AI infrastructure funding shifts toward agentic, model-agnostic enterprise tooling
Enterprise AI infrastructure funding describes investment into the platforms, databases, and operations tools that allow companies to run large-scale, model-agnostic, production AI workloads—especially agentic systems that generate code, resolve incidents, and query live data in real time—without locking into any single model vendor or consumer-facing application layer. Together, Niteshift, DeductiveAI, and PhoenixAI have attracted USD 172 million (approx. RM792 million) across seed, acquisition, and Series B rounds, highlighting where investors now see durable value: the plumbing behind AI, not the interfaces on top. These deals span AI coding infrastructure, AI-native site reliability engineering, and agentic AI database technology, forming a picture of enterprises standardizing on model-agnostic platforms while demanding vertical-specific capabilities for code, ops, and data. The question is no longer whether enterprises will use AI, but which infrastructure stack will safely power their agents at scale.
Niteshift’s $7M seed backs model-agnostic AI coding infrastructure
Niteshift is building AI coding infrastructure that separates coding agents from the models they run on, aiming to become a model-agnostic routing layer for enterprise codebases. Founded by former early Datadog engineers Sajid Mehmood and Conor Branagan, the company raised a USD 7 million (approx. RM32 million) seed led by Greylock partner Jerry Chen, with angels including Reid Hoffman and Datadog co-founders Olivier Pomel and Alexis Lê-Quôc. Niteshift argues that enterprises will hesitate to send their most sensitive code through AI model makers that are also pushing into vertical software markets. Instead of selling token-based usage, it charges per-minute infrastructure fees, positioning itself like a cloud provider for AI coding agents. The platform orchestrates between frontier models, open-source alternatives, and other providers based on project needs, entering a competitive field that includes Cursor, Cognition, Amazon Bedrock, and OpenRouter.
Elastic’s DeductiveAI acquisition signals AI-native ops as consolidation target
Elastic’s move to acquire DeductiveAI for up to USD 85 million (approx. RM392 million) shows that enterprise tooling acquisitions now prioritize AI-native site reliability engineering. DeductiveAI compresses the path from a USD 7.5 million (approx. RM35 million) seed to an exit in under a year, offering a suite of AI agents that connect directly to code, logs, metrics, traces, and events. Those agents reason over a continuously updated knowledge graph, test hypotheses in real time, and surface root causes in seconds. DeductiveAI claimed up to a 90% reduction in incident resolution time, and early deployments at DoorDash and Foursquare supported that figure. Elastic is folding this capability into its observability platform alongside earlier deals for Keep and Jina AI, plus a new agentic Kubernetes investigation workflow. The acquisition raises pressure on rivals like Datadog, Dynatrace, and Splunk/Cisco, all of which have invested in AI-assisted observability but have not bought a dedicated AI SRE startup with a knowledge graph engine.
PhoenixAI’s $80M Series B targets the agentic AI database layer
PhoenixAI, rebranded from CelerData, raised a USD 80 million (approx. RM369 million) Series B led by Sky9 Capital to power what it calls the agentic AI database. The platform is purpose-built for agentic AI workloads that fire off thousands of unpredictable real-time queries, unifying live and historical data in a single AI-native engine that delivers sub-second responses at scale. Market-leading companies including AppLovin, Coinbase, Conductor, and Demandbase run PhoenixAI in production, reporting sub-second query responses across hundreds of millions of rows and tight integration with Apache Iceberg data lakehouses and Kafka streaming pipelines. According to PhoenixAI President Rick Underwood, “Agents now fire off thousands of unplanned, real-time queries… demanding analysis across live data, historical records, and multiple systems at once, which strains existing data stacks.” By making streaming updates queryable within seconds and serving agents on the same real-time dataset as analysts, PhoenixAI aims to become the default engine for agentic AI database workloads.
Model-agnostic vs vertical-specific: competing strategies in enterprise AI infrastructure
These three deals display two main strategies in AI infrastructure funding: model-agnostic platforms and vertical-specific tooling. Niteshift is firmly in the model-agnostic camp, routing across frontier and open-source models while charging like infrastructure, not SaaS labor replacement. PhoenixAI sits closer to the platform side as well, offering an agentic AI database that any agent or model can query in real time, independent of the application layer. DeductiveAI, by contrast, is highly vertical-specific, focused on AI-native site reliability engineering inside observability platforms. Elastic’s acquisition shows how vertical tools can become acquisition currency once they reach “must-have” status for incumbents. Across coding, ops, and data, the common thread is agentic workloads: enterprises want systems that can act, not just display dashboards. Investors appear to be backing the layers where agentic AI intersects with production systems, betting that whoever standardizes those layers will win long-term enterprise tooling contracts.
