AIThority Interview With Rohit Agarwal, Founder & CEO of Portkey<\/a><\/p>\nOne platform to drive cost efficiency for the full AI lifecycle
<\/p>\n
As enterprises move from AI experimentation to production, many are discovering that AI cost is not driven only by model usage. It is also driven by the fragmented systems required to prepare data, train and fine-tune models, evaluate outputs, serve inference, and generate or process datasets with those models. To scale from development to production, AI and platform teams often start by stitching together CPU-oriented, cloud-native data processing platforms with hosted-model APIs for GPU-intensive workloads, creating brittle architectures that increase operational overhead, obscure cost visibility, and make production AI harder to govern.<\/p>\n
Anyscale closes this gap with its runtime engine, purpose-built to handle both CPU and GPU workloads at scale. Built on Ray, the open-source compute framework that powers the processing demands of leading AI companies, including Cursor, Physical Intelligence, and xAI, Anyscale enables teams to run distributed multimodal data processing, training, fine-tuning, reinforcement learning, inference, and agentic workloads on one platform inside their Azure environment.<\/p>\n
By reducing platform fragmentation, Anyscale helps teams improve GPU utilization, eliminate unnecessary data movement, streamline infrastructure operations, and replace unpredictable per-token\u00a0API economics with compute they can govern directly. Customers report up to 4 times faster experimentation and up to 90% lower AI total cost of ownership versus fragmented stacks that combine cloud-native data processing engines with hosted model APIs.<\/p>\n
An Azure Native Integration for enterprise AI security and sovereignty<\/p>\n
Anyscale on Azure is delivered as an Azure Native Integration, the result of deep engineering collaboration between Anyscale and Microsoft. The platform runs entirely inside the customer\u2019s own Azure tenancy on Azure Kubernetes Service (AKS), so proprietary data, training pipelines, and model weights never leave the boundary of the customer\u2019s account.<\/p>\n
The solution is provisioned through Azure Resource Manager (ARM) and inherits Azure\u2019s native security and identity model. Every Anyscale resource is created, tagged, and governed like any other first-party Azure resource. Enterprises apply the same Microsoft Entra ID policies, role-based access controls, resource policies, , and audit controls they already use across the rest of their Azure account. For organizations in regulated industries, including financial services, healthcare, public sector, and others with strong security postures, this also means data residency, and expanding existing governance that carries forward to AI workloads.<\/p>\n
Sovereignty here isn\u2019t a label bolted on after the fact. It is the architectural starting point: customer-owned data and customer-owned models in the customer-owned tenancy and governance estate. All deployed and managed from the Azure Portal like any other Azure service.<\/p>\n
Anyscale and Microsoft, a collaboration for the next wave of enterprise AI<\/p>\n
This wave of enterprise AI will be defined by organizations operating their own AI systems, with their own models, on their own data, on infrastructure they govern. As AI workloads become increasingly distributed, multimodal, and GPU-intensive, enterprises are looking for a unified compute layer that can support the full AI lifecycle cost-efficiently and securely.<\/p>\n
Anyscale on Azure is designed for this new era of control of enterprise AI: open source at the core, combining the proven scale of Ray and Kubernetes, and deployed where the data already lives.<\/p>\n
And because the solution is delivered natively through Azure, Anyscale consumption draws down against existing Microsoft Azure Consumption Commitments (MACC), meaning enterprises can start building sovereign AI today against budgets they have already committed, with no new procurement cycle to navigate.<\/p>\n","protected":false},"excerpt":{"rendered":"With the Anyscale on Azure public preview, enterprises can now run foundation-model-scale AI workloads, from multimodal data preparation…\n","protected":false},"author":2,"featured_media":61867,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[34548,420,7829,35401,320,7828,35402,1960],"class_list":["post-61866","post","type-post","status-publish","format-standard","has-post-thumbnail","category-microsoft","tag-anyscale","tag-azure","tag-azure-ai","tag-azure-cloud-native","tag-microsoft","tag-microsoft-ai","tag-microsoft-azure-consumption-commitments","tag-sovereign-ai"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/61866","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=61866"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/61866\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/61867"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=61866"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=61866"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=61866"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"Documentation](\"https:\/\/docs.anyscale.com\/img\/logo.svg\"){kind=logo}