Databricks Reaches $188B Valuation on Multi-AI Infrastructure

Databricks has signed a term sheet for a strategic funding round that values the company at $188 billion. Led by Coatue Management, the estimated $3 billion raise represents a 40% increase from the company’s $134 billion valuation in February 2026. The capital injection follows reported annualized revenue of $5.4 billion, growing over 50% year-over-year with positive free cash flow.

The funding is designated to accelerate a structural transition in enterprise AI workloads. Databricks is aggressively moving its platform away from “tokenmaxxing”—the default strategy of sending all queries to the most powerful frontier models—toward “valuemaxxing,” which routes tasks to the most cost-effective model capable of handling them.

Multi-AI Infrastructure and Tooling

To support heterogeneous model deployment, Databricks is expanding its governance and routing infrastructure. The Unity AI Gateway acts as a control plane for multi-AI environments, managing access, security, and spend limits across proprietary models from OpenAI and Anthropic, as well as self-hosted open-weight models.

The company is also standardizing the databases that back AI agents. In mid-July 2026, Lakebase, a serverless Postgres database purpose-built for agentic workloads, achieved SOC 2 Type 2 compliance. Databricks has also updated its enterprise natural language interface, renaming Genie Spaces to Genie Agents to reflect a shift toward autonomous business data querying.

Open-Weight Benchmarks and Hosted Models

Concurrent with the funding announcement, Databricks published internal benchmarking data validating the economic case for multi-model routing. A July 8 study evaluating coding agents on a multi-million line codebase demonstrated severe cost disparities between proprietary and open-weight models operating at similar quality thresholds.

Model	Quality Score	Cost Per Task
Claude Opus 4.8	Baseline	$1.94
GLM 5.2	Statistically Tied	$1.28

Databricks is expanding its hosted model catalog to capture these workloads. On July 15, the company announced a “day zero” partnership with Thinking Machines Lab to host Inkling. The Inkling architecture is a Mixture-of-Experts (MoE) transformer featuring 975 billion total parameters with 41 billion active parameters per forward pass, utilizing a 1 million token context window optimized for agentic reasoning and code generation.

The platform has also added support for OpenAI’s newest frontier lineup, hosting GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna directly on Databricks infrastructure.

If you are designing enterprise AI pipelines, the expansion of the Unity AI Gateway and the inclusion of models like Inkling require reevaluating your routing logic. Defaulting all prompts to a single provider is no longer the standard architectural pattern. You should profile your specific workloads against both open-weight and proprietary models to establish a baseline cost per successful task.