29 August 2026

 

Predicting the future of technology is easier when you control part of the budget required to build it.

Bill Gates anticipated pocket computers because Microsoft was helping to popularise personal computing. Jeff Bezos saw ecommerce becoming everyday infrastructure while Amazon was constructing it. Elon Musk correctly identified the potential of electric vehicles and reusable rockets, but repeatedly misjudged how quickly autonomous driving would arrive.

These are not detached futurists peering into a crystal ball. Technology leaders can direct investment, hire engineers and persuade customers that an unfamiliar product is inevitable. Their predictions may describe the future, but they can also function as strategy, recruitment tool and marketing campaign.

So which technology predictions came true, and which failed? The answer depends on more than whether the underlying invention eventually appeared. Did it arrive on time? Did it work at the promised scale? And did people use it as expected?

 

Bill Gates predicted pocket computers but overestimated digital assistants

 

In his 1995 book The Road Ahead, Bill Gates described computers small enough to carry in a pocket, video on demand, online social interaction and digital agents that would learn their users’ preferences.

His pocket-computer prediction proved remarkably accurate. Smartphones now combine communication, navigation, photography, entertainment, payments and identification in one device. Video on demand and social networking also became commonplace, although their societal effects were not necessarily what Gates imagined.

Reviewing the book 25 years later, Gates acknowledged that he had failed to anticipate how social networks would allow people to filter out opposing perspectives and harden divisions. He predicted the technology more successfully than the behaviour it would encourage.

His digital assistants were a different kind of near miss. Gates expected software to learn about its user much as a human assistant would. Siri, Alexa and Cortana arrived, but remained far more limited than the agents he envisaged.

Generative artificial intelligence has brought that idea closer. However, today’s AI agents still struggle with reliability, persistent memory and genuinely independent action.

Verdict: right technology, longer timetable and unintended consequences.

 

Jeff Bezos correctly predicted personalised online shopping

 

Amazon was still primarily an online bookseller when Jeff Bezos wrote his first shareholder letter in 1997. Online commerce, he argued, already saved customers time and money. In future, personalisation would accelerate “the very process of discovery”.

That was more insightful than merely predicting that people would buy products online. Bezos foresaw a commercial environment in which software would influence what consumers encountered.

Amazon was already experimenting with personalised recommendations. In 1997, it introduced BookMatcher, which used collaborative filtering to make suggestions based on customer ratings. Modern recommendation systems analyse vastly more behavioural data and now shape shopping, music, films, news and social media.

Bezos understood that ecommerce would not simply reproduce a shop on a screen. It would change how products found customers.

This was also a prediction Amazon could help fulfil. The company developed recommendation systems, one-click purchasing, fulfilment infrastructure and a marketplace of third-party sellers. Bezos was forecasting a market while building the machinery needed to create it.

Verdict: broadly right—and partly self-fulfilling.

 

Elon Musk got autonomous driving’s direction right but its timing wrong

 

Elon Musk offers perhaps the clearest warning against confusing technological direction with a dependable timetable.

At Tesla’s Autonomy Day in April 2019, Musk said he felt “very confident” that the company would have one million robotaxis on the road in 2020. Privately owned Teslas, he suggested, could become autonomous through software updates and earn money for their owners.

That did not happen.

Tesla now operates paid Robotaxi services using company-controlled vehicles within limited areas of selected US cities. By July 2026, the company said it had expanded unsupervised rides beyond Austin to cities including Miami, Orlando and Tampa. This represents genuine progress, but it is fundamentally different from remotely converting one million customer-owned cars into autonomous taxis.

Tesla’s consumer product is still called Full Self-Driving (Supervised). The company explicitly states that its currently enabled features do not make a vehicle autonomous and require an attentive driver who is ready to intervene.

Musk correctly identified the convergence of electric vehicles, artificial intelligence and large quantities of driving data. He underestimated the difficulty of turning an impressive demonstration into reliable autonomy across unpredictable public roads.

Yet Musk’s wider record cannot be dismissed. SpaceX made landing and reusing orbital-class rocket boosters routine, while Tesla helped move electric vehicles towards the centre of the global automotive industry.

Verdict: right trajectory, spectacularly wrong timetable.

 

Mark Zuckerberg’s metaverse prediction remains far from reality

 

When Facebook became Meta in 2021, Mark Zuckerberg said the company hoped that, within the next decade, the metaverse would reach one billion people, host hundreds of billions of dollars in digital commerce and support millions of creators and developers.

There are still several years left on that timetable, but progress has been unconvincing. Virtual reality has found durable niches in gaming, training, design and fitness, while augmented-reality and AI-enabled glasses are improving rapidly. The metaverse has not replaced the smartphone or become the primary environment for work and social interaction.

Meta reported that Reality Labs reduced its 2025 operating profit by approximately $19.19 billion. It expects similar operating losses in 2026 and acknowledges that some products may not be fully realised until the next decade.

Zuckerberg may eventually be proved right about immersive computing. But the prediction underestimated cost, discomfort, social awkwardness and the absence of a compelling reason for most people to spend hours inside a headset.

Verdict: technically possible, socially unproven and far from its promised scale.

 

Tim Berners-Lee’s open web became commercially concentrated

 

Tim Berners-Lee’s 1989 proposal for a distributed hypertext system at CERN was not a conventional forecast with a deadline. It was a technical vision intended to solve a practical problem: information about complicated, evolving projects was becoming difficult to find.

The proposed system connected information stored on different computers through links. The World Wide Web subsequently became a global layer for publishing, communication and commerce.

Its institutional outcome was more complicated. Berners-Lee designed the web around open standards and decentralised connections, but much of today’s online activity passes through a relatively small number of platforms. The underlying web remained decentralised; attention, data and commercial power became heavily concentrated.

Verdict: technically transformative, institutionally incomplete.

 

Why technology predictions get the timing wrong

 

The most accurate technology predictions tend to identify an underlying capability rather than a launch date. Computing will become cheaper. Devices will become smaller. Networks will carry more information. Software will become more personalised.

Trouble begins when visionaries translate those durable trends into precise claims about when people will change their behaviour.

Engineering can advance rapidly while regulation, infrastructure, trust and habit move slowly. A laboratory breakthrough is not a product; a product is not a market; and a market is not necessarily a social transformation.

Technology visionaries are often good at seeing what could be built. They are considerably less reliable at predicting when the rest of us will want it.