Nscale Acquires Anyscale to Test Cloud-Neutral AI at Scale
Nscale has secured a definitive agreement to buy Anyscale, raising questions about whether AI software platforms can maintain vendor independence when integrated with specialized GPU infrastructure.
A tale of scale
This week, infrastructure provider Nscale finalized a definitive agreement to purchase Anyscale, a specialist in scaling AI workloads. The acquisition puts to the test a fundamental question: can software built to work across multiple cloud providers remain truly neutral once it becomes part of a GPU-focused neocloud offering?
Adrian Bridgwater, a contributor to The New Stack, sat down with Dan Bathurst, Chief Product Officer at Nscale, to discuss the transaction and what appears to be a vertically integrated approach to AI cloud services.
According to Bathurst, Anyscale will maintain its independent identity. He explained: "What really changes — or how it's changing — is that customers now also get this first-party option, where they can have Anyscale running on Nscale fleet as a full-stack, highly-optimized solution."
Bathurst emphasized that competitive advantage will come through technical performance rather than commercial pressure. He stated: "Where we want to win is on performance, not on any sort of vendor lock-in or forcing of someone to choose Nscale as the infrastructure provider."
TOP OF THE STACK
Gemini Robotics 2 brings us one step closer to physical AGI
Google DeepMind unveiled Gemini Robotics 2 this week, an intelligence layer built from three new models designed to enable more capable physical AI systems. The models work together to equip robots with enhanced dexterity and full-body control, allowing them to collaborate and tackle complex, multi-step operations. Google says the vision-language-action approach represents progress toward more adaptable robotic systems.
WHAT ELSE IS NEW?
Mate has launched an agentic Security Operations Platform that uses full business context to investigate every security alert, enabling teams to operate at machine speed. The platform automatically resolves up to 85% of false positives, reduces investigation time from hours to minutes, and operates on a live Security Context Graph of organizational infrastructure. The company has secured $35M in new funding to expand the platform.
Roadmap.sh, long recognized for mapping career progression, now offers premium features that go beyond visualization. Roadmap Pro includes a personal AI coach and the ability to generate tailored courses on demand, providing a structured alternative to sifting through scattered documentation for those seeking to learn new technology stacks.
Subquadratic has developed a model claiming to achieve 1,000x reduction in compute requirements for long-context processing. The company's CTO, Alex Whedon, suggests the real breakthrough lies elsewhere. He discussed with The New Stack a "zero attention" architecture he believes will eventually replace current approaches and may eventually supersede even Subquadratic's own flagship model.
Webflow has demonstrated how AI integration into security operations can help engineering teams scale and automate alert triage without building a traditional Security Operations Center.