Why DigitalOcean Needs Katanemo Labs More Than GPUs Right Now

DigitalOcean acquired Katanemo Labs this week, bringing in an open source data plane project called Plano and a set of small action models designed for agentic AI workloads. Salman Paracha, co-founder and CEO of Katanemo Labs, joined DigitalOcean as senior vice president of AI. Financial terms were not disclosed and the deal is not expected to have a material impact on DigitalOcean's 2026 results. The acquisition signals something broader than a talent hire. As GPU capacity becomes easier to procure across cloud providers and bare metal vendors, the real bottleneck for production AI has shifted. Teams building agentic systems struggle less with compute and more with the operational layer that sits between models and reliable production deployment. DigitalOcean is betting that owning this middleware will matter more than selling raw GPU cycles. What Plano Brings to the Stack Plano is a framework-agnostic data plane for agentic applications, built on Envoy by its core contributors. It handles orchestration, safety guardrails, observability and model routing as out-of-process infrastructure rather than embedded framework code. Developers can wire up multi-agent systems by declaring agent endpoints and model providers in a configuration file, and Plano manages the routing, tracing and moderation policies without requiring changes to application logic. The project also includes small action models such as Arch-router and Plano-Orchestrator that handle intent classification and workflow execution for multi-agent deployments. These models complement the data plane by providing lightweight routing and orchestration capabilities tailored for real-world agentic workloads. On GitHub, Plano has accumulated roughly 6,100 stars, a meaningful signal of developer interest for a project in this category. The practical appeal is straightforward. Every team building agentic applications ends up rebuilding the same plumbing for agent routing, trace collection and safety filters. Plano centralizes that work into a shared infrastructure layer, which DigitalOcean now plans to integrate into its cloud platform. The Production Gap in Agentic AI DigitalOcean's interest in Katanemo Labs reflects a well-documented problem. McKinsey research shows that fewer than 10% of AI use cases deployed ever make it past the pilot stage. For agentic systems, where multiple models coordinate across tools and APIs, the gap between prototype and production is even wider. Observability, safety enforcement and orchestration reliability are the missing pieces, and few off-the-shelf solutions address all three. DigitalOcean reported $901 million in revenue for fiscal year 2025 with 2026 guidance of $1.075 billion to $1.105 billion. Its AI customer annual recurring revenue reached $120 million, growing 150% year-over-year. More than 70% of that AI revenue comes from inference services and core cloud products rather than bare metal GPU rental. Katanemo Labs also brings proprietary research in agentic observability that uses signal-based technology to identify informative interactions from production traces. The idea is to turn runtime data into insights about agent behavior, helping teams diagnose failures and improve performance after deployment. Competing Without Hyperscaler Scale DigitalOcean competes with AWS, Microsoft Azure and Google Cloud with a fraction of their infrastructure and R&D budgets. Its pitch to developers and mid-market companies centers on operational simplicity and predictable pricing rather than sheer scale. The Katanemo acquisition extends that positioning by adding a software layer that hyperscalers have not yet consolidated. AWS offers Bedrock Agents, Microsoft has Foundry Agent Service and Google provides Vertex AI agent capabilities. But these services tend to lock developers into specific model ecosystems and proprietary orchestration layers. Plano's framework-agnostic design gives DigitalOcean a different angle. A developer using Plano can route between OpenAI, Anthropic and open source models without rewriting application code. DigitalOcean has also scheduled a conference called Deploy for April 28, featuring leaders from Arcee AI, Character.ai, Workato and vLLM. The event is designed to position DigitalOcean as a gathering point for inference-focused developers, an ecosystem play that goes beyond infrastructure sales. The Case for Donating Plano to a Foundation Plano claims to be vendor-neutral and framework-agnostic in design, but it now sits inside a single cloud provider. For CXOs evaluating agentic infrastructure dependencies, this creates a governance question worth examining. The precedent is instructive. Google donated Kubernetes to the Cloud Native Computing Foundation in 2015, and the project became the industry standard precisely because no single vendor controlled its roadmap. Envoy, the proxy layer on which Plano itself is built, followed the same path from Lyft to CNCF. Both projects gained adoption faster under neutral governance than they would have inside their originating companies. The recent dispute between Synadia and CNCF over the NATS messaging project illustrates the opposite scenario. Synadia attempted to withdraw NATS from the foundation and relicense it under a proprietary model, triggering a public confrontation that damaged community trust. That episode reinforced why developers and enterprises care about where open source governance sits. If DigitalOcean wants Plano to become the de facto data plane for agentic infrastructure across the industry, donating it to CNCF or a similar foundation would accelerate multi-vendor adoption and signal long-term commitment to open governance. Keeping it in-house risks a different outcome. Hyperscalers could build competing data plane projects, fragmenting the ecosystem before it has a chance to consolidate. A foundation-hosted Plano would attract contributions from companies that would otherwise treat it as a competitive threat. Limitations and Open Questions The undisclosed financial terms and the company's statement that the deal will not materially affect 2026 results suggest this is a small acquisition by revenue standards. Plano remains an early-stage project with community adoption still unproven at enterprise scale. DigitalOcean also faces margin pressure from ongoing GPU capacity investments, including recent deployments of AMD Instinct MI350X GPU Droplets. Balancing infrastructure spending with software platform development will test the company's ability to execute on multiple fronts. The broader agentic infrastructure market remains fragmented. Competing efforts from hyperscalers come with massive distribution advantages that DigitalOcean cannot match through technology alone. Whether Plano's open source community grows fast enough to establish a defensible position before larger players consolidate the space remains the central strategic risk. The Katanemo acquisition positions DigitalOcean as one of the first cloud providers to own a dedicated operational layer for agentic systems. Whether the company treats Plano as a proprietary moat or contributes it to neutral governance will reveal how it intends to compete. For enterprise technology leaders evaluating inference infrastructure, the governance model of foundational open source dependencies deserves as much scrutiny as the technical capabilities themselves.