When a Red Bull Racing aerodynamics engineer walked away from the pinnacle of motorsport, few expected his next move would be to reshape the backbone of global manufacturing. Bercan Kilic’s departure from Formula One wasn’t driven by disillusionment with speed, but by a desire to apply his precision engineering skills to a problem with far broader societal impact: teaching robots to perform useful work in factories. The decision has already attracted serious capital, with his Munich-based startup microagi closing a $55 million seed round—the largest ever for a German early‑stage venture. Led by Hummingbird and joined by Northzone, LocalGlobe, Village Global, and redalpine, the round underscores growing investor confidence that the next leap in AI will happen not in chatbots but on the shop floor. While the valuation remains undisclosed, the size of the round signals a belief that microagi’s approach to bridging the gap between cutting‑edge AI models and real‑world robotics could unlock a new era of productivity.
Kilic’s motivation stems from a simple observation: the world’s most experienced factory workers are retiring, and there are not enough new entrants to replace them. During his time at Red Bull, he marveled at the sophistication of aerodynamic simulations, yet he realized that the same analytical rigor could be redirected toward creating abundance rather than shaving milliseconds off a lap time. He envisions a future where low‑cost, reliable robotic labor makes goods and services more accessible, helping to counteract inflationary pressures driven by wage growth and scarce skilled labor. This vision aligns with macro‑economic trends such as reshoring, where companies seek to bring production closer to home markets but only if automation can fill the labor void left by demographic shifts. By focusing on deployable robotics rather than purely theoretical research, microagi aims to turn engineering excellence into tangible economic value.
Unlike many robotics ventures that design their own hardware or foundational AI models, microagi takes a distinctly pragmatic route: it captures how humans perform tasks and uses that data to fine‑tune existing robotic systems. Workers wear sensor‑equipped gloves and are filmed with cameras as they carry out specific operations—whether assembling a component, packaging goods, or handling food ingredients. The resulting footage serves as a rich, task‑specific demonstration that microagi feeds into partner AI models, teaching them the nuances of motion, force, and sequencing required for each job. This method sidesteps the enormous cost and time associated with building robots from scratch while ensuring the trained behaviors are directly relevant to the customer’s environment. It also creates a feedback loop where continuous data collection improves performance over time, much like how human workers refine their techniques through experience.
The company’s consumer‑facing brand, shift, has become an unconventional yet effective engine for gathering the diverse data needed to train versatile robotics. By offering free apartment cleanings in New York and, more recently, complimentary private chef services in San Francisco, shift incentivizes people to record themselves performing everyday chores such as dishwashing, mopping, and meal preparation. These first‑person videos provide a treasure trove of dexterous manipulation examples that are scarce in traditional robotics datasets. Operating across fifteen countries and compensating over twenty thousand contributors, shift has rapidly scaled into a data‑collection network that rivals dedicated annotation houses. The model is simple yet powerful: provide a valuable service to consumers, capture high‑quality behavioral data, and sell that data to AI labs and robotics firms seeking to improve the “brains” of their machines.
Several converging forces have created a fertile environment for microagi’s rise. Artificial intelligence, after mastering language and image generation, is now being applied to physical systems where perception, control, and interaction with unpredictable environments matter. Simultaneously, the cost of essential robotic hardware—actuators, sensors, and compute modules—has fallen dramatically, making large‑scale deployment financially viable. On the labor side, advanced economies are confronting shrinking workforces and persistent shortages in manufacturing, logistics, and healthcare, prompting companies to seek automation as a strategic necessity rather than a luxury. Finally, geopolitical tensions and supply‑chain fragility have intensified pressure to reshore production, but reshoring only succeeds if robots can reliably perform the work that would otherwise be done offshore. Together, these trends have attracted significant venture capital to the physical AI sector, positioning startups like microagi at the forefront of a transformative wave.
Concerns about robots displacing human workers are common, but Kilic argues that microagi’s technology addresses a different problem altogether: the lack of enough workers to fill existing roles. He points out that in many factories, the most experienced operators are retiring en masse, and younger generations are not entering these trades at sufficient rates. In this context, robots are not taking jobs from people; they are occupying positions that would otherwise remain vacant, thereby preventing production bottlenecks and supporting wage stability for the humans who remain. By augmenting the workforce rather than replacing it, automation can help maintain output levels while allowing human employees to focus on higher‑value tasks such as process oversight, quality improvement, and innovation. This perspective reframes the narrative from one of job loss to one of workforce augmentation, a nuance that resonates with policymakers and industry leaders grappling with demographic challenges.
One of the most persistent obstacles in robotics is the staggering scarcity of data that captures how machines should move and interact with the physical world. While large language models have been trained on hundreds of billions of text tokens harvested from the internet, comparable datasets for robotic dexterity simply do not exist—a gap UC Berkeley roboticist Ken Goldberg has vividly termed the “100,000‑year data gap.” shift directly attacks this deficiency by systematically gathering first‑person, sensor‑rich recordings of humans performing a wide array of manual tasks. The resulting corpus provides the variety and granularity needed for AI models to learn generalizable motor skills, much like image nets enabled breakthroughs in computer vision. Competing services such as Scale AI, Turing, and micro1 pursue similar goals, but microagi’s integration of shift with its deployment pipeline creates a unique end‑to‑end solution that turns raw data into factory‑ready behavior.
Microagi’s technical strategy hinges on leveraging the rapid progress made by specialized AI labs in developing foundational robotics models, then customizing those models with proprietary, task‑specific data. Rather than attempting to train a monolithic model from zero, the company partners with firms like Physical Intelligence, Skild AI, and Generalist AI, adapting their architectures to the nuances of each client’s operational environment. This approach combines the breadth of pre‑trained knowledge with the precision of fine‑tuning, analogous to how developers adapt large language models for specific industries using domain‑specific text. Once the model is sufficiently capable, microagi assists with deployment on the customer’s existing robotic hardware, ensuring seamless integration into production lines. The company’s current focus spans automotive assembly, logistics warehousing, and food processing—sectors where repetitive yet variable tasks benefit immensely from adaptable automation.
The fresh infusion of capital will be allocated to three primary areas that amplify microagi’s scaling potential. First, a significant portion will fund the compute-intensive process of training and refining robotics models, which requires substantial GPU hours to process the high‑dimensional sensor data collected by shift. Second, the money will expand shift’s data‑collection footprint, both geographically and in terms of the variety of tasks captured, thereby enriching the training corpus for future model iterations. Third, microagi aims to deepen its presence in the United States, building on its initial foothold in New York to serve manufacturers across the Midwest, South, and West Coast. With a global team of thirty‑seven core employees and shift contributing roughly seventy‑five additional staff, the company is positioned to translate its Munich‑born innovation into a worldwide service network capable of supporting diverse industrial clients.
To gauge where robotics stands in its evolutionary journey, Kilic draws a useful analogy to the progression of language models. The “GPT‑2 moment” marked the point when researchers confirmed that scaling data and compute yielded predictable, linear improvements in performance. Today, the robotics field is still deciphering its own scaling laws—figuring out how much data, what model architectures, and what compute levels are needed to achieve reliable, generalizable behavior. Kilic believes the industry is rapidly approaching its “GPT‑3.5 moment,” the threshold at which models become sufficiently competent to be deployed broadly across real‑world applications without excessive custom engineering. Reaching this inflection point would transform robotics from a niche, high‑cost endeavor into a mainstream driver of productivity, much like how GPT‑3.5 enabled widespread adoption of AI‑powered chatbots and copilots.
Investor enthusiasm for microagi is fueled not only by its technology but also by the palpable urgency of its mission. Hummingbird managing partner Firat Ileri, who led the round, recalled being struck by the team’s relentless focus during a visit to the Munich headquarters, where engineers appeared to live and breathe their mission. Ileri highlighted the founders’ blend of deep technical pedigree—spanning former Mercedes F1 aerodynamics, Alan Turing Institute research, RWTH Aachen engineering, and serial entrepreneurship—and their acute awareness of Europe’s demographic trajectory. The European Union’s median age has risen to 44.9 years, up from 39.6 just two decades ago, and the European Commission projects a potential workforce contraction of up to 18.8 million people by 2050. In this context, automating factories is not merely an option; it is an economic imperative to sustain competitiveness against regions like China, which installed 295,000 industrial robots in 2024—over half of the global total—while the United States added only 34,200 units in the same period.
Looking ahead, the five cofounders have set an audacious benchmark for success: within five years, they aim to have microagi’s technology operating on tens of millions of robots worldwide, explicitly stating that deploying fewer than twenty to thirty million units would constitute a major shortfall. This goal reflects a conviction that the convergence of accessible AI models, rich task‑specific data, and affordable hardware can finally enable mass‑market robotics. For manufacturers, the practical takeaway is to begin auditing their operations for tasks that are repetitive, highly variable, and currently dependent on scarce skilled labor—these are the prime candidates for early automation pilots. Investors should look for startups that combine credible data‑acquisition mechanisms (like shift) with clear pathways to model adaptation and deployment, rather than those promising breakthroughs. By monitoring, and setting with and regulatory steps, businesses can position themselves to reap the rewards of the physical AI revolution before the window of opportunity narrows.