funding R&D, talent, and compute for multimodal AI, focused on tools that speed QA and workflows so studios can pour more energy into creativity; we’re augmenting great games, not replacing them.

How’s the supply chain after a volatile 2025?

With a balanced footprint across the U.S., Europe, and Asia, we’ve worked through component and logistics snags and continue to adapt.

On the CES floor, what stands out in robots?

There’s a dedicated robot zone this year, and live demos will separate genuine capability from tightly staged hype.

Are AI wearables moving beyond glasses?

Yes—many brands are trying non-glasses form factors that pack similar sensing and assistant features into items like headphones or even jewelry, often with multimodal AI.

What’s the mood among device makers?

Memory supply and pricing are hot topics, and companies are softening AI’s image with friendly designs, from desktop avatars to robot puppies.

That’s it for this special CES edition; we’ll be back shortly with an exclusive sit-down with Jensen Wong, so stay with us.