In 2021, a single pop-up changed the economics of the mobile internet. Apple began asking iPhone users whether apps could track them across other companies' apps and websites. Most said no. Within a year, Meta estimated the change would cost it roughly $10 billion in annual revenue.

The usual story frames this as privacy versus progress, a tax on innovation. Follow the trajectory forward, though, and a different picture emerges. The constraint is pushing engineers toward techniques that learn from data without seeing it. It is also pushing companies toward relationships built on consent rather than surveillance. That is the privacy paradox: the less data a system hoards, the more valuable it may become.

Constraints have a long history of redirecting innovation. Fuel efficiency standards pushed carmakers toward lighter materials and hybrid drivetrains. Privacy rules and rising user expectations are doing something similar to software. When collecting everything is no longer free, teams have to answer a sharper question: what data do we actually need? Many organisations avoided that question for two decades.

Answering it tends to produce better engineering. Processing data on the device rather than in the cloud reduces latency and works offline. Collecting less reduces breach liability, storage costs, and regulatory exposure. Apple turned on-device processing into a marketing pillar, and messaging apps like Signal built entire brands around knowing as little as possible about their users. In each case, a limitation became a feature customers could see and choose.

Look further ahead and the advantage compounds. Organisations that build for data minimisation today are developing architectures, talent, and processes that become harder to copy over time. Competitors that still depend on bulk collection face a widening gap as regulations tighten across jurisdictions. The early movers are positioned for a future in which restraint is expected by default.

Takeaway

Treat privacy constraints the way engineers treat physical limits. They are design parameters that force more elegant solutions, so build around them rather than routing around them.

Every data-driven business runs on an implicit bargain: users hand over information in exchange for a service. For years that bargain was invisible, buried in terms of service no one read. Now that people can see it, they are renegotiating its terms, and trust is becoming the currency.

Trust behaves differently from data. Data extracted quietly tends to be noisy, inferred, and fragile. It can vanish overnight with a new regulation or platform policy, as advertisers discovered after Apple's tracking prompt. Data shared willingly is often richer. A customer who trusts a company will tell it what they actually want: their preferences, their plans, their limits. This is the logic behind zero-party data, information people deliberately volunteer because they see the value in sharing it.

Imagine two plausible futures for a health app. In the first, it quietly harvests behavioural signals and sells insights to third parties, until a breach or exposé erodes its user base. In the second, it collects only what users explicitly share and keeps sensitive analysis on the device. The second app knows less in aggregate, but its users stay longer, share more deliberately, and recommend it to friends. Over a decade, that relationship may well be worth more.

Takeaway

Data taken without trust is a depreciating asset. Data given with trust tends to appreciate, because the relationship keeps producing it.

This shift rests on more than idealism. A maturing family of privacy-enhancing technologies now lets systems extract value from data while keeping the raw information hidden. Several have moved from research papers into products used by many millions of people.

Federated learning trains AI models across many devices and sends back only model updates, never the personal data itself. Google uses it to improve keyboard predictions. Differential privacy adds carefully calibrated noise so that patterns emerge while individuals stay hidden, and the US Census Bureau applied it to the 2020 census. Homomorphic encryption allows computation on data that stays encrypted. Zero-knowledge proofs let someone prove a fact, such as being over 18, without revealing their birthdate.

The roadmap matters as much as the current snapshot. Homomorphic encryption remains computationally expensive, and differential privacy involves real trade-offs in accuracy. But these costs are following familiar curves as hardware improves and algorithms are refined. Over the next decade, these techniques will plausibly combine and fade into infrastructure, much as web encryption went from optional to invisible once HTTPS became standard. Few people now think about the padlock in their browser. They simply expect it.

Takeaway

When evaluating an emerging technology, look past what it costs today and ask which way its cost curve is bending. That tells you when the impossible becomes the default.

The privacy paradox resolves once you lengthen the time horizon. In the short run, protecting data looks like a cost. Across a decade, it looks like an investment in architecture, trust, and resilience that competitors struggle to replicate.

For strategic planners, three questions offer a useful framework. Which of our capabilities depend on data we may soon lose access to? What could we build if we assumed we'd see less? And which privacy technologies are nearing the point where they become cheap enough to change our options?