Hardware

Intel announced its collaboration with Argonne National Laboratory to develop Aurora genAI, a generative AI model with 1 trillion parameters, five times larger than ChatGPT. Built on Nvidia’s Megatron and Microsoft’s DeepSpeed frameworks, it's designed for scientific applications and aims for completion in 2024.

Nvidia CEO Jensen Huang unveiled the DGX GH200 Grace Hopper, the most powerful generative AI processor, at SIGGRAPH. This superchip features 5TB/s memory bandwidth and 141GB capacity, enhancing AI capabilities. Nvidia also partnered with Hugging Face for seamless model running and launched new workstation GPUs. The DGX GH200 aims to revolutionize AI, with production starting in Q2 2024.

MediaTek launched its Dimensity 9300 chip, featuring an all-core architecture and generative AI, showcased by Huile's innovative tools like AI glare reduction and Image to GIF generation. Their collaboration aims to enhance mobile AI experiences and redefine user interaction with advanced technology.

The Snapdragon Summit launched on October 25, 2023, showcasing Qualcomm's Snapdragon 8 Gen 3, designed for generative AI with enhanced performance and efficiency. Key unveilings include the Snapdragon X Elite PC processor and new audio platforms, alongside Snapdragon Seamless for multi-device connectivity, signaling significant advancements in mobile AI applications.

On November 14, NVIDIA announced the H200 AI processor upgrade, featuring advanced memory for better data management. Major cloud providers will adopt it in 2024. Despite competition from AMD and Intel, NVIDIA aims to maintain its AI leadership, facing challenges like sales restrictions in China.

Qualcomm unveiled the Snapdragon 8 Gen 3 processor at the Snapdragon Summit, targeting flagship Android smartphones in 2024. It features an AI engine for generative AI, advanced camera capabilities, improved performance (30% faster), and energy efficiency. Major brands like Xiaomi plan to adopt it.

Artificial intelligence (AI) is revolutionizing content generation through AI-generated content (AIGC). Intel leads this transformation with innovations like the Evo certification program, the Core Ultra processor, and OpenVINO. These advancements enhance AI performance in PCs, supporting creativity while ensuring efficient hardware-software integration.

NEO Semiconductor launched its 3D X-AI chip technology on August 18, replacing DRAM in high-bandwidth memory to boost AI performance and cut energy use. The chip features 8,000 neuron circuits, achieving up to 100-fold AI acceleration and 99% power reduction, while supporting larger AI models with eight times the memory density.

The memory chip market is experiencing mixed trends due to AI demand in data centers and a downturn in consumer electronics. Strong demand for High Bandwidth Memory (HBM) and Solid State Drives (SSDs) contrasts with weak retail sales, impacting flash memory prices. Energy efficiency is becoming critical as AI workloads rise, shifting competitive focus among manufacturers.

Recent advancements in AI have led to a 38.4% increase in global AI server shipments in 2023, with China thriving in this sector. ZTE Corporation launched the high-performance R2350 and R2550 G6 servers to enhance AI computing capabilities. Their diverse product range positions ZTE as a key player in the expanding AI market, projected to reach $1 trillion by 2027.

Figure AI launched the Figure 02, a next-gen humanoid robot featuring a complete redesign and advanced AI, enhancing its movement, intelligence, and interaction capabilities. The humanoid robotics market is projected to grow significantly, reaching 300 billion yuan by 2035.

The storage industry is recovering, particularly due to advancements in High Bandwidth Memory (HBM). Micron Technology reported significant revenue growth, boosting market confidence, while Samsung and SK Hynix are also expanding HBM production. Despite forecasts of potential oversupply, demand driven by AI suggests a positive outlook.

Apple Reports 10% Decline in iPhone Sales Amid Competitive Challenges

The Rabbit R1, showcased at CES 2024, aims to streamline smartphone app interactions by learning user behaviors and intentions, allowing for AI-driven task execution without manual navigation. Weighing just 115 grams, it features a touch-sensitive display, a voice interaction button, and a 360-degree camera. Utilizing Rabbit's unique Large Action Model (LAM), it performs tasks through AI agents, or “rabbits,” capable of complex operations beyond traditional assistants. The device prioritizes user privacy and is available for preorder in the U.S., with shipments starting in March.

Following the launch of SearchGPT, criticisms arose over its inaccuracies, such as providing the wrong date for a North Carolina music festival. OpenAI acknowledged the errors, focusing on refining the prototype. Unlike Google's Bard, which faced significant backlash for inaccuracies, OpenAI is limiting access to mitigate issues. Analysts warn that even a small error rate could result in numerous daily inaccuracies, as LLMs like SearchGPT inherently have hallucinations. The future of search engines hinges on whether LLM creativity can coexist with traditional search reliability.