Artificial Intelligence
Boom versus bubble
Do Expect
Artificial general intelligence (AGI) within a decade. Increases in computing power will enable further breakthroughs.
Moore’s Law for AI, but more so. There are many ways to squeeze out more speed other than improving the chips themselves.
White-collar woes. Past automation improvements replaced factory workers. This one is coming for accountants, junior lawyers, and consultants.
Don't Expect
AGI in the next couple of years. It will beat us at many individual tasks, and its capabilities will continue to surprise us, but we’re still far from the point where it can do everything.
Good AI regulation. The tech is advancing too fast for regulators to keep up. Some proposed rules, like ones based on the size of AI training runs, are already obsolete.
Watch This Space
Vertical integration. In an effort to lower costs, the hyperscalers — companies with massive data-center businesses such as Amazon, Microsoft, and Google — are starting to compete with their own customers, creating chips to rival Nvidia’s, and designing their data centers to squeeze every last drop of juice out of those chips. Nvidia, in turn, is building its own data centers designed around its proprietary silicon. It’s a battle royale for slices of the AI pie — but that pie will still grow.
“Everything that moves will be autonomous.”
“Robots in the human environment, to me that’s the final frontier.”
“A trillion dollars is not cool anymore. What’s cool is a quadrillion dollars.”
Straw Man
Is this an AI investment bubble?
The billions of dollars that big tech companies are pouring into AI haven’t paid off yet, but critics forget that most transformative technologies weren’t immediately profitable. Software development alone is already bringing in billions for OpenAI, Anthropic and others, and once big companies realize how much money AI can save them — including, yes, by replacing some humans — profits will follow. Cheaper AI models like China’s DeepSeek will spur demand, too. And tech companies’ shareholders seem happy for now to carry these investments.
The New Lexicon
“Embodied AI”
AI and robotics once followed largely separate tracks, but they’ve been merging as researchers fit robot bodies with AI brains so they can learn to operate better in the physical world. Indeed, some argue that true artificial intelligence can’t flower on a server cluster, but needs a physical body to develop the intuitions and concepts that come from interacting with the world.
Artifical Intelligence
Vibers vs veterans: Why scale may not save you
Business leaders have long relied on scale as their trump card: vast resources, sweeping customer bases, and entrenched brand recognition. But AI is catalyzing a new breed of “vibers” — fast-moving, AI-first startups — who can win market share faster and cheaper than ever before.
These newcomers challenge the very notion that size alone bestows safety or success. Over the next five years, the collision between vibers and veterans will redefine competition across industries. For veterans, the key to survival is to protect their moat and embrace speed.
The vibers are so called because of “vibe coding”: creating complex software by talking to AI coding tools in ordinary language instead of writing code from scratch. Y Combinator, the world’s foremost startup incubator, revealed that its most recent graduates generated 95% of their codebases using AI tools like GitHub Copilot or GPT-4.
This allows small teams to build big businesses. Founders surround themselves with suites of AI helpers not just for coding but also for design, sales, marketing, and customer support. For instance, Greenhouse increased marketing clicks by 23% using Anyword, an AI-powered tool that optimises product messaging. Similarly, SaaS analytics startup Runway employs AI simulations to gauge user responses to new features, cutting development costs by up to 75%.
The viber approach is expanding into every aspect of business. Eric Schmidt, the former Google CEO, suggested that soon you could ask your AI to “make a copy of TikTok. Steal all the users, steal all the music, put [your] preferences in it, produce this program in the next 30 seconds, release it, and in one hour, if it’s not viral, do something different along the same lines.”
“Over the next five years, the collision between vibers and veterans will redefine competition across industries.”
For mere tens of dollars a month, a founder can conduct market research using synthetic personas that behave a lot like real humans; develop a working application by simply describing it in a few words; or build a sophisticated marketing campaign with little human input. Firms like Writer.com and Lindy exemplify this, achieving in months what traditionally required years.
In the recent past, five to seven years has been ample time for revolutionary entrants to reshape entire industries — think of how the iPhone reinvented mobile telephony, or how Uber transformed the taxi business. With AI, disruptions could be even faster and more profound.
Historically, scale, hard-earned over decades, has insulated incumbents from competition. But an AI-driven economy will be filled with many new entrants by 2030. They will move quickly, test fast, iterate rapidly. On top of that, consumers’ time and attention will be more thinly spread. Their expectations and how they make decisions may change. The new entrants will be able to respond to these changes much more quickly.
Many of these new startups’ bosses will have cut their teeth in the safe domain of a well-established, blue-chip company, and those companies will now face a talent drain. It’s increasingly commonplace for startups to achieve revenues of $10 to $20 million (or even $100 million) annually with fewer than two dozen employees — Lovable, Cursor, Submagic among them. To become a CEO founder, to step out of the safety of the institution, is now both easier than ever and offers a clearer path to riches.
To withstand these pressures, incumbent companies must establish strong, defensible moats. Effective strategies include proprietary technologies, specialised manufacturing capabilities, exclusive partnerships, and personalization of services based on user data. Companies like Netflix and Spotify embody the power of such personalization. Strong brands that embody trust and core values, such as Patagonia or Apple, will also remain formidable.
Yet moats alone won’t suffice. Incumbents must match or surpass the speed of vibers. Rapid decision-making, supported by real-time AI analytics, is crucial. So are flattening organisational hierarchies, streamlining approval processes, and swiftly iterating.
Retaining top talent will also be critical. One way is to start internal “talent studios,” which nurture innovation and entrepreneurship. This gives employees a chance to develop new ventures in-house instead of leaving to become competitors.
The next five years will separate market leaders from corporate casualties in the AI revolution. At the minimum, new competition will nibble at your margin, and the loss of talent will hamper you. Embracing the vibe will prevent it from being worse.
Semafor Business
The stories (& the scoops) from Wall St.
A twice weekly briefing from two of Wall Street’s best sourced reporters, Liz Hoffman and Rohan Goswami. Read by over 50k senior executives in business & finance. Sample the latest issue here.
