AI hardware

Nvidia is pushing a watchdog chip designed to sit alongside every AI agent, supervising its actions and intervening when behaviour goes wrong. The announcement, reported by CNBC, signals Nvidia's effort to make autonomous AI systems safer and more accountable as they spread through enterprise software. Discussion is focused on whether dedicated oversight hardware is a genuine safety breakthrough or a way for Nvidia to sell more chips.

World Labs, the AI startup founded by Fei-Fei Li focused on spatial intelligence and world models, is joining AMD. The announcement, made via the company's own blog, means the team and its technology will become part of AMD's AI efforts. The news is drawing attention on technology forums, with readers weighing what the acquisition means for the startup's ambitious research agenda and for AMD's position in the AI hardware race.

TSMC is shaping the economics of the entire AI chip industry, as the dominant contract manufacturer for leading chip designers. A report argues the Taiwan-based foundry effectively dictates what AI chip companies must pay for manufacturing capacity. With demand for AI accelerators surging, TSMC's pricing power and advanced process nodes give it leverage over customers like Nvidia, AMD and countless startups racing to ship AI hardware.

Nvidia is reportedly exploring glass substrates for advanced chip packaging, a shift that could affect the next generation of AI processors. Glass packaging offers better flatness and thermal stability than organic materials, allowing denser chip connections as AI chips grow larger and power-hungrier. Analysts see the move as part of the race to sustain performance gains beyond traditional Moore's Law scaling, with rivals such as Intel and Samsung also researching glass substrates.

VSMC is building a new 300mm wafer fabrication facility in Singapore, positioned as a strategic bet on the mature and specialty chips that underpin AI hardware even as attention stays fixed on leading-edge processors. Industry commentary frames the fab as addressing supply of foundational semiconductors that AI systems cannot function without.

Nvidia is working on a watchdog system built directly into its chips to keep AI agents under control and operating within set limits. The approach would monitor agent behaviour at the hardware level rather than relying solely on software safeguards. Details beyond the announcement remain limited, and it is not yet clear when the technology will appear in products.

Nvidia announced the Open Agent Safety Platform on Monday, combining open source software with a reference system design that uses a hardware-based watchdog to keep AI agents operating within set boundaries. The SecurityWeek report has drawn attention from security and AI communities, with discussion focused on how hardware-level safeguards could address growing concerns about autonomous AI agents acting unpredictably.

Nvidia has announced a new software tool intended to stop rogue AI systems. Details on how the tool would actually work remain limited, but it reflects Nvidia's push into AI safety as concerns grow about powerful models acting beyond the intent of their developers. Observers are watching closely given Nvidia's central role in the AI boom.

Nvidia has added LG Electronics to its Partner Network as a preferred cooling provider, aimed at easing the power and heat bottleneck facing its AI data centers. The deal gives LG a direct route to supply thermal management technology, and commentators note the real competition in AI hardware may be shifting from GPUs to cooling.

AI chipmaker Cerebras has agreed to supply cloud startup Gimlet Labs with chips capable of drawing 100 megawatts of power, a major deal in the rapidly scaling market for AI computing capacity. The agreement underscores how demand for AI inference and training hardware is reshaping the semiconductor and cloud industries, with startups securing large-scale chip supplies to compete with established hyperscalers.

China's artificial intelligence industry is entering a new phase, DigiTimes Asia reports, shifting from trying to catch up with US rivals to monetizing AI at scale. The report links the trend to domestic semiconductor efforts, including CXMT's DRAM production and advanced hardware packaging, as Chinese firms build homegrown supply chains to support commercial AI deployment and reduce reliance on foreign chip technology.

Nvidia has introduced its Open Agent Safety Platform, a reference framework for continuously monitoring AI agents directly in silicon. Announced on the company's developer blog, the platform aims to provide hardware-level safety checks as AI agents are increasingly deployed in production environments. Developer interest is growing, with discussion appearing on technology forums.

Alibaba is reportedly close to receiving Chinese approval to purchase Nvidia's RTX Pro 5500 chips for use in the country. The move would let the Chinese tech giant access advanced US-made graphics hardware despite ongoing export restrictions, and comes as Washington and Beijing continue to negotiate the terms of semiconductor trade. Investors are watching closely for confirmation, given the chip's relevance to AI and data-center workloads.

South Korean robotics firm Angel Robotics has launched phai-x1, a wearable robot platform designed for physical AI research. The company is positioning the system as a tool for researchers developing AI that interacts with human movement, an emerging focus in robotics. Details on specifications, pricing and availability have not yet been widely reported.

A new publication examines the economics of running open-weight AI models for inference, arguing that self-hosting open models can compete on cost with proprietary APIs under certain workloads. The piece analyzes hardware, utilization, and pricing trade-offs, and is drawing attention among developers and infrastructure engineers weighing whether to run models themselves or pay commercial providers.

NVIDIA has introduced an open Agent Safety Platform, described in its technical blog as a reference design for continuous in-silicon monitoring of AI agents. The platform aims to provide developers with tools to observe and safeguard autonomous agent behavior directly on hardware, as agentic AI systems move into production use.

Investors are turning their attention to three AI infrastructure stocks following AMD's investment deal with World Labs, the AI startup founded by Fei-Fei Li. The partnership has renewed focus on companies supplying the computing hardware and data center capacity behind advanced AI development, with retail traders tracking which firms stand to benefit from the deal.

Meta is reportedly developing the Muse Charm, a small dedicated AI device compared to a Tamagotchi for your AI assistant. The reveal has sparked debate over whether purpose-built AI gadgets make sense when phones already handle assistants, and whether a charm-like companion device could appeal to everyday users or remain a niche experiment.

Reports say Chinese authorities have approved Nvidia's RTX PRO 5500 chip for use by companies including Alibaba and ByteDance, despite ongoing US export restrictions on advanced semiconductors. The move would let major Chinese tech firms access Nvidia hardware at a time when Washington continues to limit what AI chips can legally reach the Chinese market.

Nvidia has assembled a group of partners including AI lab Anthropic and chip designer Arm for a new AI security platform. The coalition suggests Nvidia is extending beyond hardware into security tooling as AI deployment accelerates. Details about the platform's scope and the exact roles of each partner remain limited.