The Compendium · Part 13

Startups

2 pieces, oldest first.

May 21, 2025 · Startups

Teenage YC Founders Launch SAVA Robotics, Bring US Manufacturing Tech to Chinese Standard

The Y Combinator-backed team plans to expand from AI robot workers to fully-autonomous factories

Daniel Neuner

Last week, the founding team of SAVA Robotics, a manufacturing robotics startup out of Y Combinator’s 2025 spring batch, officially unveiled their first product, the CNC Press Brake, to the public. SAVA doesn’t just have the potential to become a unicorn company. The startup, with plans to launch fully-autonomous factories, could entirely transform the United States’ production capabilities and close the gap between East Asian competition.

“SAVA is here to bring the supply of AI-driven and data-informed robotics to manufacturing. American manufacturing can’t simply throw lots of money on inflexible, clunky hardware that replaces existing brick and mortar shops,” co-founder Alessio Toniolo told the Perrin Institution. “We need to build solutions that retrofit existing manufacturing into 50,000 small fabrications.”

How It Started

After SAVA CEO and Georgia Institute of Technology drop-out Jakob Knudsen effectively pitched the company to Y Combinator General Partner Aaron Epstein at a dinner, he and his high school-aged co-founders, Alessio Toniolo and Vedic Patel—who met their junior year of high school at Georgia Tech’s Research Insitute—were ushered into a YC interview and ultimately accepted into the highly-prestigious accelerator. A month later, they released CNC Press Brake, a 250-pound robot model they built in a shared San Francisco apartment.

“Y Combinator started with a rollcoaster, with experts telling us SAVA has no market nor technical viability,” said Toniolo. “Today, SAVA Robotics is focused on bringing manufacturing back onshore by creating technology—designed to integrate with existing infrastructure—that engages human workers rather than replacing them.”

As high school and college drop-outs without long careers of experience, the three SAVA co-founders have explored countless factories throughout San Francisco to better understand the market.

“We have been to every sheet metal shop in the Bay Area, talked to startups in cutting-edge automation, and learned from traditional companies that produce machinery for sheet metal,” said Toniolo.

“We saw automation cells collecting dust and more than half of the machinery not being utilized,” said Knudsen.

The team bore witness to antiquated tools taking up space in factories that lack modern efficiency. Additionally, they came to understand that the American manual labor force is only growing older, and technology needs to fill the gaps left by aging workers.

“Over half of skilled labor in the U.S. is retiring and over the age of 50. As manufacturing is reshored to American soil, we require a new era of versatile, intelligent robots to augment our labor force,” said Toniolo.

In-house automation is not an easy fix for these shops. The time and financial cost of renovating factories and implementing more advanced technology holds many managers back.

“Small shops can’t afford automation: robots need time-consuming programming, causing owners to turn down business,” said Toniolo.

The History of US-China Manufacturing Competition

“The consistent topic that is brought up when we visit shops is China. China is catching up to the U.S. in AI talent (DeepSeek) while still retaining the upper hand in infrastructure,” said Toniolo.

In the post-WWII era and into the 1970s, the United States dominated all tiers of manufacturing, with complete industrial ecosystems and a strong middle-tier manufacturing base. Since the 1980s, the U.S. has shifted its focus to design and software, using foreign labor and resources for raw materials and initial manufacturing while completing the late-stage production process on home soil.

“The outsourcing done in the 80s had its benefits at the time, but now we are paying the price,” said Knudsen.

In this time, China has taken over as a global supplier of manufactured goods, controlling all areas of production as did the U.S. after WWII. Using innovative tools and techniques developed by American companies, China has since become a global leader in product exports. They have, by a wide margin, the highest manufacturing output in the world: over 30% of global output, nearly twice the amount produced by the U.S.

Now, in addition to making advancements in software, China is experimenting with fully automated shops, known as “dark factories,” that function without human labor or supervision. Recognizing the gap between the U.S. and China, the SAVA team understands the advanced utilization of robotics and AI needed to achieve this level of automation, and China’s constant innovation is informing SAVA of where to build next.

SAVA’s First Robot: The CNC Press Brake

SAVA’s first robot, the CNC Press Brake, replaces a technician who would traditionally operate factory machines that form metal sheets.

“The robot uses VLMs—think LLMs with a camera—to process information from the machinery itself. Then it uses a bunch more complex models to recognize and manipulate the parts and this all feeds into the larger system of the factory,” said Patel.

The CNC Press Brake is the first of SAVA’s line of automated robot operators. Next, the team is looking to create robots that can complete other factory tasks, such as welding. All of SAVA’s groundbreaking products will be autonomous workers that make sense of the tasks they are completing, rather than relying on direct commands from human factory workers—saving shop owners significant time for other operational tasks.

“SAVA uses state-of-the-art AI vision models. Our robots see and understand the real world with aerospace-grade precision,” said Toniolo.

The SAVA Future: Fully-Automated Factories

“We are slightly shifting our focus from just the robot that does the one specific process of sheet metal to creating an end-to-end factory for sheet metal that is fully automated and can rapidly scale through acquiring smaller shops,” said Patel. “Now the vision for our company is to basically create Tesla Gigafactories that are fully autonomous for sheet metal through and through.”

 Leveraging the edge they have with their advanced robotics technology, SAVA aims to scale through vertical integration. They plan to have autonomous robots that communicate with one another to manufacture parts independently of humans.

“The factory system will give the robot a set of parts to complete, and the robot can use state-of-the-art pose estimation models to recognize and manipulate parts just based off of the CAD. Then it uses VLMs to process the CNC press brake itself to create the part,” said Patel.

“We want to build smarter factories with flexible robots, so humans don’t have to do repetitive manufacturing tasks and can focus on product design, hard engineering problems, and the advanced technology domains that will give America the edge,” said Toniolo.

Although the landscape of robotics technology and artificial intelligence is constantly changing, one thing is certain at SAVA: the team isn’t looking for a quick exit.

“For the team and I, we see this as our life’s work. When you extend the time horizon that far, the perceived risk melts along with it,” said Knudsen.

Watch SAVA’s launch video: https://youtu.be/pnrhTYbOqNc?si=fD0hjffDtvhUZjSX

June 5, 2025 · Startups

GridFree AI’s New Power Systems and the Future of Data Centers

Stealthy Houston-based startup targets data center energy and supply, integrating modularized approach with natural gas and reporting tremendous results

Nate Nadler

As AI continues to devour energy at a rate that was once reserved for entire nations, a stealthy startup bets on an off-the-grid future of computing. GridFree AI, founded in 2024 by a former BP executive and Microsoft computing guru, bridges the technology and energy sectors to make computing more efficient and environmentally friendly. Utilizing modular Power Foundries™, GridFree can convert natural gas to electricity that cools with extreme efficiency, and it is targeting the weakest link in the AI supply chain: energy. As many top AI producers search for an answer to the power problem, GridFree claims to offer a scalable and green solution.

Background & Executive Team

Founded in late 2024 by two seasoned executives, GridFree AI has made some very promising claims that could completely change the landscape of power generation in the computing world.  

The two founders have extensive experience in the technology and energy sectors, respectively. Co-founder Ralph Alexander served as the CEO and Chairman of Talen Energy, an energy infrastructure company which is currently valued in excess of $11 billion. Of similar pedigree, Co-founder Patrick Yantz was formerly part of the Global Edge Site Operations Leadership at Microsoft. Two other key leaders, Loren Long and Mark Coscio, serve as Chief Development Officer and EVP of Power Operation, respectively, and have both had extensive experience in the energy sector, with the former having served in various engineering roles across numerous companies and the latter having been the former COO at Enviva. Thanks to its combination of powerful product and seasoned executives, as of June 2025, Gridfree has raised $5 million in venture capital funding, primarily coming from Giant Venture and Amplo.

Although GridFree is yet to publicly announce the completion or commercial use of their Power Foundry™ systems, their experienced leadership team and lofty claims present a product that could truly revolutionize the computing industry.

Current Qualms with Data Centers

The cloud isn’t running out of chips – it’s running out of power.

As of today, the power demand for data centers is tremendous, at times even exceeding that of densely populated and developed nations. According to an April 2025 report from the IEA, data centers account for approximately one-half of the total power consumption of Japan, an extremely technologically advanced nation with a population in excess of 120 million people. Furthermore, Goldman Sachs stated in a February article that the energy demand for data centers could increase north of 165%.  

Northern Virginia, the undisputed capital of data centers in the US, is believed to have 70% of worldwide computing traffic run through its data centers during at least one point in the process. Despite their large scale and dominance in the industry, data centers in Northern Virginia are reported to have numerous issues related to energy demand. A 2024 report from the American Public Power Association noted that new natural gas plants “would need to be added at the rate of about one large 1,500 MW plant every two years for 15 consecutive years.”  

Clearly, there is an energy crisis brewing in the world of data centers, with energy consumption projected to grow radically due to the widespread institution of AI use across all industries.

The GridFree Premise & Potential Implications

GridFree AI proposes a decentralized approach to data center infrastructure, basing its organization on the premise of the Power Foundry™ – an off-grid, modular unit that can generate substantial power, store energy, and cool autonomously. The unit is said to seamlessly incorporate the utilization of natural gas, battery storage, and integrated liquid cooling.

Based on initial claims of increased efficiency, GridFree AI could completely revolutionize the computing industry. Touting absolutely exemplary efficiency numbers, GridFree AI’s product is reported to convert natural gas into electricity and cooling with an excess of 90% efficiency, which eclipses the grid’s 36% efficiency. Furthermore, a one-third reduction in CAPEX (capital expenditures) and OPEX (operating expenses) is claimed, potentially providing a way to enlarge margins significantly. Lastly, industry experts predict that a switch to the Power Foundry™ could result in an increase of 50% in available energy for information technology related computing.

Despite publicizing some very promising statistics, there is still much to be learned about GridFree’s plan, as there is currently not much public information about their technology being implemented. They also face numerous potential roadblocks, like a continued high consumption of natural gas, zoning and regulatory issues, and stiff competition from powerhouse competitors like AWS and Microsoft.  

While much of GridFree AI’s future is unclear, they certainly present a seemingly plausible solution to big tech’s impending energy crisis.