AI investment and its potential effects on urban digital twins

Looking back to gaze ahead

It is worth recalling the dot.com era from the end of the previous century to judge AI’s current hype. The Nasdaq Composite index – a stock index that skews toward information-technology companies – peaked at more than 5,100 points in March 2000 and then rapidly declined to a final low of just barely above 1,100 points in October 2002 (it took more than 12 years then to move beyond 5,000 points again).

The January 2000 Super Bowl event marked the height of the bubble with 14 in-game ads by dot.com companies – only one of them still active as an independent company today. Now, many analysts see the signs of a tremendous AI bubble accumulate.

A crash is likely in the making. Similarly to 2000, a bursting bubble does not mean that AI will go away, as internet-enabled companies and business models did not vanish. On the contrary, AI will flourish as the internet did. In fact, many infrastructure elements such as datacentres will become affordable for general use after lofty valuations come down.

During the late 1990s, the construction of fibre communication networks was perceived as a tremendous business opportunity. The business never became as profitable as expected, but the initial excitement created an infrastructure of dark fibre – unused but readily available communication lines – that supports today’s business models as a commodity that can be readily leveraged.

AI as an enabling technology will boost capabilities and accelerate the use of advanced digital twins. In particular, digital twins that have to work with difficult-to-capture data and not completely understood real-world dynamics will benefit tremendously. Digital twins of machineries can rely on solid understanding of physics and measurable data that sensors can cost-effectively capture.

Factory environments have many known dynamics and interactions of equipment – even workers’ likely movement patterns can be plugged into simulations. But urban digital twins attempt to capture dynamics and behaviours of relevant elements across entire cities. They are not only subject to less understood dynamics but also phenomena that are difficult – often impossible – to measure.

AI can make available data usable and create data of unmeasurable phenomena. AI in digital twins also allow the use of scenarios to better prepare for sudden events that can affect the entire system in unexpected ways. City managers thereby can develop strategies for unusual weather events, pandemic-like occurrences or localised industrial accidents with ripple effects across the urban landscape.

Digital twins and AI to plan for tomorrow’s cities 

Digital twins of urban environments are difficult to design, implement and maintain. The potential impact such digital twins can have commercially and societally promises to be substantial, however. Because of the number of parameters, intersecting dynamics and range of conceivable scenarios, the benefits AI can provide in understanding urban environments are massive. AI and digital twins reinforce each other.

AI can speed up the building of digital twins by supporting code development for virtual environments. Such applications accelerate overall design development and allow embedding design details more easily. For clients and users, AI reduces costs, enables faster implementation of digital twins, and allows for quick and inexpensive changes and alterations as requirements change or new needs arise. In addition, AI can improve the interface experience between virtual environments as well as simulations of operations and users.

Ari Lightman, professor at Carnegie Mellon University, explains: “Generative AI would be used to look at the entire simulation and turn it into a summary for humans. It could tell me things I might be missing and summarise things in a way I can understand.”

AI doesn’t only benefit digital twins but digital twins also support AI’s capabilities. Scott Likens, emerging technology leader at PwC, says: “We’re using digital twins to generate information for large language models [LLMs]…We see opportunity to have the digital twins generate the missing pieces of data we need, and it’s more in line with the environment because it’s based on actual data.” Such synthetic data of missing pieces are also finding use in other applications as “AI, XR, digital twins set to transform robotics”.

Nvidia serves the market of smart cities as city planners and managers are turning to digital twins and AI agents for urban planning scenario analysis and data-driven operational decisions, according to the company. It is providing a range of solutions to enable users to create photo-realistic, simulation-ready digital twins of urban environments to optimise city operations.

A partnership of Japanese companies is developing the digital entertainment city Namba in Osaka, Japan. The aim is to create the world’s first smart city that integrates AI, extended reality [XR], and decentralised physical infrastructure networks [a blockchain-based approach to manage decentralised networks] on a city-wide scale. The group intends to offer services beyond entertainment and tourism. Namba, being a neighbourhood within Osaka, has a limited claim to a city-wide application of the concept, however.

The silver lining of AI overinvestment 

The existence of an AI investment bubble is increasingly perceived as a foregone conclusion. AI companies and technology suppliers are now even investing in each other’s operations, adding to lofty valuations. There are obvious indications of a bubble, but positive effects can emerge from the current investment excitement. Whatever the outcome, applications for digital twins will see their timeline solidify as the immediate future of AI plays out.

If use of AI applications proves to be an all-encompassing and rapidly growing market opportunity, the immense investment of the past couple of years will be retroactively viewed as forward-looking wisdom that locked in favourable competitive positions and profits for years to come. More likely though, investors have outrun their headlights, and expectations of adoption and diffusion of AI applications over the next years are vastly overrated.

If so, there will be a shock to the system like the burst of the dot.com bubble at the beginning of the century when the Nasdaq Composite Index dropped by almost 80% within 30 months. Initial warnings existed, with the former chair of the Federal Reserve using the phrase “irrational exuberance” when discussing the development at the stock market in December of 1996. Warnings of an exuberant AI bubble are common today.

Bursting investment bubbles hurt investors and bring down many companies –25 years ago, a slew of dot.com companies vanished. But related overinvestment in infrastructure can make assets suddenly affordable, opening new opportunities. Such affordability changes cost structures that enable business models that could not have become successful at previous valuations. Infrastructure overhang – infrastructure build for rapid growth that does not materialise in the short run – leads to commodification of infrastructure elements, which can democratise a technology for incumbents and startups alike.

The over-investment in fibre during the dot.com years ended up creating dark fibre – overbuilt fibre cables for data transmission – and this infrastructure has served as a ready and inexpensive resource ever since. For AI, investment in datacentres is comparable to the fibre investment from 30 years ago.

Morgan Stanley analysts forecast datacentre spending globally of up to almost $3tn between now and 2028. The amount is staggering, and it is difficult to imagine use cases and adoption rates that will provide the required return on investment for virtually any business model. But as initial investors see their investments decrease or vanish, new players can snap up or use related infrastructures at bargain prices.

Alkesh Shah, a tech analyst with Bank of America, explains the underlying reason for such recurring dynamics: “You always overestimate how fast the change will happen, and you underestimate the magnitude of the change.”  

The impact digital twins will have on the marketplace will follow a similar dichotomy between today’s expectations of rate of change and tomorrow’s impact of such change. Digital twins require many technological bits and pieces to come together, and AI will play an important role for digital twins – if not tomorrow, then the day after tomorrow.

Martin Schwirn is the author of “Small data, big disruptions: How to spot signals of change and manage uncertainty” (ISBN 9781632651921). He is also senior advisor for strategic foresight at Business Finland, helping startups and incumbents to find their position in tomorrow’s marketplace.