Tesla has released an updated AI accelerator roadmap, rebranding its AI3-AI6 chip line as SI4, SI5 and SI6, substituting 'Super Intelligence' for 'Artificial Intelligence' in its naming. The company says its upcoming 6th-generation inference chip will deliver roughly 80 times the compute performance of its current 4th-generation hardware, while its Dojo D3 training chip will move from an SRAM-centric design to a multi-chiplet architecture with stacked HBM memory, replacing the discontinued wafer-scale approach.
tomshardware.com
· 2026-11-06
At GTC Berlin on October 21, Nvidia announced a series of multi-billion-dollar agreements with substrate, packaging, testing and optics suppliers including Ibiden, Unimicron, Samsung Electro-Mechanics, SPIL and Amkor. The deals run through 2028 and target steep capacity increases, such as 2.5x dedicated substrate capacity within three years, a 15x expansion in optical engine output by mid-2027, and a 300% rise in interposer wafer capacity by end-2028.
tomshardware.com
· 2026-10-27
Microsoft announced its third-generation Maia AI accelerator, the Maia 300, saying it will increasingly handle AI processing requests with its own silicon rather than Nvidia GPUs. The chip enters production in 2027, and Executive Vice President Scott Guthrie said Microsoft is targeting multi-gigawatt deployments of the Maia chip family across its Azure data centers.
cnbc.com
· 2026-10-12
A project called openTPU has released an open-source AI inference accelerator whose hardware design, instruction set, compiler and simulator were developed by AI agents rather than human engineers. The design was tested on an Inspur YPCB-00338 FPGA card, successfully running ten language models including LFM2.5, Qwen3, Gemma 4 and SmolLM3, with output matching the simulator bit-for-bit.
github.com
· 2026-10-06
At Apsara Conference 2026, Alibaba CEO Eddie Wu introduced the Zhenwu V900 accelerator, claiming it delivers three times the performance of the prior M890 chip and calling it China's most powerful AI chip. The V900 offers 216GB of GPU memory, 1,200GB/s inter-chip bandwidth, can scale to clusters of up to 500,000 cards, and is set for mass production in Q1 2027, earlier than the Q3 2027 date Alibaba previously targeted.
tomshardware.com
· 2026-09-23
Leaker Kepler_L2 says most of Nvidia's next-gen RTX 60 series and AMD's RDNA 5-based Radeon 10000 lineup won't launch until 2028, with AMD's midrange AT2 chip as the sole exception still on track for 2026. The claim contradicts an earlier report from Moore's Law is Dead suggesting Nvidia's RTX 6090 could arrive in early 2027, and AMD board partners have reportedly given inconsistent timelines ranging from late 2027 to early 2028.
techspot.com
· 2026-09-16
The piece explains how Google's Tensor Processing Units (TPUs), built for AI workloads in data centers, differ from GPUs and NPUs. TPUs are now replacing NPUs in Google's upcoming Pixel 11 phones, with Google claiming up to 3.5 times faster AI processing at up to 3.5 times less energy use. GPUs, meanwhile, remain versatile chips handling gaming, 3D rendering, AI training and crypto mining.
engadget.com
· 2026-08-29