OpenAI has entered a multi-year agreement to purchase Google TPU chips to expand its AI compute infrastructure amid rising demand. This move diversifies OpenAI’s hardware supply beyond GPUs, aiming to...
d-Matrix’s adoption of Nvidia’s NVLink Fusion rackscale system marks a significant shift in AI infrastructure design for 2026. This technology promises ultra-low latency communication between GPUs and...
DayOne and Sembcorp have signed a memorandum of understanding to develop integrated power and energy solutions for hyperscale data centers across the Asia-Pacific region. The partnership aims to pilot...
As AI workloads demand increasingly dense and efficient power delivery, data centers are shifting to 800VDC power architectures. This analysis explores how 800VDC reduces conversion losses, enables hi...
The AI industry's push against the 1-megawatt rack power ceiling represents a strategic inflection point requiring systemic innovation in cooling, power delivery, and architectural design to sustain s...
FuriosaAI has raised $620 million to accelerate the development of energy-efficient AI inference chips aimed at reducing data center power consumption. The funding will support manufacturing scale-up ...
FuriosaAI raised $620 million in March 2026 to scale production of energy-efficient AI inference chips designed to reduce data center power consumption. The funding reflects growing investor interest ...
Oracle announced a partnership with Google to integrate Google's TPU v4 hardware into Oracle Cloud Infrastructure, expanding AI compute options for enterprise customers. This integration aims to accel...
In early September 2026, we spotted three key trends in AI infrastructure energy and operations: self-improving AI for power optimization, MARA Holdings' massive power expansions, and hyperscalers' pu...
Red Hat has launched the OpenClaw sandbox, a containerized environment designed to secure and isolate agentic AI workloads during development. The sandbox aims to mitigate risks associated with autono...