Google is intensifying its effort to expand adoption of its in-house Tensor Processing Units (TPUs), taking direct aim at Nvidia's dominance in AI infrastructure by courting "neocloud" providers that have traditionally built their businesses around Nvidia GPUs.
Reports that Meta is considering leasing out idle AI computing capacity have rattled investors. But treating Meta's predicament as a warning sign for the entire AI industry is a classic case of overgeneralization.
At the Humanoids Summit in Tokyo at the end of May 2026, a conference that had originally focused on technology and commercialization also set aside a stage for government delegates and policy watchers. An executive at a major US robotics company said bluntly at the event: "Government intervention is no longer optional."
Apple and Broadcom have extended their custom-chip partnership through 2031 under agreements expected to exceed US$30 billion, reinforcing Broadcom's position as a critical Apple chip supplier.
Shipments of smartwatches with edge AI features leaped 70% in the first quarter of 2026, with Apple overwhelmingly leading the charge, according to Counterpoint Research. This market is taking off as health-focused users gain access to deeper insights through AI advancements, and as brands lean on wearables to offset weak smartphone sales.
Huawei has joined more than 20 Chinese technology companies and research bodies to launch OPEN NPO, the country's first multi-source agreement for near-packaged optics, in an effort to standardise high-speed optical interconnects for AI supernodes and large-scale computing clusters.
UMA, a Physical AI company, unveiled the design of its first humanoid robot at Machina Summit and introduced Real-Time Learning, an architecture that allows robots to learn new tasks from demonstrations rather than manual programming.
Nvidia has denied reports that its next-generation Kyber AI rack system could be delayed to 2028, saying its product roadmap remains unchanged, after SemiAnalysis pointed to manufacturing challenges in a key PCB midplane as a potential bottleneck for the Rubin Ultra platform.
