How should we assess Rupert Murdoch’s legacy?
A ship’s captain is responsible for the actions of his entire crew. Leaders who refuse to acknowledge responsibility for the actions of their employees should be robustly challenged, not let off the...
Artificial Intelligence
3 critical actions for when the AI bubble bursts
Iconoclast: n. A person who attacks or criticizes cherished beliefs or institutions.
This on-going column is intended to challenge the status quo. Not always giving answers, not always right, but smartly challenging our own assumptions, actions, organizations and institutions.
When the AI bubble bursts, your board will ask whether to cut AI initiatives. What will your response be?
In March 2000, Cisco Systems commanded a market capitalization of $555bn, briefly becoming the world’s most valuable company. As the infrastructure provider to the internet revolution, it seemed invincible. Within two years, however, it had lost 86% of its value. Microsoft, Intel, Oracle, and virtually the entire infrastructure layer of the dot-com boom saw similar devastation.
Yet here is what the panic obscured: Cisco’s routers kept working. The internet kept growing. While investors fixated on the spectacular collapse of Pets.com and Webvan, something quieter was happening. Traditional retailers were learning to ship products efficiently. Banks were moving transactions online. Manufacturers were rebuilding supply chains around digital connectivity. The value did not disappear – it migrated. When the AI bubble eventually bursts, as it eventually will, the story will follow a similar script.
Today’s market assigns extraordinary valuations to the infrastructure providers: the hyperscalers, the chip manufacturers, the foundation model developers. This mirrors the logic of the 2000s perfectly. But here is the critical difference: in the dot-com era, building internet infrastructure required massive capital expenditure and technical expertise. The infrastructure providers captured value because they controlled genuinely scarce resources.
AI infrastructure, by contrast, is rapidly commoditizing. Model capabilities are converging. Compute costs are falling. Within 18 months of GPT-4’s launch, comparable capabilities became available through multiple providers at a fraction of the cost. Additionally, the APIs are multiplying, and the capabilities of agents are even making an API-less world theoretically possible. When the bubble eventually bursts, the market will suddenly recognize this commoditization. The correction will be swift and brutal. Many will suffer; I probably will too. But, as in 2000, that pain will obscure the more important story.
The real value is already shifting to a different category of company entirely: the organizations that understand how to orchestrate AI within their operational reality. Not the ones talking about transformation, but the ones quietly rebuilding their business processes around AI capabilities. Not the ones buying Copilot licenses, but the ones redesigning the processes to solve problems to add and capture value. They are invisible to the market because they are not AI companies. They are insurance companies that have reduced claims processing costs by 40%. Pharmaceutical companies that have compressed drug discovery timelines. Logistics operations that have optimized routing with unprecedented precision.
Consider the parallel. In 2001, while Cisco was cratering, Walmart was methodically building the most sophisticated supply chain in retail history, using internet-enabled logistics that would devastate competitors for decades. Zara was creating a fast-fashion model that depended entirely on digital information flows. These were not “internet companies”; they were companies that became exceptional at using internet infrastructure. The market barely noticed their transformation until the value gap became impossible to ignore.
The same pattern is emerging now. I hypothesize that the companies that will dominate the next decade will not be distinguished by their AI models; those will be utilities, accessible to everyone. They will be distinguished by their ability to embed AI into workflows, linking AI to their own data to restructure decision-making around algorithmic insights, to build organizations where humans and AI systems have genuinely complementary roles, and where AI-human teams emerge as a new form of talent. This requires capabilities the market does not yet know how to value: change management at scale, process redesign expertise, and organizational cultures capable of continuous adaptation.
History does not repeat, but it rhymes.
The dot-com crash taught us that infrastructure providers capture value early but rarely sustain it. The real winners were companies that no one was tracking: mid-market manufacturers that rebuilt themselves around digital operations, regional banks that quietly dominated online banking in their markets, and B2B distributors that made procurement seamless.
For executives, this value migration creates an unusual strategic opportunity. When the AI bubble bursts, the panic will create a window. While competitors freeze, while boards demand explanations, while consultants scramble to revise their presentations, the leaders who understand this historical pattern will double down. Not on AI technology itself, that will be cheap and abundant, but on organizational transformation. On building the capabilities to absorb and deploy rapidly evolving technology. On creating cultures where experimentation is systematic rather than theatrical.
History does not repeat, but it rhymes. In 2000, the value did not vanish; it migrated from those who built the infrastructure to those who used it best. The AI bubble will burst the same way. The infrastructure will keep working. The models will keep improving. And the companies that learned to orchestrate this technology within their operational reality, rather than worship it from a distance, will inherit the future that the tech giants are currently pricing into their stocks. The only question is whether you will be ready to capture the value that is about to migrate to you.
The morning after the bubble bursts, when your board asks whether to cut AI initiatives, the right answer will not be yes or no. It will be: “Which ones are building genuine operational capability, and which ones are just summarizing emails?” The companies that can answer this honestly, that can distinguish between transformation theater and actual organizational change, will emerge from the correction positioned to capture the value that the infrastructure providers are about to lose.
“The bubble burst will make AI technology radically cheaper within months.”
Three actions for the morning after
When the market opens and the panic begins, resist the instinct to freeze. Instead, take these three steps:
First, conduct a ruthless audit of every AI initiative in your organization
Gather your leadership team and ask one question about each project: “If we stopped this tomorrow, what operational capability would we lose?” If the answer is “nothing” or involves words like “exploration” or “learning,” cut it immediately. If the answer describes a specific process that would revert to a slower, more expensive, or less effective state, protect it fiercely. The correction will force this clarity eventually. Better to impose it yourself while you still control the narrative.
Second, reallocate resources from technology acquisition to integration capability
The bubble burst will make AI technology radically cheaper within months. Your constraint is not access to models but your organization’s ability to absorb them. Shift budget from licensing fees and vendor contracts toward change management expertise, process redesign talent, and training infrastructure. Hire the people who know how to rebuild workflows, not the ones who know how to write prompts.
Third, use the panic to accelerate decisions you have been avoiding
Every organization has AI initiatives that exist primarily for political reasons: the executive sponsor who demanded a pilot, the board member who asked about ChatGPT, the department that needed to appear innovative. The bubble burst gives you permission to kill these projects without career risk. More importantly, it creates space to fund the unsexy work that actually matters: the data infrastructure projects, the process standardization efforts, the cultural changes that enable genuine transformation. Your competitors will spend the next 12 months in a defensive posture. You have a window to build capabilities that they will not notice until it is too late.
One action for you to take today: Assume the bubble has already burst, and start applying the three actions now.
Authors
Related
Decisions about the future of AI shouldn’t be made by “dark personalities”
Traits such as overconfidence and narcissism, which may lead tech leaders to make bold decisions and take risks in the pursuit of growth, are dangerous in the context of artificial intelligence....
Switzerland will suffer the shadow cost of Credit Suisse/UBS marriage
By ignoring the views of UBS shareholders, trashing the priority rule, and failing to protect Credit Suisse shareholders, regulators have tarnished the reputation of Swiss banking for years to come....
Credit Suisse: how the mighty fell
What was behind the shocking demise of a treasured institution – and how will the proposed rescue package play out?...