# DoiT acquired Attribute to trace AI costs beneath the billing layer > Source: > Published: 2026-08-17 07:36:51+00:00 [DoiT](https://www.doit.com/about?ref=runtimewire), a cloud-cost management company, acquired [Attribute](https://attrb.io/?ref=runtimewire), the AI cost-attribution business founded by [Izhak Zimmermann](https://il.linkedin.com/in/izhak?ref=runtimewire) and Liad Tropp, and folded its technology into a product designed to show which customer, feature or agent generated each piece of an AI bill. DoiT [launched the integrated Attribute product](https://www.doit.com/blog/doit-launches-attribute-ai-tokenomics-without-tags-sdks-or-code-changes?ref=runtimewire) on July 7. [CTech publicly reported](https://www.calcalistech.com/ctechnews/article/sycvenedzx?ref=runtimewire) the acquisition on August 17 and estimated the price at approximately $65 million. The exact closing date is not established. DoiT and Attribute have not confirmed the price, and [CRN reported](https://www.crn.com/news/ai/2026/doit-buys-ai-finops-startup-attribute-launches-ai-token-cost-management-product?ref=runtimewire) that the financial terms were undisclosed. Zimmermann and Tropp founded Attribute in 2023 after serving in Unit 81, an Israeli military intelligence technology unit, according to CTech. They built Attribute around a view that the tagging systems used to divide conventional cloud bills cannot keep pace with shared GPUs, LLM gateways and agentic workloads. The founders' answer was to observe consumption closer to where it occurs: inside the operating system. The runtime approach gave the founders a wedge into a crowded FinOps market and gave DoiT a way to extend cloud-cost management into the harder problem of calculating the unit economics of AI products. ### Measuring what actually runs Attribute uses a lightweight eBPF sensor to monitor runtime activity at the Linux kernel level. [DoiT says](https://www.crn.com/news/ai/2026/doit-buys-ai-finops-startup-attribute-launches-ai-token-cost-management-product?ref=runtimewire) the sensor maps GPU, CPU, memory, network, I/O and model API consumption to the process, container, pod or request responsible for generating it. Attribute then combines those observations with provider cost data. [DoiT says the product covers](https://www.doit.com/blog/doit-launches-attribute-ai-tokenomics-without-tags-sdks-or-code-changes?ref=runtimewire) calls to OpenAI, Anthropic, Google Gemini and AWS Bedrock, automatically splitting cached, reasoning, input and output tokens while tracing token usage and model requests. The resulting records can assign costs to a customer, feature, workload or AI agent rather than leaving them grouped under a shared cloud account. Zimmermann described the thesis in [DoiT's July 7 announcement](https://www.doit.com/blog/doit-launches-attribute-ai-tokenomics-without-tags-sdks-or-code-changes?ref=runtimewire): "You can't tag your way to the truth inside a shared GPU or a single Bedrock account." DoiT says installation takes about 15 minutes and requires no SDK, tagging policy or application code changes. Those deployment and coverage claims come from DoiT and have not been independently benchmarked. The product targets an increasingly practical problem. A SaaS operator may know its total OpenAI or Bedrock bill while lacking the data needed to calculate whether an AI feature carries a healthy gross margin. Shared credentials, gateways and clusters can further obscure whether a costly model request came from a paying customer, an internal experiment or an autonomous agent caught in a loop. DoiT founder and CEO [Vadim Solovey](https://www.doit.com/meet/vadim?ref=runtimewire) has framed Attribute around that gap between infrastructure spending and business accounting. DoiT's profile of Solovey says he began building data centers in 1999, co-founded DoiT with Yoav Toussia-Cohen in 2011 and still writes code most weeks. In a [July 7 post explaining the launch](https://www.doit.com/blog/why-we-are-launching-attribute?ref=runtimewire), Solovey argued that AI infrastructure was designed for speed and scale, leaving financial attribution to be reconstructed after usage occurred. The acquisition pairs that founder perspective with Zimmermann and Tropp's runtime measurement architecture. [DoiT's leadership page](https://www.doit.com/about?ref=runtimewire) identifies Zimmermann as general manager of Attribute, while [Attribute's company page](https://attrb.io/company/?ref=runtimewire) identifies Tropp as its co-founder and CTO. ### An exit after three years Attribute was founded in January 2023, according to [Startup Nation Central](https://finder.startupnationcentral.org/company_page/attribute?section=lifecycle&utm_source=openai&ref=runtimewire). [CTech reported](https://www.calcalistech.com/ctechnews/article/sycvenedzx?ref=runtimewire) that Attribute raised approximately $13.5 million from Mensch Capital, SCapital, HarelTech and IBI. Startup Nation Central lists Attribute's funding amount as undisclosed, so the $13.5 million figure remains an attributed estimate rather than a separately confirmed total. At CTech's estimated $65 million purchase price, the acquisition would represent a relatively fast outcome for a business founded roughly three and a half years ago. The structure of the consideration and Attribute's revenue were not disclosed, preventing a meaningful calculation of the return to founders, employees or investors. The acquisition gives DoiT a runtime-based approach to AI cost allocation as FinOps vendors and cloud providers add competing methods. [Finout](https://www.finout.io/blog/how-finout-connects-ai-spend-to-real-business-outcomes?ref=runtimewire) markets Virtual Tags for allocating AI spending to teams, product lines, features and customer segments, and for measuring cost per inference, customer or feature. [Vantage](https://www.vantage.sh/blog/llm-token-allocation-preview?ref=runtimewire) introduced an LLM token-allocation preview in February that joins billing records with observability data and customer-supplied metadata. [AWS](https://aws.amazon.com/blogs/machine-learning/introducing-granular-cost-attribution-for-amazon-bedrock/?ref=runtimewire) now attributes Bedrock inference costs to IAM identities and supports tags for aggregation by team, project or tenant. Attribute's bet is that those records become incomplete when many users and agents share the same infrastructure or gateway. Kernel-level observation could give DoiT finer attribution across that shared layer. DoiT's assertion that its approach is unique has not been independently established, and competitors are adding their own AI allocation features. ### DoiT is assembling a broader cost stack Attribute follows several DoiT acquisitions across cloud operations. DoiT announced its purchase of architecture mapping business [LiveDiagrams on January 14, 2025](https://www.doit.com/blog/doit-acquires-livediagrams-to-accelerate-cloud-infrastructure-optimization?ref=runtimewire), Kubernetes optimization provider [PerfectScale on February 4, 2025](https://www.doit.com/blog/doit-acquires-perfectscale-elevating-kubernetes-cost-optimization-for-finops?ref=runtimewire), and cloud-security platform [CloudWize on October 23, 2025](https://www.doit.com/blog/doit-acquires-cloud-cybersecurity-platform-cloudwize-to-converge-cloud-security-and-outcome-focused-approach-to-cloud-intelligence?ref=runtimewire). It acquired data-platform cost specialist [SELECT](https://www.doit.com/blog/doit-acquires-select-to-eliminate-snowflake-waste?ref=runtimewire) in January 2026. [DoiT says it manages over $20 billion in cloud spending for 4,500 customers across 27 countries](https://www.doit.com/blog/doit-launches-attribute-ai-tokenomics-without-tags-sdks-or-code-changes?ref=runtimewire), with the customer and country counts also listed on its [About page](https://www.doit.com/about?ref=runtimewire). Those figures position Attribute for distribution through an existing customer base that already buys cloud-finance and optimization tools. The sale also converts the founders' technical argument into a broader platform strategy: assigning AI spending to the product activity that created it while the workload is running, rather than relying solely on reconstruction from a provider invoice. Tying AI costs to individual customers and features makes the data useful beyond infrastructure monitoring. Product teams can price usage, finance teams can calculate margins and engineering teams can identify which agents or model choices are consuming the budget. Zimmermann and Tropp built the measurement layer for that calculation; DoiT bought it as AI cost attribution was becoming a feature inside major clouds and FinOps platforms.