Skip to content

Category

Hardware

56 events 7 over 7 days ↑ 133 % 17 sources report highest score 65

Importance: important NVIDIA Hardware ✓ official

NVIDIA's Vera Rubin platform won all MLPerf Training v6.1 benchmarks

NVIDIA's Vera Rubin platform debuted in MLPerf benchmarks, winning all MLPerf Training v6.1 tests and, according to the company, achieving over 10x higher energy efficiency for agentic AI inference in MLPerf Endpoints.

Importance: monitoring NVIDIA Hardware ✓ official

NVIDIA announces Rubin HPC GPU for simulations with AI agents

NVIDIA announced the Rubin HPC GPU in the Vera Rubin platform at the SC26 conference, designed to accelerate scientific and engineering simulations supported by AI agents. Technical specifications, price, and availability are not stated in the source.

Importance: monitoring Hardware 1 source

Indian startup Bharat Semi raises $24 million to develop AI image sensors

Indian startup Bharat Semi, developing programmable image sensors with integrated AI, has raised $24 million in a seed round. It has design wins with General Atomics and Kaynes Semicon, and plans commercial production for 2027.

Importance: monitoring Hardware 1 source

Company Delska opens the first AI data center in Latvia

The company Delska has opened the first Latvian data center for AI in Riga: 10 MW (expandable to 30 MW), liquid cooling for racks up to 250 kW, TIER III certification, 100% renewable energy, and use of waste heat for district heating.

Importance: monitoring Google Hardware 1 source

Google expands eighth-generation TPU lineup with 8tx and 8i Max variants

Google has released two new eighth-generation TPU variants: TPU 8tx for encrypting data during processing, and TPU 8i Max with doubled performance and memory thanks to combining two chips into a single package. These are configurations of an existing chip, aimed at expanding the offering for external customers in competition with Nvidia.

Importance: monitoring NVIDIA Hardware 1 source

NVIDIA integrates HBM3e and HBM4 memory into its accelerated computing stack

NVIDIA says it has become the first to deploy HBM3e and HBM4 memory in its GPU stack for accelerated computing. It is thus responding to the growing memory demands of large models and agentic applications amid tight supply and rising memory prices.

Importance: monitoring Hardware update ✓ 2

Nothing: Carl Pei rejects the AI wearables trend, screenless AI device targets December 2026

Nothing CEO Carl Pei has newly explained why his company rejects the AI wearables trend (pendants, pins, glasses) and called the search for the "iPhone of AI" a dead end. The upcoming screenless AI-native device is slated for release in December 2026, with details remaining unknown.

New: Nothing explicitly rejects the wearable trend (pendant, pin, glasses, companion), unlike the rest of the industry; Carl Pei expresses a different view of the AI future: 'I read this moment differently'; the new article confirms the productivity-oriented nature of the device with a screenless design

Importance: monitoring Amazon Hardware 1 source

Review highlights ethical issues of the always-recording Amazon Bee Frontier AI wearable

Victoria Song, senior editor at The Verge, describes in her review of the Amazon Bee Frontier AI wearable that she had to ask her therapist for consent to wear the device, which records continuously. The text highlights the ethical questions surrounding privacy protection.

Importance: monitoring Hardware 1 source

Huawei increases its share of the Chinese AI chip market, Nvidia falls to 34 percent

Nvidia's market share in China fell to 34 percent, while Huawei increased its share to 28 percent thanks to its Ascend chips. Huawei is preparing the Ascend 960 and 980 models with up to 384 GB of HBM memory, but according to the company it currently has no plans for the European market.

Importance: monitoring Hardware 1 source

Runware plans serial production of modular AI data centers at a factory in Romania

Runware plans serial production of modular Sonic Inference Pod data centers at a factory in Romania, aiming to reach a capacity of 30 MW per day by 2028. A second factory is to be launched at the beginning of next year, and pod installation is expected to take 48 hours.

Importance: monitoring Google Hardware 1 source

XREAL AURA glasses combine Android XR with Gemini assistance when working with apps

Google and XREAL have introduced XREAL AURA wired glasses running Android XR. According to Google, the Gemini assistant can also work with content on a mirrored laptop screen. Orders are opening in selected markets, with prices starting at 1 099 USD.

Importance: important NVIDIA Hardware update ✓ official

First deliveries of data centers based on the DSX POD design are planned for early 2027

The first worldwide deliveries of data centers based on the DSX POD design are planned for early 2027. According to NVIDIA, assembly and testing in containers before delivery can reduce installation time at a prepared site from six months to as little as 14 days.

New: The first worldwide deliveries are planned for early 2027.; According to Nvidia, installation can take 14 days instead of six months.; At a prepared site, the containers are connected to power, cooling and the network.; Partners include Dell, HPE, Lenovo and Schneider Electric.

Importance: monitoring Amazon Hardware 1 source

i14 and AWS describe changes to Laplace 3B for AWS Trainium2

The companies i14 and AWS described porting Laplace 3B to AWS Trainium2 using the beta version of TorchNeuron Native. According to the authors, training reached approximately 120 000 tokens per second; the recurrent state occupies 29.5 MB per sequence regardless of context length.

Importance: monitoring Amazon Hardware 1 source

Deepgram and AWS optimized batch transcription with Nova-3 for AWS Trainium2 chips

Deepgram and AWS ported the Nova-3 model to AWS Trainium2 chips. According to their measurements, a configuration with eight replicas on a single instance processed 11–70 % more audio per dollar than the compute options previously used. The results apply to batch transcription.