Nscale's Anyscale acquisition raises fresh questions about cloud-neutral AI software independence

GPU-focused cloud provider Nscale has agreed to acquire Anyscale, an independent software orchestration platform for AI workloads. The deal merges infrastructure capabilities—Nscale's GPU datacenters and control systems—with Anyscale's technology for scaling machine learning tasks across cloud providers. While Nscale frames it as a complementary fit, the acquisition raises fundamental questions about whether truly cloud-agnostic software can remain neutral once owned by a single infrastructure vendor.

What the merger brings together#

Nscale operates its own bare-metal GPU datacenters optimized for AI and machine learning. Anyscale, built on the open-source Ray project, provides a multi-cloud control plane that helps teams orchestrate AI workloads—from data processing and model training to inference—across AWS, Google Cloud, Azure, or any other infrastructure. Combining them creates a "full-stack AI hyperscaler," according to Anyscale CEO Keerti Melkote, where software and compute are designed and engineered in lockstep.

The neutrality pledge and the skepticism it faces#

Nscale's chief product officer, Dan Bathurst, has committed to maintaining Anyscale as a independent product that continues supporting bring-your-own-cloud deployments. He characterizes the strategy as "neutrality on the platform layer, but differentiation on the infrastructure layer"—meaning customers remain free to run Anyscale elsewhere, but will have a first-party, fully optimized option on Nscale's own hardware.

However, analyst Sanjeev Mohan from SanjMo warns that neutrality becomes hollow once best features and optimal pricing arrive on Nscale first. "The software will still run anywhere, but 'runs anywhere' and 'runs best somewhere' are different things, and buyers will feel the gap in performance and cost. At that point, neutrality is a label," Mohan tells The New Stack. He notes that controlling both silicon and software orchestration allows Nscale to tune scheduling, memory, and networking in ways a cloud-agnostic Anyscale never could—a genuine technical advantage that creates subtle lock-in.

Precedent and broader market dynamics#

Mohan argues this move reflects a broader industry pattern: independent tooling vendors are being absorbed by whoever owns the compute, because the economics of vertical integration only work when both layers are controlled. "Every optimization that only shows up on Nscale hardware is a dependency," he explains. He expects more such consolidation and sees Nscale emerging as a credible specialist cloud provider alternative, particularly strong in training and inference at scale—though not a general-purpose cloud like AWS or Azure with their sprawling managed service portfolios.