It appears to be a big year for Arm; you reportedly topped your revenue of $1 billion for the first time. What do you attribute this spike in demand to?
If you take our early history—late 90s, early 2000s—we were formed off of power-efficient chips for mobile. Apple Newton is the most famous example, a great invention for its time, but never really productized. From there, the ethos of the company has been in power efficiencies, first powering the mobile generation. We’re now in almost every mobile phone—whether it’s an iPhone or an Android.
That’s where most people know us from, but we've diversified across other segments. The data center is one of the fastest-growing for us—think hyperscalers like Google, Amazon, and Microsoft. We expect to have about 50% market share in cloud infrastructure by the end of 2025, largely due to the AI tailwind. Automotive has also been gaining traction —we announced a major collaboration, and customers like Tesla, Rivian and BMW are all using Arm. We’re seeing a future where 70% of inference could move to the edge, including wearables and earbuds.
Your recent £250m deal with Malaysia gives the country access to advanced chip design and blueprints. Given the global rise in semiconductor demand, which specific regions are you seeing the most interest from?
The U.S continues to be a big market for us. We're obviously a global company and seeing demand growth in all of our regions, and APAC continues to be a big market—Malaysia, Japan, Korea, China, India. We just announced with our partners that we will bring our technology into the Gulf region as well.
With mounting pressure to reduce electronic waste and the rising energy demands of AI and data centers, what steps are you taking to make your chip designs more sustainable?
We work closely with customers to design chips using our energy-efficient processors. Our instruction set is streamlined compared to alternatives, which makes us inherently energy efficient. Chips account for about ~50% of AI data center energy use; the rest goes to supporting systems like cooling. We’re also optimizing software to align with our processors and give customers a full stack—from software to hardware. Grace Blackwell from NVIDIA is a prime example; it’s the leading super chip for hyperscalers and is built on Arm. Amazon, Microsoft, Google—they're all designing AI systems using Arm because of the higher performance per watt. That allows them to be more sustainable, or put more computers in the same size data center.
Our key differentiator is optimizations at the hardware level. We do things at the instruction set, board-level design, and even work with partners on how components are laid out on the board. Compared to alternatives, we can optimize energy efficiency by up to 40%. That’s a huge cost saving when you think about wattage in AI data centers going forward.
Arm recently launched Compute Subsystems to support artificial intelligence applications across cloud and edge environments. Can you break down some real-world use cases?
Arm CSS offers a ready-made foundational system with built-in software and tool support, which has been a real breakthrough for our customers in terms of getting products to market faster. We’ve seen design start to tapeout in as little as 13 months and the ability to save up to 80 years of engineering effort. A really great example using Arm Compute Subsystems is Microsoft. They used Arm Compute Subsystems for Neoverse to develop their own custom silicon, called Cobalt, which now runs in their Azure datacenters. Building on our foundation, Microsoft accelerated development significantly.
A good example of AI at the edge is a Nest thermostat, which runs on Arm. Your car—brake assist, powertrain, steering and Tesla self-driving mode—where it processes all sensor data in and around the car to make decisions, all run on Arm. Think about AirPods—they're all based on Arm. Your iPhone uses Arm for photographic filters, tagging people or locations in photos—that’s all done on Arm’s compute platform. Gemini is another good example—it runs on our platform.
Balancing demand for high-performance chips with minimizing energy consumption must be a constant push and pull. How do you approach that?
It is a constant push and pull. One of the things we do well is working closely with our ecosystem. We design chips years ahead of when they ship, so we must consider sustainability from the start. We work closely with fabs like TSMC to ensure process nodes are power efficient years in advance. We optimize the instruction set. We optimize how our chips talk to each other and the software that runs on them. It's a constant reevaluation—there’s no one-size-fits-all answer.
What would you say are the major barriers to achieving your objectives in 2025?
There’s obviously a lot of uncertainty in the macroeconomic climate. That said, there’s still a lot of demand for AI. You’ve seen the CapEx numbers from the Magnificent 7 during earnings season. We’re not seeing a slowdown in interest or demand for our compute platform.
Over 300 billion Arm chips have been shipped, but despite that, people might not fully understand where we play. So a big focus is educating the world on who we are and where we're headed. We're trying to show how we can drive energy efficiency, which is key to scaling AI.
I always say, “No power, no AI.” That holds true for where we need to go next with artificial intelligence.
