Nvidia Rival Cerebras Secures $1B Funding at $23B Valuation for Wafer-Scale AI Infrastructure
Cerebras Systems, a leading challenger to Nvidia in the artificial intelligence hardware market, has raised $1 billion in fresh funding at a reported $23 billion valuation, marking one of the largest recent investments in next-generation AI infrastructure. The funding round highlights growing investor appetite for alternatives to traditional GPU-based computing as demand for large-scale AI training and inference continues to surge worldwide.
Founded in Sunnyvale, California, Cerebras is best known for its wafer-scale chip technology — an unconventional approach that builds a processor from an entire silicon wafer rather than cutting it into smaller chips. This design allows far more cores, memory, and bandwidth to be placed on a single device, significantly reducing the communication delays that often slow down massive GPU clusters. The company says its systems are designed specifically for training and running very large AI models faster and more efficiently.
The new capital will be used to expand manufacturing, scale cloud and data-center deployments, and accelerate development of its AI supercomputing platforms. Cerebras has increasingly positioned itself not just as a chipmaker, but as a full AI infrastructure provider, offering integrated hardware and software systems optimized for frontier-scale models.
The funding comes at a time when AI developers and enterprises are seeking diversified compute options amid tight GPU supply and rising infrastructure costs. By focusing on large, unified processors instead of distributed chip clusters, Cerebras aims to simplify AI model training while lowering energy and networking overhead.
Industry analysts say the investment signals confidence that specialized architectures can compete alongside — and in some cases outperform — conventional GPU stacks for certain AI workloads. With fresh capital and a sharply higher valuation, Cerebras is now seen as one of the most serious competitors in the race to power the next generation of artificial intelligence systems.
Ex-OpenAI and DeepMind Researchers Raise $150M to Tackle AI Hallucinations
A startup founded by former researchers from OpenAI and DeepMind has raised $150 million in new funding to develop tools that help detect and debug AI hallucinations — one of the most persistent challenges in generative artificial intelligence. The funding round signals rising investor focus on AI reliability, transparency, and safety infrastructure as adoption accelerates across industries.
The company is building diagnostic and interpretability tools that allow developers to look inside large AI models and better understand how they arrive at specific outputs. By mapping internal reasoning patterns and activation pathways, the platform aims to help engineers identify where and why models produce fabricated or inconsistent information. The goal is to move AI systems away from “black box” behavior toward more inspectable and controllable performance.
AI hallucinations — confident but incorrect outputs — remain a major barrier to enterprise deployment in high-stakes fields such as finance, healthcare, law, and scientific research. Current mitigation methods often rely on external guardrails or post-generation filtering. The new tools instead focus on model-level debugging, enabling targeted fixes and behavior steering during development and fine-tuning.
According to the company, its technology can highlight which internal components of a neural network are responsible for certain claims or errors, allowing teams to adjust or constrain those mechanisms directly. Early trials with model developers and research labs have shown improvements in factual consistency and response traceability.
Investors backing the round include major venture firms and strategic technology players, reflecting a broader shift toward funding AI infrastructure layers beyond model training and compute. Analysts say interpretability tooling could become a core part of the AI software stack, similar to testing and monitoring frameworks in traditional software engineering.
The company plans to use the new capital to expand its research team, scale its platform, and partner with model builders seeking greater control and accountability in AI systems.
Fibr’s $7.5M Could Make Every URL an Intelligent Experience — Here’s How
AI startup Fibr has raised $7.5 million in fresh funding to build technology that turns ordinary web links into intelligent, interactive experiences. The company’s core idea is simple but ambitious: every URL should not just display information, but understand it — and respond to users in real time.
Traditionally, webpages are static destinations. Users open a link and manually scan, search, and interpret the content themselves. Fibr aims to change that by adding an AI layer that automatically analyzes each page and converts it into a dynamic, queryable interface. Instead of reading everything line by line, users can ask questions, request summaries, extract key data, and get personalized explanations directly from the page.
Fibr’s platform uses large language models and content-mapping systems to parse structure, meaning, and context across a webpage. Once processed, the link effectively becomes “interactive.” A product page could answer feature questions instantly. A research article could generate simplified summaries or highlight methods and findings. A long policy document could be converted into key points and action items within seconds.
The technology is designed for broad use across education, marketing, e-commerce, and enterprise knowledge systems. For teams, it could turn internal documents and dashboards into conversational resources. For learners, it could make dense material easier to understand. For businesses, it could increase engagement by making landing pages responsive and adaptive to visitor intent.
The new funding will support product development, hiring, and integrations with browsers and workflow tools. Fibr is also working on developer features that allow companies to embed intelligent link behavior directly into their platforms.
As the web becomes more crowded and information-heavy, tools that reduce friction and increase clarity are gaining investor attention. Fibr’s bet is that the future of browsing is not just clicking links — but talking to them.
DeepMind Alumni Raise $255M to Launch NEXUS, AI Platform for Enterprise Spreadsheets
A new AI startup founded by former DeepMind researchers has raised $255 million to launch NEXUS, a platform designed to bring advanced artificial intelligence directly into enterprise spreadsheet workflows. The sizable funding round reflects growing investor confidence in AI tools that enhance everyday business software rather than replacing it.
Spreadsheets remain one of the most heavily used tools across industries, powering financial models, forecasts, reporting, and operational planning. Yet they are also a major source of inefficiency and risk, often relying on complex formulas, manual updates, and fragile logic chains. NEXUS aims to modernize this environment by embedding AI that understands both data and spreadsheet structure.
The NEXUS system allows users to interact with spreadsheets using natural language. Instead of manually writing formulas or building multi-step functions, users can describe what they want to calculate, analyze, or predict. The AI then generates formulas, runs models, flags inconsistencies, and explains results in plain language. It can also detect anomalies, suggest corrections, and simulate scenarios based on historical patterns.
Beyond individual productivity, the platform focuses on enterprise collaboration and governance. Features include automated audit trails, logic validation, version intelligence, and policy-aware formula generation. This is intended to reduce costly spreadsheet errors while improving transparency across teams handling sensitive financial and operational data.
The founders say their approach combines large language models with specialized reasoning systems trained on structured data and quantitative workflows. That hybrid design helps the AI handle numerical reasoning and multi-step calculations more reliably than general-purpose assistants.
The new capital will support product development, enterprise integrations, and global expansion. Early pilots are underway with finance and analytics teams.
As AI shifts from experimental tools to embedded workplace infrastructure, NEXUS is betting that the spreadsheet — long considered humble but indispensable — is the next major frontier for intelligent automation.
Ex-Waymo Engineers’ Bedrock Robotics Raises $270M to Automate Construction Sites
Bedrock Robotics, a startup founded by former Waymo engineers, has raised $270 million in new funding to bring autonomous technology to construction sites, signaling strong investor confidence in robotics for heavy industry. The company is focused on developing AI-powered systems that can automate construction equipment and workflows, aiming to improve safety, speed, and efficiency across large-scale projects.
Construction remains one of the least automated major industries, despite rising costs, tight labor supply, and persistent safety risks. Bedrock Robotics plans to address these challenges by adapting self-driving and machine perception technologies — originally developed for autonomous vehicles — to operate in rugged, constantly changing job-site environments.
The company is building autonomy software that can be integrated into heavy machinery such as excavators, loaders, and haul vehicles. Using sensor fusion, real-time mapping, and task-planning AI, these systems can navigate uneven terrain, avoid hazards, and carry out repetitive operations with minimal human input. Human supervisors remain in the loop, but the machines handle much of the continuous precision work.
Bedrock’s platform is designed to work with existing fleets rather than requiring entirely new vehicles. This retrofit-friendly approach could make adoption easier for contractors and infrastructure firms that want automation without replacing expensive equipment. The system also collects operational data that can be used to optimize workflows, reduce downtime, and improve project forecasting.
The newly raised capital will support expanded field trials, hardware partnerships, and commercial deployments across major construction and infrastructure projects. The company is also investing in simulation and digital twin technology to train and validate autonomous behavior before machines reach live sites.
Industry analysts say the large funding round reflects a broader shift as autonomy moves beyond roads and warehouses into complex outdoor industries. If successful, Bedrock Robotics could help reshape how construction projects are executed, bringing software-driven precision and scalability to one of the world’s most physically demanding sectors.
