The return of dark strategy - and how to govern it

Alphabet, Amazon, Meta, and Microsoft are on course to spend more than $660 billion on artificial intelligence infrastructure in 2026.

That sum is larger than the GDP of Belgium and represents a 36 per cent increase on what these firms spent on AI last year. It is also increasingly funded by debt.

What is striking, for anyone trained in conventional strategy, is how precarious the analytical case this spending remains.

Meta founder and CEO Mark Zuckerberg told investors that “making a significantly larger investment here is very likely to be profitable.” A similar message has been echoed by Google CEO Sundar Pichai and Amazon CEO Andy Jassy.

This is not a net present value calculation but a collective act of conviction, running well ahead of revenue that could justify it.

The belief among the mega-players is that whoever cuts capital expenditure first, loses.

Rediscovering cunning in corporate strategy

If you look at Elon Musk’s $44 billion purchase of Twitter, Jeff Bezos’ acquisition of the Washington Post, and Zuckerberg’s $80 billion metaverse gamble, the analytical case has almost invariably arrived after the decision, not before.

What is on display is something older, more primal, and harder to teach.

The ancient Greeks had a name for it. They called it metis: cunning intelligence, the capacity to act decisively in conditions where formal analysis fails.

In Homer’s epic poem, The Odyssey, the hero Odysseus triumphs not by meticulous planning but by reading the moment.

Two millennia later, Machiavelli codified a complementary tradition: a prince must be both fox and lion, manage appearances, and move decisively when fortune opens a window.

That same ethos is on show today. Let’s call it ‘dark strategy’, not because it is evil, but because it runs on emotions and instincts that modern management has preferred to keep in the shadows – conviction, ego, ideology, cunning, and the naked desire for power.

Twitter: finding value in an overpriced deal

Musk is the purest current exemplar of dark strategy. Text messages released during litigation revealed that his acquisition of Twitter was driven by personal frustration with content moderation. At first, it seemed he had overpaid for a declining platform.

However, X’s firehose of real-time content was harnessed to feed xAI. Within three years that was valued at roughly $200 billion.

And in February, SpaceX absorbed xAI in the largest merger ever recorded, valuing the combined company at $1.25 trillion.

Musk’s disastrous overpayment became the enabling acquisition of a vertically integrated empire reaching from the Earth to orbit.

Dark strategy is not confined to business. The tariff programme unveiled by the White House in April – including a 100 per cent duty on patented pharmaceuticals and a flat 50 per cent on steel, aluminium, and copper — was not the result of a conventional policy process.

When the Supreme Court struck down the administration’s emergency-powers tariffs in February, the response of the White House was to reroute rather than retreat.

An analytical policy process would have taken six months. But within six weeks the tariffs were back, re-imposed under national-security authority instead.

This is the tempo of metis. Act first, absorb the legal and economic consequences as they arrive, and justify your decision after the fact.

When an analytical approach is too slow

The pattern is neither heroic nor pathological. It is what the primal instincts of conviction, pride, and ideology produce when formal analysis can no longer keep up with the rate of change.

Machiavelli argued that “fortune favours the bold” precisely because cautious, analytical actors adapt too slowly when the terrain shifts.

This is why hyperscalers such as Alphabet, Amazon, and Microsoft have embarked upon the boldest corporate bet in history.

For these hyperscalers, capital intensity – which measures the amount of fixed assets, such as machinery and technology, required to produce goods or services compared to other factors such as labour – now runs at more than 45 per cent of revenue. These levels were previously reserved for utilities and heavy industry.

Depreciation on these AI assets is projected to reach roughly $400 billion a year by 2030, more than the combined profits of all four firms during 2025. The case for that kind of spending lies not in a formal analysis but in a collective belief about the shape of an industry that does not yet exist.

How to harness dark strategy

The question is not how to eliminate dark strategy. That is not possible and attempting to do so would eliminate much that is valuable within it.

Instead, the key challenge is how to govern it. Odysseus’ cunning succeeds because he remains capable of reflection and course-correction.

In contrast to this, another mythological figure Icarus fails because his conviction hardens into rigidity and feedback is filtered out. He ignores his father’s warning not to use his man-made wings to fly too close to the sun. As a result, the wax holding the wings together begins to melt and he falls to his death.

In business, the difference is institutional, not temperamental. The standard model of corporate governance, in which independent directors provide analytical challenge to management, was designed for contexts where chief executives were accountable professional managers.

It fits less well for founders with supermajority voting stock, direct lines to presidents, and personal fortunes larger than most national economies.

Harnessing their instinctive choices requires a set of structured counterweights: designated internal dissenters with authority to force reconsideration, mandatory pre-mortems before irreversible commitments, independent red teams reporting outside the chief executive’s direct line, and a cultural expectation that challenging the founder’s conviction is itself a job, not insubordination.

The same applies at the level of the state. Strategy was never purely analytical. It was always human, emotional, and political.

The classical traditions of metis and Machiavellian virtù named these forces plainly, because their authors had no illusion that consequential power could be tamed by reasoning alone.

Our own managerial tradition has been sanitised, preferring the language of frameworks to the language of cunning.

It will take years before we can determine whether the current conviction about investing in AI infrastructure was Odyssean cunning or Icarian folly. By that time, the bet will be irreversible.

The true domain of strategic action lies in the shadows somewhere between analysis and recklessness. It is time to bring it into the light.

Further reading:

Increase the odds of success in digital transformation

Multidexterity: How Netflix and Lego achieved achieved digital success

Four keys to using big data to unlock better strategy

Why imitating innovation can be a successful strategy

Loizos Heracleous is Professor of Strategy at Warwick Business School and teaches Strategy and Practice on the Full-Time MBA, Executive MBA, Executive MBA (London), Global Online MBA, and Global Online MBA (London).

Learn more about our developing agility and resilience as a competitive advantage with our Executive Education programmes Leading Through Innovation and Strategy Into Action at WBS London at The Shard.

Discover more about Strategy and Organisational Change. Receive our Core Insights newsletter via email or LinkedIn.