Carmakers are testing neuromorphic technologies based on spiking neural networks. What will cars do with their newly found brain power?

BrainChip converts CNNs to spiking networks to reduce power and enable on-chip learning at the edge.

Tenstorrent and Rapidus have partnered to jointly develop semiconductor IP for AI in the device based on 2-nm logic semiconductors.

The software licensed from Intel will form the basis of Mihira’s orchestration layer, but Mihira will be building its own IP on top.

ChipNeMo, trained on Nvidia’s internal repositories of code and text, can answer questions about chip design, summarize bugs and write scripts for EDA tools.

Condor Galaxy 1 is the first supercomputer-scale cluster of AI chips from any startup, validating the market for non-GPU hardware at scale.

The two companies have demonstrated SiFive’s RISC-V host processor works with BrainChip’s Akida NPU IP for edge AI SoCs

Intel joined RISC-V and is pumping money into the ecosystem, while Nvidia's Arm deal fell apart. How will this affect RISC-V adoption?

Processor in memory technique can enable extremely energy efficient AI inference and will be available on GloFo’s 22FDX process

AMD’s Vamsi Boppana admits software is a journey, but the open-source community can "help bridge the gap."

BrainChip’s Akida neuromorphic chip powers keyword spotting in the Mercedes Vision EQXX.

The startup, previously targeting at-scale AI training, is now going after generative AI inference with a new chip.

New product cadence speeds up to 6 months, market offers little competition for general-purpose FPGA under 100k logic cells

Look out for dozens of interesting AI chips, Intel, and Tesla keynotes and tutorials on CXL and MLIR at Hot Chips this year.

An EE Times Exclusive interview with Ayar Labs CEO Charlie Wuischpard explores the future of optical chip-to-chip interconnects.

Graphcore Bow IPU AI processor, the first commercial WoW chip, boosts performance up to 40% using same design

In this podcast, we discuss BrainChip's second-generation architecture for spiking neural network acceleration.

A founding father of tinyML co-founded Useful Sensors to help appliance manufacturers add AI capabilities to everyday objects with AI-enabled sensor modules.

Prophesee’s approach to commercializing its retina-inspired event-based sensors, and where the technology will go from here

Prophesee’s fifth-generation sensor is smaller, less power hungry and cheaper—all geared towards always-on #IoT applications.

AI chip startup Tenstorrent’s CEO and CTO have swapped jobs in a move "planned since Keller joined."

On this podcast: GPU giants battle it out, plus a conversation with SiFive’s Chris Lattner on developing AI chips software-first.

Embedded ML development platform company Edge Impulse will double headcount in 2022

Percieve’s chip uses information theory to run multiple large networks simultaneously

Lemurian Labs invented a new logarithmic number format designed for AI acceleration and is building a chip for data centers.

We discuss the commercial readiness of spiking neural networks and the potential for spiking LLMs with Intel's Mike Davies.

Sizeable Series C funding for Groq nonetheless dwarfed by competitors’ raises in this space

In this podcast, Sally speaks to Silicon Labs' Matt Maupin and Dan Kozin about the company's wireless IoT chips with AI acceleration, what use cases need this capability, how the accelerator fits...

AMD unveils its vision for AI, including its best bet to compete with Nvidia’s flagship H100 in the data center, the MI300X.

In the latest round of MLPerf inference benchmark scores, Habana performs competitively on LLM inference with Gaudi2.

Arm’s Jade Alglave invented a formal methodology for preventing concurrency bugs, which are otherwise extremely difficult to find and fix.

Open-source LLMs help promote transparency and foster trust, and researchers can use these models to identify risks and develop safeguards.

Intel and AMD are good candidates to play the unstoppable secular trend of our digital future. Read why recent corrections provide entry opportunities for both.

The legendary CPU architect Jim Keller discusses Tenstorrent’s strategy, RISC-V and the future of AI in an exclusive interview with EE Times.

MLPerf training and tinyML inference benchmarks show debut training scores for Nvidia H100, plus wins for Habana Gaudi2 and GreenWaves GAP9.

MLPerf Inference scores have become a two-horse race. Qualcomm, Nvidia and Intel all made the case they'd scored moral victories.

Synsense and Prophesee have joined forces to integrate a single-chip, event-based image sensor and neuromorphic processor.

Google, Graphcore, and Habana submit scores for brand new hardware accelerators while Nvidia sticks with 2-year-old A100

Groq CEO Jonathan Ross predicts increasing opportunities for inference hardware with LLM techniques like prompt engineering and reflection.

Quietly, but not secretly, building an end-to-end embedded developer platform for machine learning on microcontrollers

AMD unveils its vision for AI, including its best bet to compete with Nvidia’s flagship H100 in the data center, the MI300X.

Lightelligence Hummingbird connects 64 cores in an all-to-all topology via photonic links in silicon interposer.

Taiwanese AI accelerator maker Neuchips has demo’d Llama2-7B inference at 240 tokens/second on a 4-chip PCIe card.

Quadric has optimized its architecture for both AI and standard computer vision algorithms aimed at edge applications.

Eta Compute’s tool chain has gained a compiler and middleware that streamline edge AI development, plus a network zoo, sensor board, and tie-in with Edge Impulse for data ops.

Nvidia once again faced down Qualcomm in the data center and edge divisions of MLPerf

New TinyML benchmark has metrics for latency and energy consumption - and the first round of results are out

MLPerf training and tinyML benchmarks show off the range of AI-capable systems and their progress.