★★★★½ 4.6/5 — A rare book that is both urgent and actionable. Suleyman writes with the authority of someone who built the technology and the humility of someone who knows it could go badly wrong.
Best for: Policy thinkers, technology leaders, concerned citizens, and anyone who wants to understand AI beyond the hype cycle.
Reading time: ~7 hours (320 pages) | Difficulty to apply: Moderate — ideas are accessible; the policy context requires some engagement.
The Coming Wave in one minute
Two of the most powerful technologies ever created are arriving at the same time, and we are not ready. Mustafa Suleyman — who co-founded DeepMind and helped pioneer modern AI — argues in The Coming Wave that artificial intelligence and synthetic biology are not simply the next step in technological progress. They are a wave of a different kind: general-purpose, rapidly self-improving, cheap enough for almost anyone to access, and capable of reshaping every industry, every government, and every aspect of daily life within a generation.
The central challenge Suleyman calls the containment problem: how do you capture the enormous benefits of these technologies while keeping them under meaningful human control? History offers a few successes — nuclear weapons were never widely proliferated, CFCs were eventually banned — but those successes relied on high barriers to entry and identifiable actors. The coming wave has neither. Suleyman does not predict doom. He maps a difficult but navigable path, laying out ten concrete steps that could, if taken seriously, allow humanity to pass through what he calls the narrow gate.
Key takeaways
- Two waves, not one: AI and synthetic biology are converging simultaneously, each amplifying the risks of the other — a lab that can design proteins can also design pathogens.
- Cost curves are decisive: DNA synthesis costs a millionth of what it did in 2000; AI compute follows a similar trajectory. Once a technology becomes cheap, containment becomes exponentially harder.
- General-purpose is the key word: Unlike a car or a hammer, these technologies do not do one thing. An AI system that writes code also writes disinformation; a DNA synthesizer that makes vaccines also makes risks.
- The containment problem is structural: The economic incentives to proliferate technology are enormous; the incentives to contain it are diffuse and long-term. No single actor can solve this alone.
- Previous containment succeeded for specific reasons: Nuclear weapons required nation-state-level resources; biological weapons conventions worked because verification was possible. Neither condition holds for AI.
- Nation-states are losing the capacity to govern: The largest AI companies now have more compute, more talent, and more global reach than most governments. The traditional regulator-regulated relationship has inverted.
- The omnicrisis is not one crisis but many: Suleyman identifies simultaneous risks in employment, political stability, military balance, and the reliability of information — crises that AI could trigger at once.
- Containment does not mean stopping AI: Suleyman is explicit — you cannot and should not try to halt the wave. The goal is to shape it: to build in accountability, auditability, and control from the start.
- Ten steps exist, and they are achievable: From safety research to choke points to civil society, Suleyman charts a concrete path that does not require global agreement on day one.
- Time is the most critical variable: The gap between when AI becomes highly capable and when governance catches up is the danger zone. The work must begin now.
What is The Coming Wave about?
The Coming Wave is about the simultaneous arrival of artificial intelligence and synthetic biology as transformative, general-purpose technologies — and the urgent question of whether humanity can contain their risks while capturing their benefits. Suleyman, who co-founded DeepMind, argues that these two technologies share characteristics that make them uniquely difficult to govern: they are rapidly cheapening, autonomously improving, dual-use, and general-purpose in ways no previous technology has been.
About the author
Mustafa Suleyman grew up in north London and studied philosophy and theology at Oxford before dropping out to help found a Muslim youth helpline. In 2010, alongside Demis Hassabis and Shane Legg, he co-founded DeepMind — arguably the world’s most influential AI research laboratory — which was acquired by Google in 2014. Suleyman served as VP of AI Products at Google before co-founding Inflection AI in 2022, creator of the Pi personal AI assistant. In 2024, Microsoft brought him in to lead Microsoft AI. His co-author Michael Bhaskar is a writer and technology analyst whose previous books examine the economics of creativity and curation. Together they bring both insider engineering knowledge and broad cultural perspective to one of the most consequential debates of our time. Explore all Mustafa Suleyman book summaries →
Key concepts at a glance
| Concept | What it means | Use it when |
|---|---|---|
| The Coming Wave | AI and synthetic biology arriving simultaneously as general-purpose, cheapening technologies | You need to grasp why this moment is historically unprecedented |
| The Containment Problem | The structural tension between incentives to proliferate technology and the need to control it | Discussing AI governance, policy, or corporate responsibility |
| The Omnicrisis | Multiple simultaneous societal disruptions — economic, political, military, epistemic — triggered by uncontrolled AI | Evaluating worst-case AI scenarios or stress-testing policy proposals |
| General-Purpose Technology | A technology (like electricity or the internet) that transforms many sectors at once, not just one | Comparing AI to past revolutions and explaining why AI regulation is harder |
| Choke Points | Physical or regulatory control points — GPU exports, model weights, talent pipelines — that can limit proliferation | Thinking practically about what governments can actually do |
| The Narrow Gate | Suleyman’s metaphor for the difficult but navigable path to safe AI — not utopia, not disaster, but hard-won equilibrium | Framing why effort now matters more than prediction about outcomes later |
Part 1: The Wave — What Is Coming and Why It Is Different
Suleyman opens by distinguishing the coming wave from every previous technology wave. Steam power, electricity, and the internet were transformative — but they did one thing at a time, required massive capital, and degraded in power as they spread. The technologies arriving now are the opposite. A large language model trained once can be deployed infinitely. A DNA synthesizer capable of making life-saving vaccines is the same device that could, in principle, make something far more dangerous. Cost curves are the engine of the wave: the price of sequencing a human genome fell from $100 million in 2001 to under $600 today. The price of AI compute is falling faster still.
What makes this wave qualitatively different is its general-purpose nature. When GPT-4 scored in the top 10 percent of bar exam takers, it did not become a lawyer — it demonstrated that the same system capable of legal reasoning is also capable of writing code, generating images, composing music, and planning logistics. AlphaFold solved a 50-year-old problem in protein structure prediction and deposited the results into a public database where any researcher, anywhere, could access them the next morning. This is the wave: fast, cheap, multi-domain, and self-reinforcing.
Part 2: Containment — Why It Has Worked Before and Why It May Not Work Now
The middle section of the book is its most intellectually rigorous. Suleyman surveys the history of technological containment: the 1987 Montreal Protocol that phased out CFCs and began repairing the ozone layer; the Nuclear Non-Proliferation Treaty and the decades of deterrence that (so far) prevented nuclear war; the Biological Weapons Convention that banned state-sponsored bioweapons programs. These were genuine successes. The question is: why did they work, and do those conditions still hold?
Suleyman’s answer is sobering. Containment worked when: (1) the technology required nation-state-level resources, limiting the number of potential bad actors; (2) verification was physically possible — you could count missiles, inspect facilities; and (3) the costs of defection were visible and immediate. None of these conditions reliably apply to AI. Building a competitive AI system today requires resources within reach of well-funded startups, not just superpowers. AI models, once trained, are software — they can be copied at zero marginal cost. And the harms from poorly aligned or misused AI are often diffuse, delayed, and invisible until they are not.
The deeper problem is structural. Nation-states, Suleyman argues, are being outpaced. The ten largest AI companies collectively spend more on research and development than the scientific agencies of most governments. They employ the majority of the world’s leading AI researchers. They have data centers in every jurisdiction and user bases of hundreds of millions. Governments regulate by writing laws; companies build by shipping products. The tempo difference is decisive.
Part 3: The Omnicrisis — What Failure Looks Like
Suleyman does not dwell on science-fiction scenarios of robotic uprisings. His concern is more immediate and arguably more credible: a world in which AI and synthetic biology arrive faster than institutions can adapt, producing not a single catastrophe but a constellation of simultaneous crises — what he calls the omnicrisis.
Economic disruption is the most visible dimension. Routine cognitive work — data entry, basic legal research, customer service, content moderation — is already being transformed. But Suleyman argues the deeper risk is to the social contract: if productivity gains accrue primarily to capital rather than labor, the political consequences could be severe in democracies that depend on economic consent. The political dimension is equally pressing. AI-generated disinformation operates at a scale and personalization impossible for human propagandists. Deep fakes of political figures are already being used in elections. The epistemic commons — the shared pool of agreed-upon facts that democratic deliberation requires — is under pressure.
Military AI adds a third layer. Autonomous weapons systems that can select and engage targets without human confirmation represent a categorically different risk than conventional weapons. The speed of AI-driven conflict could compress decision cycles to the point where meaningful human oversight is effectively impossible. And synthetic biology, running in parallel, means that the same decade that produces transformative AI medicines could also lower the barrier to engineered pathogens.
Part 4: Through the Narrow Gate — Ten Steps to Containment
The final section is the book’s most hopeful and most pragmatic. Suleyman rejects both techno-utopianism (AI will solve everything) and techno-pessimism (it is already too late). Instead he charts ten concrete steps — not a complete solution, but a framework for beginning serious work.
Safety research must become a first-class discipline, not an afterthought. We need to understand what AI systems are doing and why. Liberal democracies must lead because open societies have the most to lose from surveillance capitalism and AI-enabled authoritarianism. Corporate responsibility is not optional: companies building these systems must accept accountability for their downstream effects. International consensus — imperfect, slow, frustrating — remains the only viable path to preventing a race to the bottom on safety standards. Regulation and legislation must move faster; the EU AI Act and the US executive orders are starts, not finishes. Auditing and liability give regulation teeth: if companies are liable for harms, they have incentives to prevent them. Choke points and controls — export restrictions on advanced semiconductors, oversight of model weights, scrutiny of research pipelines — are the practical levers governments actually possess. Public investment in AI safety is essential; the field cannot remain exclusively corporate. Civil society and a free press provide the independent watchdog function that neither governments nor companies can provide for themselves. And finally, individual choices: how citizens demand accountability from the companies and politicians who shape AI policy.
Suleyman is careful not to promise that these steps will be sufficient. He closes with a characteristically honest assessment: the gate is narrow, the path is hard, and there is no guarantee of success. But he has walked through enough of this terrain to know the contours, and his map is worth having.
Who is The Coming Wave best for — and who should read something else first?
Read this if you work in technology, policy, or any field that will be materially affected by AI in the next decade — which is to say, almost any field. It is also the right book for concerned citizens who want to understand the AI debate beyond headlines. Leaders who shape organizational AI adoption will find Suleyman’s governance framework directly applicable.
Consider reading first: If you want a more technical grounding before the policy argument, our summary of Algorithms to Live By offers an accessible entry point to computational thinking. If you’re already deeply familiar with AI governance debates, Suleyman’s prescriptions will feel familiar, but his personal authority and the clarity of his prose make the book worth reading anyway.
Questions to reflect on
- What would “containment failure” look like in your life, your industry, or your country — and how would you know if it was happening?
- Do you trust corporations, governments, or neither to govern AI effectively? What would need to change for your answer to shift?
- Which of Suleyman’s ten steps do you think is most achievable in the next five years, and why?
- How has AI already changed your relationship with information, work, or community — and have those changes been net positive?
- If you were advising a government on AI policy today, what would be your single highest-priority recommendation?
🔥 Ready to understand the most important technology debate of your lifetime?
The Coming Wave is the clearest, most authoritative case for why AI governance matters — written by the person who built the technology.
How to apply The Coming Wave (7-day plan)
- Day 1 — Understand the wave: Read or listen to Part I of the book. Alternatively, watch Suleyman’s TED talk on the containment problem. Write down one area of your life where AI has already arrived.
- Day 2 — Map your AI exposure: List every AI tool you used this week (search, writing, navigation, recommendations). For each one, ask: who controls it, and on what terms?
- Day 3 — Research your country’s AI policy: Find your government’s current position on AI regulation (most have published something in the past two years). Identify one gap between the policy and Suleyman’s ten steps.
- Day 4 — Study a containment success: Read a short account of the Montreal Protocol or the Nuclear Non-Proliferation Treaty. Note what made it work, and which of those conditions apply to AI today.
- Day 5 — Explore one safety organization: Visit the website of one AI safety research group — Anthropic, DeepMind’s safety team, ARC Evals, or the Center for AI Safety. Read their current research summary.
- Day 6 — Evaluate one piece of AI-generated content: Find an article, image, or video you suspect was AI-generated. Try to verify it. Reflect on what the experience tells you about the epistemic risks Suleyman describes.
- Day 7 — Commit to three actions: Choose three concrete behaviors: one professional (how you discuss or deploy AI at work), one civic (an organization you’ll support or a policy you’ll advocate for), and one personal (how you’ll stay informed).
Frequently asked questions
What is The Coming Wave about?
The Coming Wave argues that artificial intelligence and synthetic biology are arriving simultaneously as general-purpose technologies that will reshape every aspect of society. The book’s central concern is the “containment problem”: how to keep these technologies under meaningful human control while still capturing their benefits. Suleyman draws on his experience as a DeepMind co-founder to explain why this challenge is structurally different from anything humanity has faced before, and then maps ten concrete steps toward a viable solution.
Who is Mustafa Suleyman and why should I trust his perspective?
Mustafa Suleyman co-founded DeepMind in 2010 — one of the world’s most influential AI research labs — and was present at the creation of many technologies he writes about. He later co-founded Inflection AI (creator of the Pi assistant) and became head of Microsoft AI. His credibility is unusual: he is neither a journalist reporting from outside the industry nor a pure academic. He built the wave he is describing, which gives his concerns about containment particular weight. His co-author Michael Bhaskar brings broader cultural and economic context.
What does “containment” mean in The Coming Wave?
Containment is Suleyman’s term for the set of mechanisms — regulatory, technical, social, and international — that keep a powerful technology under meaningful human oversight. It does not mean banning or stopping AI; Suleyman explicitly argues that is neither possible nor desirable. It means building in accountability, auditability, and control from the start, and creating institutions capable of monitoring and governing the technology as it evolves. He points to the Montreal Protocol on CFCs as an imperfect but instructive historical example of successful containment.
Is The Coming Wave pessimistic about AI?
No — it is urgent but not despairing. Suleyman is explicit that he is neither a techno-utopian (AI will solve everything automatically) nor a doomsayer (it is already too late). He believes the benefits of AI and synthetic biology are real and worth pursuing: cures for diseases, dramatically increased productivity, scientific breakthroughs. His concern is that without serious governance work, the risks will outpace the benefits. The final section of the book — the ten steps to containment — is genuinely constructive, though Suleyman is honest about how hard the work will be.
What are Suleyman’s 10 steps to containment?
The ten steps are: (1) prioritize AI safety research; (2) liberal democracies must lead; (3) demand corporate responsibility; (4) build international consensus; (5) enact regulation and legislation; (6) implement auditing and liability; (7) establish choke points and controls on hardware and software; (8) increase public investment in AI safety; (9) empower civil society and a free press; and (10) exercise individual responsibility in how we use and advocate around AI. Suleyman frames these not as a complete solution but as necessary starting points — a coalition of efforts rather than a single silver bullet.
How accurate are the book’s predictions so far?
Published in 2023 and updated in 2024, the book has held up well. AI capabilities have continued to expand rapidly (GPT-4 and its successors, multimodal systems, AI coding tools). The governance gap Suleyman identified has, if anything, widened — regulatory frameworks remain fragmented, export controls on semiconductors have emerged as a key choke point as he predicted, and the economic disruption to knowledge work has accelerated. Some critics note the book underweights positive developments in alignment research; others feel it understates how quickly safety culture has grown inside major labs since 2023.
Is The Coming Wave worth reading if I am not technical?
Absolutely. Suleyman is a skilled writer, and the book reads more like a thoughtful policy essay than a technical manual. He assumes no prior knowledge of machine learning or molecular biology; when he introduces technical concepts, he does so through narrative and analogy. Readers who engage with current events — in business, politics, or culture — will find the book immediately relevant. The most demanding parts are the historical sections on containment precedents, but those are also the most intellectually rewarding for non-technical readers who want the full picture.
Related summaries
- Nexus by Yuval Noah Harari — How information networks shape civilization, from ancient myths to modern AI
- Power and Prediction by Ajay Agrawal, Joshua Gans & Avi Goldfarb — The economic logic of how AI redistributes power across industries
- The Age of Surveillance Capitalism by Shoshana Zuboff — The business model that helped build the wave Suleyman describes
- The Growth Reads: Best AI Books →