AI Infrastructure Funding Becomes the New Enterprise Battleground
AI infrastructure funding refers to venture and strategic investment in the platforms, databases, and operational tools that help companies deploy, manage, and evaluate AI systems across different models and workloads, instead of building every AI workflow from scratch on a single provider’s stack. Over the past few weeks, that layer has become one of the hottest segments in enterprise software. Startups focused on developer tooling investments, agentic AI database engines, AI ops tooling, and safety platforms for agents have collectively drawn more than USD 200 million (approx. RM920 million) in fresh capital and one major acquisition. The pattern is clear: buyers and investors want model-agnostic platforms that sit between frontier labs and production applications. These companies are not competing on yet another foundation model; they offer the infrastructure that lets enterprises use Anthropic, OpenAI, open source, and specialised agents without handing over control of code, data, or operations.
Niteshift Bets on Model-Agnostic Coding Infrastructure
Niteshift has raised USD 7 million (approx. RM32.2 million) in seed funding to build model-agnostic infrastructure for AI coding agents. Founded by former Datadog engineers Sajid Mehmood and Conor Branagan, the company argues that enterprises will avoid routing sensitive code through frontier labs that now compete in vertical markets. Instead of replacing tools like Claude Code, Niteshift is a routing layer that orchestrates between frontier and open-source models based on project needs. The startup charges per-minute infrastructure fees, positioning itself as a cloud-style service for AI coding workloads rather than a labour replacement tool. Backers such as Greylock’s Jerry Chen and angels including Reid Hoffman and Datadog’s co-founders show confidence that unbundling code agents from their underlying infrastructure will be a durable opportunity. In a field crowded by Cursor, Cognition, Amazon Bedrock, and OpenRouter, Niteshift’s pitch rests on its experience scaling observability infrastructure for large engineering teams.
Elastic’s DeductiveAI Deal Signals AI-Native Ops Tooling Is Acquisition Currency
Elastic’s agreement to acquire AI SRE startup DeductiveAI for up to USD 85 million (approx. RM391 million) compresses the path from seed funding to exit into months and shows how AI-native ops tooling has become strategic. DeductiveAI raised USD 7.5 million (approx. RM34.5 million) in seed capital in late 2025, then focused on AI agents that connect directly to code, logs, metrics, traces, and events. Those agents reason over a live knowledge graph to test hypotheses and surface root causes in seconds, with reported incident resolution time cuts of up to 90 percent in early deployments. According to VentureBeat reporting cited in the source, DoorDash alone saved more than 1,000 engineering hours annually by using Deductive’s automation. Elastic will fold this capability into its observability suite, alongside earlier acquisitions such as Keep and Jina AI, to offer autonomous resolution instead of only smarter alerting—putting pressure on Datadog, Dynatrace, and Splunk.
Agentic AI Databases and Spatial Training Pull in Massive Rounds
Beyond ops, new infrastructure is forming around agents themselves. PhoenixAI, an agentic AI database previously known as CelerData, raised USD 80 million (approx. RM368 million) in Series B funding to scale an AI-native engine built specifically for agent workloads. Its database unifies live and historical data to answer thousands of unpredictable real-time queries with sub-second responses, and it already runs in production at companies such as AppLovin, Coinbase, Conductor, and Demandbase. In parallel, General Intuition is in talks to raise about USD 300 million (approx. RM1.38 billion) at a valuation just over USD 2 billion (approx. RM9.2 billion) to advance spatial AI agent training. Spun out of Medal, it trains models on two billion gameplay videos each year from ten million monthly users, using first-person clips to teach agents spatio-temporal reasoning. Unlike other world-model players, General Intuition builds models specifically to train agents, not to sell model access directly.
Coval and the New Reliability Layer for Voice and Chat Agents
A parallel wave of infrastructure targets reliability for voice and chat agents. Coval raised USD 28 million (approx. RM129 million) in Series A funding, bringing total capital to USD 31 million (approx. RM142.6 million), to expand what it describes as an evaluation platform for voice AI. The company provides simulation, observability, labelling, and monitoring for autonomous voice and chat agents, and counts Zoom, Deepgram, and other large enterprises as customers. As more organisations adopt voice AI in customer service, finance, and healthcare, Coval aims to replace fragile manual QA with tens of millions of automated evaluations. Founder and CEO Brooke Hopkins, who previously worked on evaluation systems for autonomous driving, argues that every company will have a voice agent and needs infrastructure to deploy it with confidence. With more than USD 7 billion (approx. RM32.2 billion) invested in voice AI in the first quarter of 2026, safety and reliability platforms are emerging as a critical layer.
