Five funding deals tracked by Crunchbase News reveal the expanding scope of startups using artificial intelligence and automation beyond the traditional software stack. The projects that attracted funding this month are working on floating nuclear reactors, evaluating models’ ability to control robots, maintaining commercial properties, purchasing construction materials, and recording agricultural operations by voice.
This article is part of a monthly column tracking startup deals that may not receive sufficient attention. The five companies do not share a single business model, but their common thread is an attempt to connect software or artificial intelligence with physical and field operations that are difficult to execute through a computer alone.
Nuclear power aboard barges
Bluecore Energy, based in Long Beach, California, raised a $50 million seed round led by Silverton Partners, just two months after emerging from stealth, when it had secured $10 million in pre-seed funding. The investor list included Slauson & Co., Harlem Capital, Black Angel Group, HartBeat Ventures, and other investors.
The company is developing compact, integrated, water-cooled small modular reactors designed to operate aboard floating barges. The idea is to manufacture the systems and then transport them to locations that need electricity, rather than building a new plant and its infrastructure on land over the course of years. Bluecore is initially targeting the Port of Long Beach, while identifying ports and artificial intelligence data centers as potential demand gaps. The company says it is working with the Nuclear Regulatory Commission and the U.S. Coast Guard to obtain the necessary approvals, meaning the funding does not yet equal commercial or regulatory readiness.
AI in supply chains and maintenance
Saudi Arabia-based BRKZ, headquartered in Riyadh, secured $31 million, including a $13 million Series B equity round co-led by Aramco’s Wa’ed Ventures and 500 Global, along with an $18 million growth debt financing commitment from Stride Ventures under a previously announced facility. This brought the company’s total raised to more than $83 million, according to Crunchbase.
BRKZ operates a marketplace connecting construction companies with building-material suppliers, offering procurement, logistics, and financing services. It says it has collected approximately 38 million structured data points that it uses in a materials-pricing engine trained on nearly 40,000 requests for quotations. According to company figures, price forecasts fall within 5% of the final transaction price in 84% to 89% of cases. An AI agent also reads images of cement delivery receipts sent via WhatsApp, matches them with orders, and verifies delivery, processing roughly three-quarters of these operations without human intervention.
Viabot, based in Santa Clara, California, raised $24 million in a Series A round led by Walden International to expand a fleet of robots aimed at commercial properties. The robots perform tasks such as sweeping, removing debris, and landscaping, while providing what the company calls “low-intervention security” through patrols. Viabot is adopting a “robot-as-a-service” model rather than attempting to build a multipurpose humanoid robot, focusing on recurring outdoor work.
Who evaluates physical AI?
Robocurve, a three-month-old startup based in San Francisco, raised $10 million in seed funding led by Initialized Capital, with participation from Y Combinator, Notable Capital, Halcyon Futures, Decasonic, and other investors.
The company wants to operate as an independent auditor of physical AI by testing how well advanced models can control real robots and publishing the results publicly. It says its initial research found that general-purpose large language models can outperform specialized robotics vision-language-action models on some simple tasks. Robocurve was also established as a Public Benefit Corporation, and says AI labs do not determine its research agenda, evaluation methodology, or published results.
The company intends to fund academic teams and provide them with robotic arms to build open-source tests. More than 200 institutions have enrolled in its program, including researchers from 19 universities among the world’s top 20 universities, according to the company. Robocurve offers a combined $500,000 in funding and free robotic arms to participating groups. The independence and effectiveness of these evaluations remain open questions requiring practical testing, especially because the preliminary figures cited in the article come from the company itself.
A voice interface for agricultural workers
Tellia, which operates between San Francisco and Paris, raised $5 million in a pre-seed round led by Revent, with participation from Grey Silo Ventures, Jeriko, and Fund F. The company is developing an AI platform powered by voice for farmers, agronomists, and workers in the sector.
Instead of stopping work to fill out a form or write notes, users can send a voice memo, a WhatsApp message, or an image to record data, create reports, and set reminders. Tellia says its system turns unstructured information into records linked to the correct field, crop, and team. It says its technology is being used across one million acres, including at Campos Brothers Farms and Duckhorn wineries in the United States, as well as by agricultural organizations in Europe.
What do these deals mean?
The deals do not provide evidence of final commercial success, but they show where investors are directing their attention: tools that reduce manual work in specific operations, add a measurement and accountability layer to autonomous systems, or attempt to address infrastructure and energy constraints. Crunchbase data cited in the article indicate that physical AI companies raised $47.4 billion across 521 deals in the first half of 2026, while robotics companies alone raised more than $21 billion during the same period.
At the same time, the constraints remain clear. Floating reactors need a regulatory approval pathway, pricing-accuracy or automation claims require independent verification, and deploying robots in properties requires dealing with varied field environments. Even in agriculture, the system’s value depends on the quality of voice and image data and on linking it to the correct record. Therefore, the actual development here is not merely adding an AI interface, but attempting to integrate these systems into a measurable physical workflow whose results and risks can be assessed.