Nscale, a London-based AI cloud provider that owns its own power generation and data centers, announced on July 30, 2026 that it will acquire Anyscale — the startup behind the widely used open-source Ray distributed computing framework — in a deal Bloomberg reports is worth approximately $1.65 billion. All roughly 200 Anyscale employees are expected to join Nscale, and the company will continue operating under its own brand rather than being folded into Nscale’s identity.
Anyscale’s story starts with Ray itself: an open-source Python framework built by the same founding team to let developers scale compute-heavy workloads across clusters of machines without rewriting their code for distributed systems. When GPT-3 landed in 2022 and large language models became the center of gravity for enterprise software spending, Anyscale pivoted its commercial platform toward exactly the problem every AI team eventually hits — training, fine-tuning, and serving models across fleets of GPUs, plus data curation, inference, and reinforcement learning at scale. The company says its platform can cut total cost of ownership on AI workloads by as much as 90%, and it reported 70% sequential revenue growth in its most recent quarter, evidence that the pitch is landing with real enterprise buyers, not just open-source enthusiasts. Governance of the Ray project itself transferred to the PyTorch Foundation under the Linux Foundation in October 2025, a move that kept the open-source core neutral even as Anyscale built a commercial business on top of it.
Nscale itself is a relatively young company with an unusual origin story for a GPU infrastructure provider. Founded in 2023 by Josh Payne and Nathan Townsend as a spinout of Arkon Energy, a renewably-powered cryptocurrency-mining data center business, Nscale pivoted that same power-and-cooling expertise toward AI compute once the GPU shortage made owning your own energy supply a genuine competitive advantage rather than a cost center. It has raised aggressively to fund that bet: a $30 million seed round in December 2023, a $155 million Series A a year later, then a $1.1 billion Series B in September 2025 led by Aker ASA — at the time the largest Series B round in European history — followed by a further $2 billion raised in March 2026 at a reported $14.6 billion valuation, alongside a separate $1.4 billion delayed-draw term loan and a $790 million project-financing package for a hydro-powered data center campus in Narvik, Norway. That capital base, and the fact that Nscale already owns real renewable power generation rather than just leasing capacity, is what makes a $1.65 billion cash-and-stock acquisition of a software company plausible for a three-year-old startup in the first place.
Why a GPU Cloud Company Wants an Orchestration Startup
The deal is a clean example of a trend reshaping the “neocloud” category — the wave of AI-focused infrastructure providers, Nscale among them, that emerged to rent out GPU capacity as an alternative to the big three hyperscalers. Raw GPU rental is a commodity business with thin, shrinking margins once enough providers are competing for the same enterprise contracts. The software layer that sits on top of the hardware — the orchestration, scheduling, and workload-management tools that decide how efficiently those GPUs actually get used — is where the real defensibility and pricing power live. By buying Anyscale outright rather than just integrating with Ray as a third-party dependency, Nscale is betting it can offer customers a single, vertically integrated stack from power generation through to AI workload orchestration, rather than competing purely on GPU price per hour against every other neocloud and hyperscaler in the market.
That’s a meaningfully different strategy from simply buying more GPUs. It’s an attempt to own the layer where a customer actually experiences whether their AI infrastructure spend is working — and it puts Nscale in more direct competition with the orchestration and MLOps tooling built natively into AWS, Google Cloud, and Microsoft Azure, not just with other GPU-rental shops.
What This Means for Philippine Founders
Most Philippine AI startups don’t buy GPU capacity directly from a neocloud like Nscale — they’re on AWS, Google Cloud, or a reseller, often through startup credit programs. But this deal is still worth watching for two concrete reasons. First, it signals where AI infrastructure pricing pressure is actually headed: as neoclouds like Nscale build genuine software moats instead of competing purely on raw compute price, the “race to the bottom” on GPU-hour pricing that has made experimentation cheap for early-stage founders may not continue indefinitely in its current form — vertically integrated providers have less incentive to compete purely on price once they can differentiate on efficiency and orchestration instead. Second, Ray itself is increasingly the default tool Philippine engineering teams reach for when they outgrow a single-GPU prototype and need to actually scale training or inference — it’s open source, well-documented, and now backed by a company with real enterprise revenue and $1.65 billion of fresh institutional confidence behind it. A founder deciding what to build a scaling layer on top of now has one more concrete signal that Ray isn’t a bet on an obscure framework disappearing next year — it’s infrastructure a real acquirer just paid real money to own outright.
Share this article