NVIDIA … The Thinking Machine … exploring the genius of Jensen Huang, the growth of NVIDIA … and the transformation of AI into the defining industrial system of the 21st century
June 26, 2026
In 1993, I was working in Sunnyvale, in the heart of Silicon Valley.
One evening, three guys walked into a local Wendy’s burger restaurant and started talking about gaming, and their frustration with the limits of computing power. Why couldn’t tech companies create better chips? Why did they just serve the average productivity-seeking user, rather than hard-core users like them?
One of those 3 guys was Jensen Huang, a 30 year old mid-career chip engineer. Born in Taiwan, his family had emigrated to the US when he was 9, and he’d grown up in Oregon. They agreed to start their own business, focused on high powered chips. 30 years later, that business became the world’s first $5 trillion company.
The three founders were looking for a name which evoked speed, power, and desirability in graphics computing. They liked the word “invidia” which is Latin for envy. And from that burger, that frustration, emerged the (stylised in capitals) corporate name which now leads the world of technology, NVIDIA.
The Thinking Machine
Stephen Witt’s The Thinking Machine is one of the best narratives to explain the rise of NVIDIA and, more broadly, the transformation of artificial intelligence into the defining industrial system of the 21st century.
While it is structured as a corporate biography of Jensen Huang and NVIDIA’s evolution, its deeper purpose is to explain something far larger: how computing moved from being a tool of software companies into becoming the foundational infrastructure of the global economy.
At its core, the book is not really about chips. It is about the emergence of a new industrial stack, one in which compute power becomes the scarce resource that determines which companies, countries, and technologies can progress. In Witt’s framing, NVIDIA is not just a successful hardware company. It is the builder of the “thinking machine”: the distributed computational substrate that enables modern AI systems to exist at all.
The result is a story that sits at the intersection of entrepreneurship, semiconductor engineering, platform economics, and geopolitical strategy.
NVIDIA’s central idea: compute becomes intelligence infrastructure
Witt’s most important argument is that artificial intelligence did not emerge simply because of breakthroughs in algorithms or data availability. It emerged because of a parallel revolution in hardware—specifically, the GPU.
Originally designed for video game graphics, GPUs turned out to be uniquely suited for the kind of parallel computation required for deep learning. This accidental alignment between gaming hardware and neural networks created the conditions for the modern AI boom.
But Witt pushes the argument further. He suggests that once AI models began scaling, compute itself became the limiting factor of intelligence. The ability to train and run large models depends not just on clever software, but on access to vast, highly specialised computational infrastructure. From this perspective, NVIDIA did not just “win” a technology cycle. It became the gatekeeper of cognitive capacity at scale.
For CEOs, this reframes AI entirely. It is not a software category. It is an infrastructure dependency—similar to electricity in the industrial age or oil in the 20th century.
Jensen Huang and the philosophy of long-cycle thinking
A significant portion of the book is devoted to Jensen Huang, who has remained over three decades as the company’s long-serving CEO. Witt portrays Huang not as a conventional Silicon Valley entrepreneur, but as a leader shaped by long time horizons, technical obsession, and extreme resilience.
Unlike many tech founders who pivot frequently or chase market trends, Huang is characterised by consistency. NVIDIA’s strategy over decades has been remarkably stable: invest heavily in parallel computing architectures long before their commercial payoff is obvious.
Witt highlights a key leadership pattern: Huang repeatedly commits to architectures and platforms that take years—sometimes decades—to become fully realised markets. This includes the shift from gaming GPUs to general-purpose compute, and later to AI-specific architectures.
The strategic implication is profound. NVIDIA’s success is not the result of reacting quickly to AI. It is the result of anticipating a world in which parallel computation becomes the basis of intelligence itself.
For CEOs, Huang’s leadership model suggests that in deep technology industries, advantage accrues not to the fastest adapters, but to the most persistent system-builders.
The GPU revolution: from graphics to general intelligence
One of the most important narrative threads in The Thinking Machine is the transformation of the GPU from a niche gaming component into the central engine of AI.
Initially, GPUs were designed to render images for video games by processing thousands of small calculations in parallel. This architecture was ideal for graphical rendering, where many pixels must be processed simultaneously. However, researchers in machine learning discovered that neural networks also rely on parallel computation—specifically matrix multiplications across large datasets. This unexpected alignment meant that GPUs could accelerate AI training by orders of magnitude compared to traditional CPUs.
Witt emphasises that this was not a planned transition. It was a convergence of separate technological trajectories: gaming demand on one side, academic machine learning research on the other.
Once this convergence was recognised, NVIDIA began investing heavily in software ecosystems (particularly CUDA) to lock developers into its platform. CUDA effectively transformed GPUs from hardware products into programmable intelligence infrastructure. This shift is critical. It means NVIDIA is not just selling chips—it is selling a full-stack computational environment that defines how AI is built.
CUDA: the hidden moat
One of the most strategically important sections of the book concerns CUDA, NVIDIA’s proprietary software layer that allows developers to write programs for GPUs.
While hardware competitors can theoretically build similar chips, CUDA created a deep ecosystem lock-in. Over time, thousands of AI researchers and engineers built workflows, libraries, and systems around NVIDIA’s architecture. Witt describes this as one of the most powerful “invisible moats” in modern technology. It is not just technical superiority, it is ecosystem dependency.
For organisations trying to compete with NVIDIA, the challenge is not simply building better chips. It is recreating an entire developer ecosystem, which took more than a decade to mature.
For CEOs, this illustrates a broader principle of platform power: control of developer experience becomes control of the market.
The scaling laws and the AI demand explosion
Witt also situates NVIDIA’s rise within the emergence of scaling laws in AI research—the empirical observation that model performance improves predictably with increases in data, compute, and model size.
This insight transformed AI from a research domain into an industrial scaling problem. If performance improves with scale, then competitive advantage goes to the organisations that can deploy the most compute. This created an exponential demand curve for GPUs. Companies like OpenAI, Google, Meta, Amazon, and countless startups began competing for access to NVIDIA’s hardware.
Witt highlights a key structural shift: AI stopped being a marginal research activity and became a compute-hungry industrial process. This fundamentally changed NVIDIA’s position in the value chain. It moved from being a component supplier to being the central enabler of frontier AI.
The new industrial stack: chips, systems, and intelligence
One of the most important conceptual contributions of the book is its implicit mapping of the new AI industrial stack.
At the bottom layer are semiconductor fabrication processes, where physical constraints determine what is possible. Above that are chip designers like NVIDIA. Above that are system integrators building data centres. Above that are cloud providers. And finally, at the top layer, are AI model developers and applications.
Witt’s argument is that NVIDIA sits unusually close to the foundation of this stack, giving it disproportionate influence over everything built above it. This structure mirrors earlier industrial revolutions. Just as control of steel, oil, or electricity determined economic power in previous eras, control of compute infrastructure now determines AI capability.
For executives, this implies that competitive advantage in AI is not only about models or data, but about access to and control of computational infrastructure.
Supply chains and physical constraints
A major theme in the book is the physical reality behind digital intelligence. Despite AI often being framed as an abstract software domain, Witt repeatedly emphasises that it is grounded in extremely tangible constraints: fabrication plants, lithography machines, energy consumption, and global supply chains.
NVIDIA does not manufacture its own chips. Instead, it relies on TSMC in Taiwan, ASML in Europe, and a complex global network of suppliers. This introduces geopolitical fragility into the AI ecosystem. The entire global AI boom depends on a small number of highly specialised manufacturing nodes.
Witt uses this to highlight a paradox: AI is often described as dematerialised intelligence, but it is in fact one of the most materially constrained technologies in existence.
Nvidia as a platform company, not a hardware company
One of the most important reinterpretations in the book is that NVIDIA should not be understood as a semiconductor company in the traditional sense.
Instead, Witt frames it as a platform company for computational intelligence. This distinction matters. Traditional hardware companies compete on price, performance, and manufacturing efficiency. Platform companies compete on ecosystem control, developer lock-in, and network effects.
NVIDIA’s dominance is therefore not just technological. It is structural. The company has created a self-reinforcing ecosystem in which:
- developers build on CUDA
- researchers optimise for NVIDIA architectures
- cloud providers standardise NVIDIA hardware
- AI labs depend on NVIDIA GPUs for training
This creates a compounding advantage that is extremely difficult to dislodge.
The geopolitical dimension
Although not always framed explicitly as a geopolitical book, The Thinking Machine inevitably becomes one. NVIDIA sits at the centre of global competition between the United States and China over AI capability.
Access to advanced chips has become a strategic lever. Export controls, supply chain restrictions, and national AI strategies all reflect the recognition that compute is now a strategic resource.
Witt shows that NVIDIA’s position is unusual: it is simultaneously a private company, a global infrastructure provider, and a geopolitical chokepoint.
For CEOs, this raises a fundamental question: in strategically important industries, where does corporate strategy end and geopolitical exposure begin?
The economics of scarcity
A recurring insight in the book is that AI is defined not by abundance but by scarcity—specifically scarcity of compute.
Despite rapid innovation, demand for GPUs consistently exceeds supply. This creates pricing power, long waiting lists, and strategic allocation decisions by NVIDIA.
Witt highlights how this scarcity has reshaped the economics of AI development. Companies are forced to make trade-offs between model size, training time, and deployment scale based on compute availability rather than purely technical ambition.
This reinforces NVIDIA’s central position: it effectively controls the bottleneck of modern intelligence production.
The emergence of AI as industrialisation
Perhaps the deepest argument in The Thinking Machine is that AI represents not just a technological shift, but a new phase of industrialisation.
In earlier industrial revolutions, societies learned to harness energy (steam, electricity, oil) to amplify physical labour. In the AI era, societies are learning to harness compute to amplify cognitive labour. NVIDIA sits at the centre of this transition, providing the infrastructure for machine intelligence at scale.
Witt suggests that we are still at the early stages of this transformation. Just as electricity took decades to reshape economies, AI infrastructure will gradually reshape every industry—from finance and healthcare to manufacturing and logistics.
Strategic implications for CEOs
Although The Thinking Machine is not written as a business strategy manual, its implications for executives are clear.
First, AI should be treated as infrastructure investment, not software adoption. Competitive advantage will depend on access to compute, not just algorithms.
Second, platform dependencies matter. Organisations building AI systems are increasingly dependent on a small number of infrastructure providers, creating concentration risk.
Third, supply chain resilience is now part of AI strategy. Semiconductor geopolitics, energy availability, and hardware access are strategic variables.
Fourth, long-cycle thinking is essential. NVIDIA’s success illustrates the value of sustained investment in foundational technologies long before they become obvious winners.
Finally, AI should be understood as an industrial system rather than a tool. It requires rethinking organisational design, capital allocation, and operating models.
The machine beneath the intelligence revolution
The Thinking Machine ultimately reframes the entire AI revolution. It shifts attention away from applications and models and toward the underlying infrastructure that makes intelligence at scale possible.
Stephen Witt’s central contribution is to show that NVIDIA is not merely a successful company riding an AI wave. It is the builder of the wave’s physical substrate—the compute layer that enables modern artificial intelligence to exist.
The book’s lasting insight is simple but profound: every intelligence system has an industrial base. In the case of AI, that base is compute, and NVIDIA is its dominant architect.
For CEOs, the message is clear. Understanding AI requires understanding not just what it does, but what it is built on. And what it is built on is increasingly the most strategically important resource in the global economy.
More from the blog