For my father,

who was born into a world of oil lamps and bullock carts,

and has lived to see machines that think.

He did not need to understand the technology to understand what mattered most: that power always moves — and the wise know where it is going before the powerful notice it has left.

At ninety-four, he has outlasted empires, currencies, ideologies, and now perhaps the very idea that only humans can think. This book is my attempt to make sense of what comes next — written in his honour, and in his spirit.

献给亲爱的父亲大人

他生于油灯与牛车的年代，却亲眼见证了会思考的机器降临人间。

权力从不停歇——智者在权贵察觉之前，已知晓它的去向。

— 陳德強教授博士 · Prof Dr Tan Teik Kheong
IEEE · 2026

Why Power, Not Technology

Every era has a technology that rewires the global economy. Steam rewired manufacturing. Electricity rewired cities. The internet rewired commerce and communication. Artificial intelligence is rewiring everything simultaneously — and doing it faster than any previous technology transition in recorded history.

But this book is not primarily about the technology. It is about the power. Specifically, about how economic and strategic power is transferring — systematically, structurally, and largely invisibly — across the entire AI ecosystem. From apps to infrastructure. From software to compute. From consumer brands to hidden infrastructure owners. From public markets to private capital.

Throughout this book, seven questions recur. They are the analytical framework that cuts through the noise: Who owns the layer? Who pays whom? Where does the money flow? What is the bottleneck? What is the hidden dependency? What happens to retail investors? And what changes globally — and for whom?

Ask these questions consistently, about every AI announcement you encounter, and the power shift comes into sharp, uncomfortable focus. The answers are rarely the ones that the press releases suggest.

The Day Apple Borrowed a Brain

In June 2024, Apple stood on stage at its annual Worldwide Developers Conference and announced a partnership with OpenAI to bring artificial intelligence to its products. The crowd applauded. The headlines celebrated. Nobody asked the obvious question: why couldn't Apple build this itself?

Apple's market capitalisation hovered around three trillion dollars. It had its own custom silicon — the M-series chips that had outpaced Intel's roadmap by years. It had cash reserves exceeding the GDP of most countries. And yet, when it came to the most important new capability in consumer technology, it had to go to a startup founded in 2015 and ask to borrow its brain.

The public conversation about artificial intelligence is almost entirely focused on the wrong things. We talk about chatbots. We talk about whether AI will write our emails, replace our doctors, or take our jobs. Meanwhile, the actual power shift is happening somewhere most people never look — in data centres the size of football stadiums, in semiconductor fabs in Taiwan that are the single most strategically important buildings on earth, and in private capital markets where sovereign wealth funds are writing cheques that make venture capital rounds look like rounding errors.

Apple had the money. It had the engineering talent. It had the distribution. What it did not have was the accumulated expertise in transformer architectures, the proprietary training data at the required scale, and the institutional knowledge of how to manage the catastrophically expensive process of training and iterating on frontier models. These things take years to build, not quarters.

The deal Apple struck with OpenAI is a microcosm of the broader AI power structure. A consumer-facing brand with enormous distribution, dependent on an infrastructure-layer company for the intelligence that makes its products work. The brand takes the credit. The infrastructure takes the rent.

To understand the full implications, think of intelligence as a stack with five layers: raw compute at the base, training infrastructure above it, then training data, then the model itself, then deployment infrastructure at the top. Apple could not build layers one through four quickly enough to matter. So it licensed the output of those layers from the company that had built them, and added its own contribution at layer five.

Every organisation that has made a similar choice — which is to say, most organisations that have adopted AI capabilities in any serious way — has created a similar dependency. And the sum of those individual dependencies is the infrastructure of the AI power shift: the concentrated ownership of the intelligence stack in a small number of organisations that now have leverage over an enormous proportion of the world's AI-dependent workflows.

There is a concept in finance called sovereign debt — the obligations a nation owes to external creditors that constrain its policy choices in a crisis. The AI era is creating a form of sovereign technological debt that operates by similar logic. Nations that have built their critical digital infrastructure on platforms and services they do not own have created obligations to external providers that constrain their policy choices in ways they may not fully recognise until a crisis forces the constraint into view.

The June 2026 kill switch was that crisis for much of the developed world. The lesson was available earlier — in the Huawei entity list episode of 2019, when every telecommunications operator that had built networks on Huawei equipment had to reconsider its infrastructure. The world chose not to learn it then. The 2026 episode was considerably harder to ignore.

This book is organised around seven questions: Who owns the layer? Who pays whom? Where does the money flow? What is the bottleneck? What is the hidden dependency? What happens to retail investors? And what changes globally — and for whom? The answers are surprisingly concentrated, and they point consistently toward the infrastructure layer as the location of durable power in the AI era.

The New AI Empire

There is a question nobody asks at AI conferences. Speakers present stunning benchmarks. Investors explain why this cycle is different. And nobody asks: who actually owns this? The new AI empire is not owned by the companies whose logos appear on the products. It is owned by a smaller, quieter group who positioned themselves in the infrastructure layer before the consumer race began.

At the base of the stack sits NVIDIA — a company whose position has no historical precedent. In 2025 it generated $130 billion in revenue, up from $27 billion two years prior, at gross margins exceeding 57 percent. Every major AI lab trains on NVIDIA hardware. The company is the landlord of the AI economy, collecting rent from every competitor in the model race.

Microsoft's $13 billion investment in OpenAI, structured as Azure compute credits, was a masterstroke of infrastructure lock-in. Amazon's commitment of up to $20 billion in Anthropic followed identical logic. These are not charitable investments. They are long-term lease agreements: the model company gets capital, the hyperscaler gets a guaranteed, high-volume tenant paying cloud fees for years.

China's AI situation in 2026 is a strategic paradox: the world's largest population of AI users, constrained by export controls that deny access to the most advanced hardware. But the export controls produced unintended consequences. DeepSeek's R1 model demonstrated that frontier-level reasoning capability could be achieved with restricted hardware using novel efficiency techniques — innovation under constraint that embarrassed the assumption that hardware access was the binding constraint on AI capability.

The strategic implication for the rest of the world: the binary choice between American and Chinese AI dependence is becoming a more nuanced landscape. Open source models from Chinese labs are available globally, not subject to American export controls, and can run on domestically owned infrastructure. The kill switch, designed to protect American strategic interests, inadvertently made the case for Chinese AI sovereignty to every government in the world.

The investors who captured Anthropic's 235-fold valuation increase were not retail investors. They were Google, Amazon, and a small group of institutional players. By the time retail investors could buy shares on a public exchange, the extraordinary returns had already been made. The IPO is not the beginning of the opportunity. It is the closing of the first chapter.

Models Are Not Enough

There is a persistent fantasy in the AI industry: a brilliant team trains a model so capable that it disrupts everything. Intelligence wins. Infrastructure is just a detail. This fantasy has produced some of the most expensive failures in technology history.

Every frontier AI model company faces five structural dependencies: compute, energy, data, capital, and talent. Training a frontier model costs between $50 million and $500 million. Data centres will consume 950 terawatt-hours by 2030 — equivalent to Japan's entire national electricity consumption. OpenAI lost an estimated $5 billion in 2024 on $3.7 billion in revenue. The researchers who can advance frontier models number in the hundreds globally.

Of all the dependencies, talent is the most structurally significant in the long run. The concentration of AI research talent is extraordinary — perhaps in the low thousands globally, concentrated in San Francisco, London, and a handful of academic centres. Senior researchers command total compensation of $1 million to $5 million per year. The most sought-after can negotiate equity positions worth tens or hundreds of millions.

DeepSeek's demonstration that frontier-level results can be achieved despite hardware constraints proved that the talent dependency, not the compute dependency, may be the binding constraint. Innovation under constraint produced approaches the unconstrained American labs had not prioritised.

Understanding what models cannot do is as important as understanding what they can do. The most fundamental failure mode is hallucination — producing fluent, confident-sounding text that is factually incorrect. Hallucination is not a bug that can be fixed by making the model bigger. It is a consequence of the fundamental architecture, which predicts the next token without any intrinsic mechanism for verifying whether predictions correspond to facts.

Context sensitivity, opacity, and knowledge currency compound the challenge. These failure modes are reasons to adopt AI with clear eyes, designing workflows that put model capabilities to work where failure modes are least consequential — not reasons to avoid AI adoption entirely.

The Hidden Toll Roads

In the California gold rush of 1849, most miners did not get rich. The people who got rich sold them shovels, jeans, food, and financing. The AI era is following the same pattern — except the toll roads are harder to see, the tolls are being collected at every layer of the stack simultaneously, and the scale of the extraction dwarfs anything the gold rush could have imagined.

NVIDIA supplies chips to OpenAI, Anthropic, Google, and every lab that can access its products. The harder the model companies compete, the more chips they buy. TSMC in Taiwan manufactures those chips — the single greatest structural vulnerability in the global AI economy. No TSMC, no advanced AI chips. No advanced AI chips, no frontier models.

Microsoft signed a deal to restart the Three Mile Island nuclear plant to power its data centres. Amazon acquired a data centre campus co-located with a nuclear plant in Pennsylvania. Google signed long-term power purchase agreements with small modular reactor developers. The energy toll road is the least glamorous and potentially the most durable position in the entire AI investment landscape.

Electric utilities, nuclear operators, construction companies, data centre REITs, fibre connectivity providers — the AI build-out is touching every sector that provides the physical substrate on which digital infrastructure is built. The companies that own energy infrastructure in markets near major data centre clusters are seeing demand for power that exceeds their previous growth forecasts by multiples, not percentages.

Data centre REITs that built facilities in Data Centre Alley in 2015 or 2020 are collecting rents from tenants who could not replicate the infrastructure at any price in 2026. The toll road has become, in the most literal sense, a moat. You do not need to pick a winning model. You need to own the land the winner's data centre sits on.

The Kill Switch: Who Controls Your AI?

On June 12, 2026, Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei ordering that Fable 5 and Mythos 5 be subject to export controls for all foreign nationals. Because American 'deemed export' rules treat access by a foreign national as equivalent to an export, the company had no choice but to disable both models for every user on earth.

Every hospital that had integrated Fable into its diagnostic workflow. Every law firm that had automated document review with Mythos. Every defence contractor, every government agency — all discovered in a single weekend that their infrastructure could be switched off, not by a cyberattack, but by a letter from a cabinet secretary in Washington.

To understand the architecture, trace the dependency chain: end user → deploying organisation → software vendor → cloud platform → AI model company → American government. The architecture of dependency made the kill switch not only possible but essentially inevitable once the conditions for its use were met. Every step in the chain was individually rational. The aggregate result was a systemic vulnerability that no single actor could see clearly or had the incentive to correct.

America's strategy of restricting advanced AI chips to China appeared decisive. The reality was more complex. DeepSeek's R1 demonstrated that frontier-level capability could be achieved with restricted hardware using novel efficiency techniques. The lesson that containment strategies have been teaching since the nuclear era repeated itself: restricting a capability can accelerate the development of that capability in the restricted party.

AI sovereignty exists on a spectrum from complete dependence to complete independence. The practical objective for most nations is to move meaningfully away from complete dependence. The first step is inventory: understanding what AI systems your critical functions depend on and where those systems' key components are controlled. The second is prioritisation. The third is diversification. The fourth — the longest-term and most resource-intensive — is domestic capability investment.

French Prime Minister Lecornu: 'We cannot rely on tools developed by foreign powers.' Canadian Prime Minister Carney: 'Nobody has done anything wrong. But we will have done something wrong if we just accept this.' The map matters. But first you have to read it.

The Labour Bargain Breaks

For two centuries, the labour bargain has been the foundational compact of industrial society: technology makes workers more productive, workers command higher wages, higher wages support consumer demand, demand drives growth. The AI era is testing whether that bargain still holds — and the early evidence is deeply uncomfortable.

OpenAI's research on Codex (June 2026) documented what the agentic transition looks like inside the organisation that built the tools. Within one year, Codex became the primary tool for every department — Legal, Finance, Recruiting, not just Engineering. Seventy percent of users assign tasks exceeding one hour. Twenty-five percent assign tasks exceeding eight hours. Over a quarter of work done by Finance and Legal staff is engineering or coding — tasks entirely outside their training.

Karpathy coined 'vibe coding' in early 2025 — describe what you want, accept what comes back. He declared it obsolete a year later, replacing it with 'agentic engineering': orchestrating agents with oversight rather than writing code. He then joined Anthropic, stating the model is the bottleneck on what an agent can do. The XDA benchmark confirmed: Claude Code outperformed rivals not on functionality but on UX judgment — understanding design intent, not just syntax.

The professional contract — invest years acquiring expertise, receive compensation reflecting its scarcity — is being disrupted by AI reducing the effective scarcity of expertise. The tasks where AI is most capable are precisely those where the risk of hallucination means human oversight remains essential. The cost savings from AI are real; the human cost of providing oversight is also real; and the net impact on employment is less dramatic than either enthusiasts or alarmists claim.

Who Gets Left Behind — and Who Decides

Nations without chips. Companies without data. Workers without skills. Investors without access. The AI power shift produces winners at extraordinary speed — and produces a permanent underclass of the unprepared at the same speed. But the question of who gets left behind is not only economic. It is geopolitical, generational, moral, and — most importantly — a matter of deliberate choice rather than inevitable fate.

By mid-2026, the United States and China controlled approximately 90 percent of global AI computing power. The next tier — UK, Canada, France, Germany, Japan — had research capability but not infrastructure sovereignty. The tier below — Southeast Asia, South Asia, Africa, Latin America — was consuming AI products built entirely on infrastructure they did not own, running on models trained primarily on data that underrepresented their languages and realities.

The nations left behind in AI capability are not merely economically disadvantaged. Anthropic's Mythos model demonstrated the ability to discover thousands of critical vulnerabilities in every major OS and browser, completing autonomous end-to-end attacks on enterprise networks. The time from vulnerability discovery to active exploitation collapsed from 2.3 years in 2018 to 20 hours in April 2026. The most powerful AI is simultaneously the most capable offensive cyber weapon ever built — and the only meaningful defence against it at scale.

Technologies that produce very concentrated benefits and broadly distributed costs are politically fragile. AI is following a pattern that looks more like nuclear than like the automobile: financial benefits accreting overwhelmingly to infrastructure owners, productivity benefits distributed unevenly among knowledge workers in high-income countries, costs — displacement, deskilling, cybersecurity risk, environmental impact — broadly distributed and falling most heavily on communities least positioned to benefit.

The concentration of AI infrastructure in a small number of American companies is likely transitional. The opacity of AI systems is likely transitional. The displacement-without-replacement dynamic is also likely transitional. What is not transitional is the shift in the architecture of cognitive capability — the way AI systems have become a permanent part of the infrastructure of thinking, of decision-making, of knowledge work. The integration of AI into human processes is a one-way door.

Which future we inhabit — broadly distributed human flourishing, or concentrated AI-enabled advantage producing instability — will be determined by the choices that are being made now. These choices are not predetermined. They are being made, continuously and consequentially, by a distributed set of actors who mostly do not frame their choices as contributions to the long-term architecture of the AI era. Framing them as such is not idealism. It is strategic clarity about what is at stake and who gets to decide it.

The Map Is Yours

The AI power shift is not a future event. It is a present reality, unfolding faster than most people have registered, with consequences already visible in the balance sheets of the world's largest companies, the geopolitical calculations of the world's most powerful governments, and the career trajectories of the world's most educated workers.

First: power in the AI era lives in the infrastructure layer. The companies that own the chips, the energy, the cloud, and the real estate collect rent regardless of which model wins.

Second: model companies are structurally dependent on infrastructure they do not own. The hyperscalers have more leverage than they appear.

Third: private capital captured the extraordinary returns before retail investors could participate. The IPO wave of 2026 is the end of the first chapter, not the beginning.

Fourth: the kill switch is real. Any nation or institution that has built critical infrastructure on another nation's AI has accepted a dependency with an eviction clause. June 2026 was a warning. It will not be the last.

Fifth: the labour bargain is being renegotiated at machine speed. The question is whether you are learning to orchestrate agents or being orchestrated by them.

Sixth: AI-enabled cyberwarfare has crossed a threshold from theoretical to operational. Every organisation not rethinking its security posture is operating on borrowed time.

Seventh: the distribution of AI's gains is a choice, not a destiny. It is being answered right now by decisions that could be made differently.

— Prof Dr Tan Teik Kheong · 陳德強教授博士
IEEE · 电气电子工程师学会 · 2026

Prof Dr Tan Teik Kheong is a technologist, academic, and strategist with forty years of experience across engineering, policy, and global industry.

His professional journey began with a defining chapter in the early history of broadband networking. As Senior Vice President representing Asia at the global ATM Forum, he was instrumental in bringing broadband infrastructure to scale across the Asia-Pacific region, establishing himself as one of the foremost evangelists of the broadband era that laid the groundwork for the modern internet economy.

He subsequently served as Chair of the IEEE 802.11 Wireless Next Generation Committee, contributing to the foundational standards that gave the world Wi-Fi. His corporate career has taken him through Cisco, NXP Semiconductors, Accenture, FINRA, and IMDA, where he shaped digital policy at a national level. He has taught at Stanford University and MIT, and has collaborated with Oxford.

His contributions have been recognised through his nomination to the International Who's Who of Networking, his designation as an ATM Forum Ambassador, and his appointment as an Internet Society (ISOC) Ambassador. The AI Power Shift is his third book.