Power & Sites · United States

Axelera AI: Data Center Inference Performance in the Power Envelope of Embedded Systems

Credited to EE Times · eetimes.com · 23h ago

Photograph · EE Times

Partner Content

Physical AI is hitting three walls: power, economics, and fragmentation. Fabrizio Del Maffeo explains how to break them.

Recorded at the AI Infra Summit 2026 in Santa Clara, California, Axelera AI co-founder and CEO Fabrizio Del Maffeo sits down with EE Times Sally Ward-Foxton on why intelligence was never the bottleneck to AI adoption. Infrastructure is.

The three walls to scaling physical AI

Del Maffeo’s keynote framed three barriers standing between today’s pilots and real deployment.

Power and physics. The data center was built for training: persistent, high-wattage, liquid-cooled, and tolerant of moving data long distances. Physical AI is the inverse. Inference happens where data is created—in factories, vehicles, hospitals, and city infrastructure—inside power envelopes measured in tens of watts. Shrinking a training architecture does not fix that, because the dominant cost is not computation; it is data movement. Axelera’s Digital In-Memory Compute performs matrix-vector multiplication, roughly 80% of the inference workload, inside the memory cell itself. Less movement, less energy, more work per watt.

Economics and control. Cloud inference made sense for experimentation. For production workloads running continuously, the bill never stops growing, and in regulated industries, the data cannot leave the building in the first place. Del Maffeo argues the decision has shifted from a cost question to a control question: your models, your data, your cost structure.

Fragmentation. Every new accelerator brings its own toolchain, conversion step, and lock-in. Developers pay that tax, and it is why projects stall between demo and deployment.

Start at the edge, then scale up

Most semiconductor companies design for the cloud and strip the architecture down for smaller devices. Axelera did the opposite: solve the hardest, most power-constrained environments first, then carry that efficiency up. Europa, launched at the show, is what that looks like at enterprise scale.

Europa is a purpose-built inference accelerator, delivering 629 TOPS at INT8 with eight second-generation AI cores, 16 on-chip RISC-V vector processors for pre- and post-processing, LPDDR5 at 200 GB/s, hardware HEVC/H.265 decode, and a Kudelski KSE3 secure enclave, at 35 watts typical during LLM inference. It ships as the Axelera Edge 232p, a half-height, half-length PCIe card, and the Axelera Server 250p, a full-height, full-length card with four AI processing units. Standard slot, existing server, no new rack, no new cooling.

Software is the missing link

Silicon is the easy half. Del Maffeo makes the case that hardware and software have to be co-designed, and that manual optimization is where developer velocity dies. The Voyager SDK takes PyTorch models directly with no intermediate conversion, covering CNNs, transformer vision models, VLMs and LLMs, with quantization and graph optimization handled by the compiler. Voyager Wingman, the new agentic developer assistant, turns plain language into a working pipeline, runs it on real hardware, and iterates until it performs. Built on RISC-V, open by design, no proprietary lock-in.

Ecosystem validation

A chip alone deploys nothing. Del Maffeo closes on the partner ecosystem, from defense and agricultural robotics to retail analytics and rail, and how system integrators, model builders, and hardware vendors working against one toolchain are collapsing the traditional 12- to 18-month deployment cycle.

https://axelera.ai/ai-accelerators/aipu/europa

https://axelera.ai/news/axelera-ai-launches-europa-delivers-physical-and-enterprise-ai-through-growing-partner-ecosystem-including-dell-and-supermicro

https://axelera.ai/ai-software/voyager-sdk#wingman

Power & SitesUnited States

Caterpillar and CoreWeave shorten the learning loop for physical AI

Credited to SiliconANGLE · siliconangle.com15h ago

Useful

Demand for new data centers, power plants and highways is fueling a construction boom, even as the industry wrestles with declining productivity and a shortage of skilled machine operators. Physical AI, which lets machines perceive their surroundings and act on them, is emerging as one answer.

Caterpillar Inc. has years of experience with autonomous equipment in mining. Extending that technology to construction requires systems that can adapt to more variable conditions, according to Brandon Hootman (pictured, right), vice president of physical AI platforms and construction autonomy at Caterpillar.

“Once that mine site gets instantiated, it does change, but it doesn’t change frequently,” he said. “You go to construction as an industry. Polar opposite of mining. If you think about all of the dynamics of what has to happen, taking a structured system and matching it to an unstructured environment, it’s really, really challenging to do.”

Hootman and Richard Ahlfeld (left), senior vice president of Physical AI at CoreWeave Inc., spoke with theCUBE Research’s Dave Vellante and John Furrier at the Fully Connected event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed why physical AI reshapes the infrastructure behind autonomy and how the two companies are shortening the learning loop for construction machines. (* Disclosure below.)

Physical AI changes what data centers must deliver

AI clouds were first built to train foundation models, then tuned for agentic inference at scale. Training an autonomous excavator means ingesting telemetry and vision data, simulating a digging scenario a million times and layering reinforcement learning on top. CoreWeave recently launched a Physical AI Field Engineering service that embeds its engineers with customers’ domain experts, according to Ahlfeld.

“Physical AI now is an entirely different beast,” he said. “That requires, first of all, a lot of storage. It requires a different infrastructure.”

Caterpillar’s digital ecosystem already holds about 18 petabytes of federated data from machines, dealers and customers, Hootman noted. Training autonomous equipment also requires perception data synchronized with machine control and performance data.

“That 18 petabytes that we’ve collected so far is just a drop in the bucket to the amount of data that it actually takes to go train physical AI to operate in a world like a construction site,” he said. “But you take [Light Detection and Ranging] data, camera data, that multi-second control data and performance data coming back in, you’re talking about terabytes of data within a given day for just one machine.”

CoreWeave named Caterpillar among its enterprise customers in its second-quarter results announcement. Caterpillar began working with the company this year, drawn by both graphics processing unit capacity and applied expertise, according to Hootman. Working with Nvidia, the partners use AI models to annotate and label incoming field data.

“Now you’re taking things that were months to maybe weeks,” Hootman said. “And now you’re getting it down to hours and your feedback loop between that happening and being able to use it in your simulation environment, or your training environment, happens within the given workday.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Fully Connected event:

(* Disclosure: TheCUBE is a paid media partner for the Fully Connected event. Neither CoreWeave, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from our Marketplace portal page and links: https://siliconangle.com/aws-marketplace/

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.

Atlassian lays groundwork for humans and AI agents to work side by side

Google goes nuclear to power its AI data centers, committing to buy 890 megawatts of energy from Constellation

Meta teams up with Bret Taylor's Sierra Technologies on new standards for AI agent commerce

Multiply Labs raises $75M to optimize pharmaceutical manufacturing with robots

In the agentic era, standing privilege must go: Key thoughts from the SailPoint Navigate keynote

Anthropic folds Project Glasswing into an expanded three-tier Cyber Verification Program

INFRA - BY MARIA DEUTSCHER. 10 HOURS AGO

SECURITY - BY ZEUS KERRAVALA. 10 HOURS AGO

SECURITY - BY DUNCAN RILEY. 12 HOURS AGO

Join our community

Original · SiliconANGLE

Front Method