A $1 Trillion Order Book and a Palo Alto Living Room: What GTC 2026 and Hard Things Taught Us This Week About Physical AI
Co-authored by Vitaly Golomb
Palo Alto, CA: Nvidia just disclosed a $1 trillion order book. Uber committed to deploying robotaxis in 28 cities. And in a Palo Alto living room on Thursday, a founder with 20 years of operational scars told a group of investors and builders the one thing that will determine who actually captures value in this market.
This was the week Physical AI stopped being a thesis and became a capital allocation decision. Here is what we learned — and what it means if you are building, funding, or advising companies at this frontier. We live in a world defined by uncertainty. We have to be flexible in everything we do.
Last Monday, Jensen Huang took the stage at the GPU Technology Conference (GTC) at SAP Center in San Jose in front of 10,000 attendees — with 30,000 more watching from 190 countries — and laid out a hardware and software roadmap that redefines what is possible in AI infrastructure. Then, on Thursday evening, we co-hosted an intimate gathering of founders, investors, and builders in Palo Alto through Mavka Capital and Foley & Lardner’s new invite-only series, called “Hard Things.” A single speaker, Veena Radhakrishna of Cartesian Kinetics, gave us the operator’s framework for what it actually takes to build and ship in this market.
Two events. Opposite scales. The same conclusion: Physical AI is no longer a research topic. It is an infrastructure buildout, a regulatory challenge, a capital formation opportunity, and a company-building discipline — all at once. And the teams that get the architecture right, from their tech stack to their cap table, will define the next decade.
Nvidia reported $215.9 billion in fiscal year 2026 revenue. Quarterly data center revenue: $62.3 billion. Eleven consecutive quarters above 55% growth. And the headline number: at least $1 trillion in high-confidence purchase orders for Blackwell and Vera Rubin systems through 2027 — double the $500 billion Jensen projected just one year ago at GTC 2025.
That is not a forecast. It is an order book. The companies writing those checks, from hyperscalers, cloud providers, sovereign AI programs to AI-native startups, have moved from exploration to industrial commitment. For founders raising capital in adjacent markets, this is the demand signal you underwrite against. For investors, it is the infrastructure baseline that makes vertical Physical AI plays investable at scale.
Jensen organized his keynote around five layers: accelerated computing, AI factories, open models, agentic systems, and physical AI. Nvidia’s strategy is to be present in all five simultaneously. There are different implications for every company building on this stack.
OpenClaw (the open-source agentic AI framework launched in January 2026) became, by some measures, the fastest-growing open source project in history. Jensen compared it to Linux, HTTP, and Kubernetes. It is one of those foundational standards that crystallized entire computing eras.
Nvidia’s enterprise layer is NemoClaw, a security-hardened stack that enables companies to deploy autonomous agents without exposing proprietary data. Jensen’s framing was direct: every SaaS company will become an Agentic-as-a-Service company. Every software business will rethink its product around autonomous, goal-directed agents. According to Jensen, Nvidia’s own engineers are already 100% on agent coding tools.
Between the two of us, we have advised on hundreds of SaaS transactions and capital raises in this Valley. The signal is unambiguous: if your software company does not have a credible agentic roadmap today, you are building a product whose market is evaporating. This is not incremental disruption. It is categorical. And the legal, governance, and IP frameworks around autonomous agents are being written now. The companies that get ahead of that curve, with the right counsel, will have a structural advantage.
What the Room Heard
What struck us watching the conversation unfold was how singularly universal Veena’s framework turned out to be, even in a room of people who span hardware, software, robotics, manufacturing, and finance.
The founders in the room were all nodding, not because they build warehouse automation systems, but because the underlying discipline she described, staying architecturally and strategically flexible while still committing to a direction, is the hardest and most important thing any deep tech founder has to do.
It’s easy to talk about pivoting. It’s much harder to build an organization that can actually pivot without losing its core identity, its team’s conviction, or its customers’ trust. Veena has done it repeatedly, over two decades, in one of the most technically demanding and capital-intensive domains in technology. That is worth paying very close attention to.
The Through-Line GTC to Hard Things
So, here’s what this week showed us, at every scale — from Jensen Huang’s arena keynote to Veena’s conversation in a Palo Alto living room.
Physical AI is not a trend. It is a platform shift. Jensen’s trillion-dollar order book and Uber’s 28-city robotaxi deployment are the demand signal. Veena’s twenty years of patient, flexible, compounding engineering work is the supply signal. The two are converging, and the window to get positioned at that intersection is narrowing faster than most people realize.
If you’re a founder, investor, or enterprise leader navigating this moment, the question is no longer whether to engage with Physical AI. The question is how to position yourself to move fast without making commitments you can’t undo.
Practical Implications
For the founders, investors, and advisors who read this column: If you are building in Physical AI, Veena’s framework is your operating manual. Design for retrofit. Choose the intelligence layer over the hardware layer wherever you can. Let your customers’ real environments teach you things your roadmap can’t predict. Stay flexible long enough to be around when the market clarifies, because in Physical AI, the market always clarifies later than the models predict and faster than the incumbents are ready for.
If you are investing in Physical AI, the GTC announcements are your market map. Nvidia has now laid out a clear full-stack platform (computing, open models, simulation, and networking) that will lower barriers and accelerate timelines across every vertical. The companies that figure out how to build on top of that platform, rather than competing with it, will be the ones generating outsized returns. Look for the Veenas of the world that are domain experts with deep operational knowledge, flexible architectures, and genuine customer traction. That combination, in a market this large and this early, is rare and enormously valuable.
If you are an enterprise buyer, the conversation is not about whether to deploy Physical AI. It’s about how to do it without making irreversible commitments prematurely. Veena’s retrofit approach is directly applicable to how you should think about integrating autonomous systems into your operations. Preserve optionality. Start with orchestration. Don’t rip and replace until you have real operational data.
If you need help navigating the legal, capital, and strategic dimensions of Physical AI — structuring deals, raising capital, or preparing for M&A — that is exactly what Foley & Lardner and Mavka Capital are here for. These are not simple transactions. They require advisors who understand the technology and the market. Reach out.
One More Thing
Hard Things is just getting started. Our first gathering was everything we hoped it would be — a room full of genuinely sharp people, a speaker who told the unvarnished truth about what it takes to build in this domain, and the kind of conversation that doesn’t happen at a 30,000-person conference.
We’ll be hosting more of these, with Mavka Capital and other partners, throughout the year. The series is simple: the most important insights at the frontier of Physical AI don’t come from keynote stages. They come from people like Veena, who have been in the trenches long enough to know what actually matters — and who are generous enough to share it.
Thank you to Kateryna Mamyko, Tiaan De Nysschen, and Sahar Mor for making Thursday’s evening what it was. And thank you to Veena for reminding us all that the hardest things are always worth doing — as long as you’re flexible enough to keep learning while you do them.
See you at the next one.
Further reading
- The Moat Is Dead. Long Live the Runner by Vitaly Golomb
- Software Valuations in Early 2026: The Age of Darwinism, and M&A Readiness in the Age of the “AI Apolcalypse” by Louis Lehot
- Blitzhires: The New Fast-Moving M&A Deal by Louis Lehot
About the authors
Louis Lehot is a veteran Silicon Valley emerging growth technology company lawyer at Foley & Lardner LLP. He advises founders, investors, and boards across the technology, life sciences, and clean energy sectors. Follow his musings on business and legal developments in the Global Silicon Valley on his Substack. Vitaly Golomb is the Managing Partner of Mavka Capital, an advisory investment bank focused on supporting deep tech entrepreneurs in raising capital and M&A, Managing Partner of Mavka Ventures, an early-stage embodied AI fund, and author of Accelerated Startup. Together, they co-host the Hard Things invite-only event series. For legal, fundraising, M&A, or strategic advisory on Physical AI and agentic AI, reach out to Louis or Vitaly directly.