The anatomy of AI investment

The AI startup frenzy is anything but hype. The first half of 2025 saw global investment in AI startups shoot to more than $205 billion, according to Crunchbase, with $145 billion raised in North America alone. And though observers may argue over their long-term viability, no one disputes the seemingly endless tsunami of new companies entering the fray. Some estimates put the number of AI startups globally at more than 10,000, with more than 5,000 emanating from the U.S.

All this activity only adds to the pressures of the venture firms trying to determine which companies have the tech and teams to develop and market the next great breakthrough. But for the investment groups focused on a small but growing corner of the market – industrial AI – the stakes are even higher.

Unlike “traditional” AI that may take the form of a human resources chatbot, or a marketing forecasting dashboard, industrial AI is the application of models and automation in mission-critical systems within energy, transportation, manufacturing, financial, healthcare, and the like. Issues common to enterprise-grade AI, like hallucinations, drift, biases, etc., are not tolerated in industrial AI. As such, selecting the right startups with whom to invest in this crazy environment is tough.

“The fervor and volatility of the AI startup landscape presents a perfect storm for VCs, really,” says Gayathri Radhakrishnan, a partner at Hitachi’s investment arm, Hitachi Ventures. “You start by applying the myriad concerns you have with all startups, like cashflow, business structure, vision, team make-up, IP, etc. Simultaneously, you have to consider the chaos swirling around AI. And then on top of it all, you have to put on the industrial lens and ask, can these guys make it?”

For Hitachi Ventures, which invests in firms across the technology spectrum, due diligence in the AI space requires a focused vision on the sector that best aligns with its business.

What an industrial AI investment looks like

That, however, can be easier said than done. The explosion of AI startups has created challenges reminiscent of the cloud computing hype cycle in the early 2010s. “When cloud became popular, every company claimed to be a cloud company,” Radhakrishnan says. “Today, every company claims to be an AI company. We must cut through the noise to understand whether AI is core to their mission or just a feature.”

By 2023, the challenge of selecting AI investments only intensified as capital poured into general-purpose foundation models, like ChatGPT. “The space was quite crowded and even early-stage Series A funding rounds got quite expensive,” she says.

Looking for a promising but less saturated niche and one that aligned best with Hitachi’s heritage in operation technology (OT) and information technologies (IT), as well as its deep expertise in AI, Hitachi Ventures turned its attention toward AI for industrial and physical environments.

Gaining originality in modeling

One of its early investments is Archetype AI Inc., a Palo Alto startup that is building a foundational AI model that interprets data from sensors – including sound, vibration, temperature, and pressure – to perceive, understand, and reason about the physical world. The company’s ultimate goal is to encode the entire physical world, which would enable it to predict equipment failures, optimize industrial processes, and create digital twins of real-world operations.  

Hitachi backed Archetype in December 2023, just seven months after the startup incorporated – an unusually early stage for venture investment – and well before physical AI started to become a mainstream investment. The fact that Archetype AI was pushing boundaries made it both promising and risky, a common balancing act for industrial AI investors.

“What Archetype is doing is squarely in our thesis, but there was nobody else doing what they were doing at that time,” Radhakrishnan says. “We were a little uncomfortable because of that, but we were also comfortable with being uncomfortable. Sometimes the investments that could be the big winners have no precedent.”

As a corporate venture fund, with Hitachi as its sole limited partner, Hitachi Ventures operates with twin objectives. “Our first responsibility is to deliver strong returns,” Radhakrishnan explains. “But we also want to create a broad, strategic advantage for Hitachi.”

This means the firm’s deep dives into emerging technology spaces serve multiple purposes, including identifying investment opportunities while also educating the broader Hitachi organization about market trends. “Startups are often early indicators of technology waves,” she notes. “They’re a lighthouse that illuminates what’s coming in the distance.”

Achieving cognitive resonance

Hitachi Ventures’ investment in Xaba Inc., illustrates this dual approach. The Toronto-based startup develops cognitive control systems for robotics, enabling machines to respond intelligently to their environment in real-time. Traditional robots are pre-programmed for specific tasks; if they encounter an obstruction, they either stop or push through it. Xaba’s xCognition technology, which combines physics-based modeling with AI learning, enables robots to perceive obstacles and automatically adjust their path, essentially acting like a “brain” for the robot. This allows the robot to reason, adapt, and generalize across tasks.

What convinced Radhakrishnan wasn’t just the technology, but the Xaba founder’s depth of expertise, one of the many nuances that goes into funding decisions. “VCs are jacks of all trades who know everything, but who only know it an inch deep,” she says. “The founder comes with a strong technical background combined with industry knowledge, and he understands his space really well.”

When Radhakrishnan introduced the xCognition technology to counterparts at Hitachi Rail Ltd., they immediately saw its value. The Hitachi subsidiary, which operates in more than 50 countries, is now deploying Xaba’s robots for precision machining on the surfaces of locomotives. These tasks require sub-millimeter accuracy that previously demanded extensive manual labor.

But the Xaba team didn’t stop there. While AI co-pilots are all the rage for coding, Xaba has developed an automated code generator for Programmable Logic Controller (PLC), called PLCfy, that enables democratization of industrial automation. PLCfy provides a drop-in AI layer that augments existing PLCs with modern capabilities, like predictive control, anomaly detection, and adaptive optimization, without ripping and replacing hardware.

The flip side of consensus 

Yet while Xaba seems like the ideal Hitachi Ventures’ investment, it also demonstrates how challenging the industrial AI space is: Initially, the partners in the firm’s investment committee couldn’t agree whether it was worth pursuing.

“Sometimes when you can’t get consensus, those are the deals where you think, ‘Maybe there’s something there,’ ” she says. “Who would have thought that an online company selling books would redefine global compute needs? Or that a social media company connecting friends would impact enterprise storage buying behavior? If the impact of an AI startup is obvious to everyone, you’re probably not investing in the next Google, Amazon, or Facebook.”

# # #

Gayathri Radhakrishnan is a partner at Hitachi Ventures. With more than $1B AUM, the company is setting new standards for corporate ventures, fostering partnership and access for ambitious founders transforming the world around us. From advanced AI and robotics to sustainable energy solutions and beyond, Hitachi Ventures sees the potential in investing in companies that dare to dream big and disrupt the status quo. The firm’s expertise, coupled with Hitachi’s global resources and commitment to innovation, enables it to identify and nurture promising startups with the potential to drive significant impact and shape the future of technology.

For more, visit Hitachi Ventures.