AMD's Helion introduces an open rack architecture prioritizing CPU density for agentic AI infrastructure, prompting reevaluation of power, cooling, networking, and economics in next-generation data centers.
Alternative AI Compute
deals
emerging tech
unexpected combo
Major cloud providers and AI developers, including AWS and OpenAI, are rapidly adopting novel hardware configurations to meet intense AI demands. This shift involves integrating specialized accelerators and exploring disaggregated architectures beyond traditional GPU clusters, fundamentally altering data center compute blueprints.
→
AMD has introduced Helion, an open rack design for agentic artificial intelligence workloads, offering CPU and GPU options with planned availability soon.
Apple is enabling resource sharing across multiple macOS devices to boost GPU power for large-scale artificial intelligence tasks.
Duos Technologies Group, Inc. has sold its GPU-as-a-Service business unit to Axe Compute, signifying a strategic shift in its operational focus.
CoreWeave's Vera Rubin NVL72 is available, with Cognition being the first customer, offering the NVIDIA Vera CPU for agentic artificial intelligence workloads.
CoreWeave is expanding its Forge Platform to target enterprise customers, aiming to compete with hyperscalers by evolving into a comprehensive AI cloud provider.
Huawei Cloud has launched a comprehensive suite of enterprise artificial intelligence products, establishing an open agentic cloud framework to enhance AI capabilities for businesses.
Google's Gemini artificial intelligence model can now generate cartoon or lifelike avatars that lip-sync generative speech.
According to the CEO of Percona, an optimal database solution specifically designed for artificial intelligence agents is currently unavailable in the market.
TypeSafe AI has developed a new artificial intelligence model capable of playing the video game Doom.
Home-based GPU networks are emerging as potential supplements to AI data centers, with pilots offering homeowners capacity via wall-mounted appliances.
Tencent aims to monetize its substantial hardware investments by renting out the infrastructure for artificial intelligence workloads, potentially generating immediate profits.
Google's DiLoCo initiative aims to harness idle compute resources across an entire data center network, exploring new paradigms for distributed artificial intelligence training.
Gobi X is developing technology focused on increasing energy availability for artificial intelligence, prioritizing societal benefit over consumption.
Onsemi has acquired Synaptics for $7 billion in an all-stock transaction to accelerate the development of technologies supporting physical artificial intelligence.
An artificial intelligence-powered brain-computer interface has enabled a patient with ALS to work full-time despite being unable to speak.
Amazon Web Services is transitioning its network infrastructure for new non-graphics processing unit deployments to a Resilient Network Graphs architecture, replacing the existing Clos fabrics.
Alibaba has unveiled its new AI chip, the Zhenwu M890, featuring 144GB of GPU memory, designed for both training and inferencing workloads.
Tenstorrent has announced the general availability of its Galaxy Blackhole AI compute platform, featuring RISC-V based systems with 32 Blackhole accelerators in a 6U chassis, priced at $110,000.
DARPA is seeking to accelerate the development of compact, deep-ocean autonomous craft, indicating a shift towards autonomous underwater vehicles beyond aerial drones.
Bitcoin miner Cango has launched "EcoHash," a new offering that will provide high-performance computing and artificial intelligence inference services.
Kelsey Hightower suggests IT professionals can leverage existing automation by rebranding it as "zero-token architecture" to enhance productivity in the era of agentic artificial intelligence.
South Korean startup Rebellions has raised $400 million to scale its artificial intelligence inference capabilities and plans US expansion, reflecting a market shift towards efficiency-focused AI deployments.
While traditional x86 CPUs are common in AI servers, alternative architectures like ARM and RISC-V, along with ASICs, offer potential improvements in energy efficiency and sustainability when paired with appropriate workloads and software optimizations.
Multi-line insurance policies can offer data centers improved administrative efficiency and fewer coverage gaps, although they may also impose aggregate limits and limit customization options.
Elon Musk's ambitious 'Terafab' initiative, intended to address future constraints in chip supply, power availability, and overall scaling for artificial intelligence, has generated discussion within the infrastructure sector.
At GTC 2026, vendors including Dell and HPE introduced innovations that extend beyond hardware to encompass advanced orchestration, storage, and infrastructure solutions designed to support scalable artificial intelligence deployments by addressing critical data challenges.
When a data center unexpectedly becomes a liability, understanding the implications and potential consequences is crucial for stakeholders.
Amazon Web Services is collaborating with chip manufacturer Cerebras on artificial intelligence inference disaggregation technology, which will be implemented within AWS data centers using the Amazon Bedrock platform.
As data centers grow to campus scale to support artificial intelligence infrastructure, operators are adopting comprehensive, layered physical security models enhanced by technologies like artificial intelligence, Internet of Things, and video analytics.
RISC-V proponent SiFive has adopted Nvidia's proprietary NVLink Fusion interconnect technology, a decision that casts doubt on the future viability of competing interconnect standards like UALink.
OpenAI has committed to deploying 750 megawatts worth of Cerebras' large, SRAM-heavy accelerators through 2028, aiming to enhance its ChatGPT inference capabilities and real-time agent performance.