Datadog acquires Adaptive ML to boost agentic AI research

Datadog has acquired Adaptive ML, a frontier AI startup focused on Reinforcement Learning Operations (RLOps), in a move the NASDAQ-listed observability vendor says will accelerate its in-house research into agentic large language models and world models. Financial terms were not disclosed.

Adaptive ML, which positioned its platform as a way for enterprises to build, own and deploy their own specialised agents and models, will be folded into Datadog AI Research, the company's internal lab. The acquisition is the latest signal that established observability platforms are moving beyond monitoring and logging into AI-native operations, where agents detect, diagnose and remediate infrastructure issues autonomously.

The deal

Julien Launay, co-founder and chief executive of Adaptive ML, said the pairing gives his team the production-scale signal its research had previously lacked. "The hardest part was production scale," Launay said. "With Datadog's unmatched access to real-world infrastructure, we can accelerate towards continuous intelligence." The company will work directly within Datadog AI Research rather than operating as a standalone unit.

Datadog's chief scientist Ameet Talwalkar framed the acquisition as a complement to existing lab work, citing the goal of converting Datadog's telemetry streams into what he called "first-party intelligence." The company says it spends more than $1 billion annually on R&D, and has already shipped agentic products under its Bits product line, including Bits Investigation, Bits Code and Bits Security Analyst, which it says have collectively handled hundreds of thousands of customer investigations. Its Toto 2.0 time-series forecasting research was also cited as evidence of the lab's output to date.

Market context

The acquisition fits a broader pattern of consolidation in the AI operations space. Observability vendors including Datadog, Dynatrace and New Relic have each been building or buying capabilities that layer AI reasoning on top of traditional metrics, logs and traces pipelines. The RLOps category that Adaptive ML was carving out sits at the intersection of MLOps tooling and reinforcement learning infrastructure, an area that has attracted a number of well-funded startups alongside offerings from hyperscalers.

For Datadog specifically, integrating RLOps capabilities could sharpen its ability to train and fine-tune specialised agents against its own high-volume telemetry, a genuine competitive advantage given the scale of data it processes from its global customer base of Fortune 500 companies and high-growth AI firms. Whether that advantage translates into durable product differentiation will depend on how quickly competitors can replicate similar feedback loops from their own data estates.

Regulatory and standards read-across

Acquisitions of AI research teams have drawn increasing scrutiny from competition authorities on both sides of the Atlantic. The UK CMA and the European Commission have both signalled interest in so-called acqui-hires and talent-led deals in the AI sector, examining whether they constitute de facto mergers that circumvent standard notification thresholds. Datadog has not indicated whether the Adaptive ML deal was notified to any regulator, though its scale and the private startup status of the target make it unlikely to trigger mandatory filing in most jurisdictions.

Separately, the EU AI Act's provisions on general-purpose AI models and high-risk deployment will be relevant as Datadog's Bits agentic products mature; systems that autonomously act on infrastructure at scale may attract closer scrutiny as enforcement guidance firms up through 2027.