North America
US-Based AI Evaluation Startup Patronus AI Raises $50M to Scale Digital Simulation Platform for Agent Testing
The funding will help Patronus AI expand its simulation platform, enabling AI companies to stress-test autonomous agents before real-world deployment.
San Francisco-based AI startup Patronus AI has raised $50 million in a Series B funding round led by Greenfield Partners, as demand grows for tools that evaluate the reliability of increasingly autonomous AI agents. The round also attracted participation from Notable Capital, Lightspeed, Datadog, and Samsung, bringing the company’s total funding to $70 million.
Founded in 2023 by former Meta AI researchers Anand Kannappan and Rebecca Qian, Patronus AI develops simulation environments that allow AI companies to test how autonomous agents perform across complex, real-world scenarios before they are deployed. The company says its customer base now includes nearly every frontier AI lab alongside a growing number of AI startups.
Building Digital Worlds to Test AI Agents
As AI systems evolve beyond conversational assistants into agents capable of executing multi-step tasks independently, ensuring their reliability has become a growing challenge for developers.
Patronus AI addresses this by creating what it calls “digital world models”—simulated versions of websites, software platforms, and enterprise systems where AI agents can safely perform complex tasks. Within these environments, agents are evaluated after reinforcement learning training to determine whether they complete objectives accurately or rely on unintended shortcuts that could lead to failures in real-world applications.
The company compares its approach to the simulation techniques used by autonomous vehicle developers such as Waymo, where synthetic environments expose systems to rare or unpredictable situations before they encounter them on public roads.
According to Glenn Solomon, Managing Director at Notable Capital, demand for Patronus AI’s evaluation platform has become “nearly insatiable,” with the company’s revenue increasing fifteenfold over the past year.
Expanding AI Evaluation Beyond Benchmarks
Traditional AI benchmarks often measure model capabilities through standardized tests, but Patronus argues they provide limited insight into how autonomous agents perform when executing extended workflows in dynamic environments.
Instead, its platform focuses on verifying agent behavior while identifying cases where AI systems complete tasks incorrectly by exploiting shortcuts rather than following intended reasoning processes.
Kannappan said the company is currently concentrating on use cases where outputs can be objectively verified, including software engineering and financial workflows. However, the long-term vision extends to more complex environments where AI agents may need to operate autonomously for hours, days, or even weeks while maintaining consistent performance.
“We’re focused today on problems that can be verified immediately,” Kannappan said, adding that the company ultimately aims to build environments capable of evaluating agents during much longer and more sophisticated workflows.
Positioning for the Next Phase of AI Adoption
Rather than competing directly with companies providing human feedback for reinforcement learning, Patronus AI positions itself as an automated evaluation platform designed to measure AI agent performance without relying on human reviewers.
The company says its primary competitors are the internal evaluation teams that major AI labs have built to test their own models. As enterprises increasingly adopt autonomous AI systems for software development, finance, and other mission-critical applications, Patronus believes demand for scalable agent evaluation infrastructure will continue to accelerate.
With the new funding, the company plans to further expand its simulation platform and develop more advanced digital environments capable of testing increasingly sophisticated AI agents before they are deployed in real-world settings.
North America
US-Based Cloud Startup Blacksmith Raises $45M Series B at $550M Valuation to Scale AI Code Validation Platform
The Y Combinator-backed startup will expand computing capacity as AI coding agents generate more software that needs to be built, tested and reviewed.
US cloud infrastructure startup Blacksmith has raised $45 million in a Series B funding round led by Peak XV Partners, valuing the company at $550 million as AI-generated code creates growing demand for software validation infrastructure.
Existing investors Y Combinator and GV also participated in the round, which closed in March 2026 and was publicly announced on August 12. The financing brings Blacksmith’s total funding to $58.5 million, following a $3.5 million seed round and a $10 million Series A.
The company is scaling at a time when AI coding tools are accelerating software development but also increasing the volume of code that engineering teams must test and review before release.
AI Coding Creates a New Infrastructure Bottleneck
Blacksmith said continuous integration, or CI, jobs running on its platform have grown between 5% and 10% each week since the beginning of 2026.
CI systems automatically build and test software changes, allowing developers to identify problems before new code is merged or released.
More than 6,000 companies now use Blacksmith, including Supabase, Clerk, Ashby and Mercury. That represents a sharp increase from roughly 800 organizations when the company announced its Series A in September 2025.
Blacksmith attributes part of that growth to the adoption of coding agents such as Claude Code and OpenAI’s Codex. As engineering teams use AI to produce more code and open more pull requests, the company sees validation becoming an increasingly important part of the development process.
Co-founder and CEO Aditya “JP” Jayaprakash said AI has made writing code considerably easier without creating the same improvement in validation.
Engineering teams adopting coding agents are generating several times more pull requests, he said, turning CI into a potential bottleneck because every piece of AI-generated code still needs to be built, tested and reviewed before deployment.
Building Infrastructure for Continuous Integration
Founded in 2024 by Jayaprakash, Aayush Shah and Aditya Maru, Blacksmith participated in Y Combinator’s Winter 2024 batch.
The startup initially focused on infrastructure for GitHub Actions workloads, which developers use to automate software builds, testing and other development processes.
Blacksmith operates computing infrastructure specifically designed for CI workloads, combining dedicated computing resources with caching and storage systems intended to accelerate testing.
The company says developers can migrate existing GitHub Actions workflows to Blacksmith by changing a single line in a workflow file.
It also claims its hardware runs twice as fast while its service costs 60% less than GitHub-hosted runners, based on Blacksmith’s own comparisons.
Expanding into AI-Assisted Software Development
Blacksmith has expanded beyond CI infrastructure with Codesmith, a cloud-based coding agent that developers can use to delegate software development tasks.
The company says Codesmith can build features and fix bugs while operating within the broader software validation workflow.
Its Autofix feature monitors pull requests for failed CI checks and review feedback. When it detects an issue, the system can attempt to diagnose the problem, generate a fix and commit the resulting change.
Blacksmith is also developing Codesmith QA, which is designed to autonomously test software changes before they are merged.
Most of the Series B capital will go toward expanding the infrastructure required to support that growth. Blacksmith currently manages hundreds of thousands of computing cores and plans to increase capacity by approximately tenfold in the coming months.
The expansion reflects the company’s broader bet that as AI agents take on a larger share of software creation, the infrastructure responsible for validating their output will become an increasingly critical part of the development stack.
North America
U.S.-Based FinTech Company Yellow Card Raises $40M to Expand Stablecoin Payment Infrastructure Across Global Markets
The strategic funding brings Yellow Card’s total equity financing to more than $120 million and will support the expansion of its Global USD Accounts and stablecoin payment rails.
Yellow Card, a US-based stablecoin infrastructure company focused on Africa and other emerging markets, has raised $40 million in strategic funding from SC Ventures, Sony Innovation Fund, Polychain Capital, Blockchain Capital, and other strategic investors.
The latest investment brings Yellow Card’s total equity financing to more than $120 million. The company will use the capital to scale its Global USD Accounts, expand its stablecoin payment infrastructure, and deepen its presence across Latin America and Asia-Pacific.
Scaling Global USD Accounts
Yellow Card’s Global USD Accounts provide businesses with a single platform for managing US dollars, holding and swapping stablecoins, managing treasury operations, and collecting or disbursing local currencies through domestic payment networks.
The company currently supports local payment rails across more than 50 countries, enabling businesses operating across multiple markets to manage cross-border financial operations through a unified infrastructure.
Yellow Card said the new funding will allow it to bring Global USD Accounts to more businesses while expanding local payment rails and currency coverage globally. Customers using its infrastructure include Visa and Western Union.
Connecting Banks to Stablecoin Infrastructure
Yellow Card is increasingly positioning itself as an infrastructure provider connecting traditional financial institutions and global businesses with stablecoin-based payment networks.
Chris Maurice, CEO and Co-Founder of Yellow Card, said the investment reflects confidence in the infrastructure the company has developed to help businesses move money without relying entirely on traditional correspondent banking networks.
Maurice added that connecting banks directly to stablecoin infrastructure represents a significant opportunity, potentially improving access to US dollars for businesses underserved by traditional cross-border banking systems.
From Africa to Global Markets
Founded by Chris Maurice and Justin Poiroux, Yellow Card was created to make cross-border money movement more affordable and efficient, particularly in emerging markets.
The company has grown into one of Africa’s largest stablecoin on- and off-ramp infrastructure providers and now employs more than 200 people across 20 countries.
Yellow Card holds licences, authorisations, or registrations across 22 jurisdictions in North America, Europe, and Africa. It has facilitated more than $10 billion in transactions, supports over 50 currencies, and has partnerships spanning Visa, Mastercard, PayPal, and Coinbase.
Expanding Beyond Africa
Although Yellow Card built its business around African and emerging markets, the company is now accelerating its international expansion, particularly across Latin America and Asia-Pacific.
The new capital will support the development of additional domestic payment connections and currency coverage, strengthening the infrastructure that links its Global USD Accounts with markets worldwide.
Alex Manson, CEO of SC Ventures, said stablecoin adoption will increasingly depend on reliable infrastructure and practical applications, highlighting Yellow Card’s role in enabling businesses across Africa to move value efficiently between markets.
Building Global Stablecoin Payment Rails
Yellow Card has gradually shifted its business model from primarily providing retail cryptocurrency access toward serving businesses and financial institutions.
The transition became increasingly visible around its $33 million Series C round in 2024, as the company focused on larger and more consistent transaction volumes generated by corporate customers.
With more than $120 million in total equity financing, Yellow Card aims to become an infrastructure layer connecting banks and businesses to stablecoin payment rails, supporting faster cross-border transactions and expanding access to dollar-based financial services across emerging and global markets.
North America
U.S.-Based DeepTech Startup Discovered Materials Raises $9M to Develop AI-Driven Materials for Advanced Chips
The Lightspeed-led round will support the expansion of Discovered Materials’ team, laboratory infrastructure and AI research agents as it develops new materials for next-generation semiconductors.
DeepTech startup Discovered Materials has raised US$9 million in a Seed funding round led by Lightspeed India Partners, with participation from Y Combinator, Peak XV Partners, and several angel investors.
The round also attracted backing from prominent technology investors and entrepreneurs including Paul Graham, Gokul Rajaram and Thariq Shihipar.
Discovered Materials will use the fresh capital to expand its team and laboratory infrastructure while scaling its AI-powered research agents designed to accelerate the discovery and development of advanced semiconductor materials.
Using AI to Accelerate Materials Discovery
Discovered Materials is developing an AI-powered research platform focused on discovering new materials for semiconductor chips.
Traditionally, developing and validating new materials for semiconductor manufacturing can take years of laboratory research and testing. The startup aims to shorten this process by combining artificial intelligence with materials science and experimental research.
Its AI research agents can help analyse potential materials, identify promising candidates and guide experimentation, allowing researchers to explore new material combinations more efficiently.
The company believes this approach could compress parts of the traditional materials research and development cycle from years into significantly shorter periods.
Tackling the Heat Challenge in AI Chips
One of Discovered Materials’ initial areas of focus is improving heat dissipation in advanced semiconductor chips.
The rapid growth of artificial intelligence is driving demand for increasingly powerful GPUs and other computing infrastructure, but higher computing performance also generates substantial amounts of heat.
Modern GPUs can reach heat densities of around 140 W/cm², making thermal management an increasingly important challenge for data centres and semiconductor manufacturers.
Discovered Materials is developing new thermal materials designed to transfer heat away from chips more effectively, potentially improving the performance and efficiency of advanced computing systems.
Supporting Next-Generation Chip Architectures
Improved thermal materials could also play an important role in emerging semiconductor technologies such as 3D chip stacking.
This architecture places multiple layers of semiconductor components on top of one another to increase computing density and performance. However, concentrating more components within a smaller physical area creates additional thermal management challenges.
By developing materials capable of dissipating heat more efficiently, Discovered Materials aims to help enable increasingly dense and powerful chip architectures.
Co-Founder Akash Ramdas said advances in computing have driven technological progress for decades, but current chips remain significantly less power-efficient than the human brain.
He added that the company has already developed new thermal materials that achieve performance comparable with products that took major chemical companies years to develop.
Lightspeed Backs AI-Driven Semiconductor Innovation
Hemant Mohapatra, Partner at Lightspeed India Venture Partners, highlighted the growing materials challenge created by increasing demand for AI computing.
He said progress in semiconductor performance is increasingly constrained by the time required to move new materials from research into production.
Lightspeed sees Discovered Materials’ combination of materials science expertise and advanced AI engineering as an opportunity to accelerate this process, initially focusing on improved heat dissipation for increasingly powerful chips.
Scaling AI Research and Laboratory Infrastructure
The $9 million Seed round will allow Discovered Materials to increase its research capacity by expanding its laboratory, hiring additional talent and scaling its AI research agents.
The company’s broader objective is to create materials that can improve the efficiency and performance of next-generation semiconductor technology as demand for AI computing continues to increase.
By combining artificial intelligence with physical experimentation, Discovered Materials aims to accelerate materials innovation for advanced chips, data centres and emerging semiconductor architectures.
-
MENA Startups7 days agoEgypt-Based Exits MENA and ACE Acquire Avanz Capital Egypt in Seven-Figure Deal to Expand Private Equity and Asset Management
-
MENA Startups6 days agoMENA-Born Enterprise Voice AI Startup HeyBreez Raises $2.5M Seed to Accelerate Global Expansion
-
News4 days agoDubai Entrepreneurs Who Raised $400K on Kickstarter Launch 12-Week Accelerator Teaching Teens to Build and Crowdfund Real Startups
-
Asia Startups1 week agoSingapore-Based RetailTech Startup Graas Raises $17M Series B, Acquires Trustana to Expand AI-Powered Commerce Platform
-
MENA Startups6 days agoSaudi EdTech AILA Raises $3M Pre-Series A to Scale AI-Powered Personalized Learning Across Regional and Global Markets
-
MENA Startups7 days agoIraq-Based E-Commerce Startup Rozenama Raises $150K Pre-Seed to Expand Platform Nationwide
-
MENA Startups7 days agoUAE-Based Startup Tax Star Raises $1.75M to Expand AI-Powered Tax Compliance Platform Across GCC
-
Asia Startups7 days agoIndian FinTech Startup Rezolv Raises $12.5M to Expand AI-Powered Lending Platform

You must be logged in to post a comment Login