Inside the Belly of the Physical AI Beast

The Italian flag outside is misleading. INNOVIT HW & Robotics Program — it’s pure Silicon Valley.

A brief note before we begin: this article is long — because it had to be.

1. HW & Robotics — The Turn: Why Physical AI Became Impossible to Ignore

Silicon Valley is back to building things that weigh, that move, that take up space. Things that must endure tolerances, friction, gravity.
In his latest keynote, Jensen Huang spoke almost exclusively about this: machines, factories, robots, physical processes. He called it Physical AI.

Not a new term on these pages, but Jensen… he frames it as the ultimate business for NVIDIA. One word formed in my mind: “Finally.”

In 2008 I founded my first IoT startup. In 2016 I wrote a book about it.
It was a brutal journey: like working at the edge of a future that was taking shape, but — as Gibson would say it — simply wasn’t distributed yet where I was.

Then IoT grew arms, legs, wheels. And eventually a mind: AI.
The acceleration hit abruptly.

Take MultiplyLabs, for example.
When Fred Parietti (Founder & CEO) asked me to coordinate the effort to improve the physical design of their pharmaceutical robot, the context was precise: a technical cell, a defined process, a clear boundary. The work was solid — ambitious in the right measure. It even earned a GOOD DESIGN AWARD. A nice signal, but not the point.

The point came later. Two years later, Fred stepped onto the NVIDIA main stage.
And not as a PR stunt — as a sign of the times. A company with fewer than one hundred people had expanded and strengthened its core product:

• using AI to simplify and automate workflows once managed by customers
• introducing humanoid robots that extended automation well beyond the sterile cell where everything began

These are the foundations that transform an entire industry. This acceleration is formidable — and widespread. Physical AI is happening now, across a thousand industrial micro-segments. (I am so proud of you, guys!)

Weeks Just weeks later, a company I’ve followed since day zero — and a mentor in past cohorts — closed a $35M Series A in Physical AI. Kudos!

I had this scenario in mind when INNOVIT — for the third time — involved me in designing the Hardware & Robotics cohort. An edition where we didn’t want to explain the shift, but build an environment where ignoring it would be impossible.

A place to learn from the inside.
Inside the belly of the system.
Inside the belly of the beast.

⸻

2. The New Industrial Mind — A Cultural and Technological Shift

Italy is a country that builds real things: cars, ships, industrial robots, sensors, space components. A deep industrial chain, often made of small actors by American standards, yet critical inside global supply networks.

Just consider that none of the Teslas driving around Palo Alto could run without a couple of components made in Catania — in my faraway Sicily.

INNOVIT Hardware & Robotics startups are born from this DNA.
Small in size, inventive by instinct.

In Italy we design robotic hands, collision-avoidance systems for cobots, long-range drones, marine AI, critical components for every kind of industry.

The technical talent is there. But scaling hardware is something else entirely.

It requires a mindset that blends engineering, design, software, operational pace, and the ability to communicate value — even to those who don’t immediately see the elegance of a well-architected system.

This is why we created a place where founders could stretch their product mentally and imagine it as part of a system. A flexible, adaptive system that cannot be built unless software drives the atoms. Software-Defined Architecture became our mental map. Not a methodology, but a way of widening the field:

• making software speak to metal
• making mechanical engineering converse with the cloud
• designing hardware with its future scalability already in mind

This is where the two halves of the world come together.

In Silicon Valley all this is taken for granted — but the opposite is also true: Silicon Valley produces very little hardware.

The gap between thinking and manufacturing is enormous. You go from a maker room to a Chinese factory with almost nothing in between.

Italy sits in the complementary position.

We have mechanical engineering, material science, companies that know how to turn a concept into a production line. But historically, we lack the systemic, narrative, vector-driven mindset that Silicon Valley assumes by default.

This is where the relationship becomes clear:

Physical AI doesn’t originate from one of the two poles.
It emerges when they interlock.

SV brings the software-first mindset.
Italy brings the body, the matter, the factory.

And in between lie design, systems architecture, and the ability to scale what today exists only as a prototype.

This is the flow we trained inside INNOVIT.
And when this mental shift happens, everything else changes with it.
This is the New Industrial Mind.

⸻

3. Robotics Conversation.
Two weeks inside C15 aren’t a program: they’re a force field.

Founders’ mind bend direction before you even notice.
Almost everything happens behind closed doors. Close-range sessions, 1-to-1s, candid conversations with the people who actually build Silicon Valley.

It isn’t a school. It’s a practice ground.
In the morning you talk with someone who put a device into millions of hands.

In the afternoon you pitch to an engineer who dismantles your architecture in front of you. The working tables change like a jazz lineup:

Google next to an automotive company, a founder who exited next to one whose startup crashed. You’d bring one question in the morning and end up with ten — all sharper.

Later we open the doors…

Embodied Intelligence Event (TechCrunch 2025)

The scheduling overlap with TechCrunch opened a door: creating an official side event. The theme: Embodied Intelligence. Natural for anyone from interaction design; less common in mainstream vocabulary.

On stage, three perspectives performing like instruments in the same band:

• Marco Angelici (ST) — silicon that senses
• Andrea Esposito (ex Tesla Optimus) — mechanics that become gesture
• Kiril Stoynov (NVIDIA) — where AI finds power and body

The room was packed.
And most importantly: 81% of attendees had never been inside Casa Italia before.

The inevitable question arrived: “Why give a robot a human form?”

The panelists answered clearly:
it’s not a fetish.
The world is built around the human body.
Hands are the unit of measure of our environment.

I added only one line:
We — humans — are the ultimate interface.

VIP Panel @DemoDay Event. The moment where founders cross from practice into reality.

On stage, four architectures of the robotics system:

• Francesco Brianti — STMicroelectronics | silicon that perceives, processes, safeguards
• Peter Marcotullio — SRI International | the trajectory that brings research to market impact
• Stefano Landi — AWS IoT & Physical AI | the hybrid cloud–edge geometry made real by billions of connected devices
• Modar Alaoui — Silicon Valley Robotics & Humanoid Summit | the aerial view of a fast-forming ecosystem

The synthesis I offer the audience is this:
We must prepare for a form of inclusivity that stretches everything we know — not only humans of different cultures working together, but humans, robots, and AI agents sharing the same operational space.

This isn’t futurism.
It’s a transition already underway.

4. CLOSING — Founders: Who’s Already Building the Future

When the lights go down at the end of the program, the real magic is usually this: awareness — and a group.
Startups leave more conscious of the challenge, but also more confident in what they know, in what they’ve refined, and above all: aware they’re not alone.
They’re each on a trajectory.

My job is helping them navigate it — each in their own way.
Here are the ones I want to highlight:

Ganiga, for example.
One of only three Italian startups brought to San Francisco by TechCrunch, already INNOVIT alumni: AI and robotics to sort waste at the point of generation, while producing geo-positioned analytics on consumption. Useful to brands. Valuable to public spaces. And above all: a missing piece of urban infrastructure. They’ve sold hundreds of robots — even to Google. Nicolas Zeoli is probably Italy’s most unpredictable founder — in the best possible sense.

Then there’s Mediate, working on the future of human–robot coexistence and collaboration. At the heart of Mediate’s platform is a powerful real-time perception system that precisely detects human presence around industrial machines and robots — distinguishing operators from inanimate objects, dynamically adapting to movement, and ensuring safety without slowing down productivity.
It’s the kind of enabling technology that quietly unlocks entire categories of automation — the “infrastructure layer” most people never notice, but every factory (and every human–robot interaction) desperately needs.

Then there’s ViberAlert.
A founder I had met before — in another INNOVIT program, at CES, and around the ecosystem — who then completely reinvented himself. Vincenzo Maresca’s team, boosted by investors who later became true partners, quite literally blossomed inside INNOVIT. Their technology saves motorcyclists’ lives. They have traction, major customers, and a real U.S. presence. A rare example of what maturing truly means — something far deeper than simply growing.

Then Dgree, bringing AI to open water.
A story that started in an INNOVIT cohort supported by Regione Toscana.
If yachting has a strong Italian footprint worldwide, this company — led by Giovanni Palamà — has built the only deterministic marine AI trained on fundational data collected through real navigation. They’re uniquely positioned to become the “AI of the seas.” Here, Physical AI reaches its deepest meaning: watch a robot cross the ocean alone — no network, no cloud, no margin for error — and you immediately understand what it means to design for an environment that does not forgive.

Cognivix is the most “American” of the group. They build micro-intelligence units that transform manufacturing: systems that learn by observing, correct errors in real time, and intervene locally without waiting for a server thousands of kilometers away. Their California team, led by Daniele Bernardini, has already delivered major projects and is now supporting U.S. reshoring initiatives.

and there is much more…
Across the three years I’ve led INNOVIT’s deep-tech programs, companies like Oversonic — the only Italian-made anthropomorphic cobot — or competitors Prensilia and BionIT Labs passed through. Italy’s excellence in robotic hands is real — and BionIT is already American.

Mature companies like Rain and Macnil, or those already strongly established in the U.S. like Icop, Galdi, SIR, and many more.

All share one thing:
They experienced momentum together.
They sharpened their ambition.
They built concrete tools to grow in America.

That’s what my work (also) served — and I’m proud of them.
Now, back on “my Yacht”.
because someone has to test this marine AI, right?