Good morning from Hong Kong.
Like a lot of people, I now use ChatGPT just about every day. Prompt, response; prompt, response. It’s simple, fast, cheap, and (mostly) reliable—almost like magic. But that magic depends on a sprawling network of very real infrastructure—GPUs, servers, data centers, cooling systems, undersea cables—and consumes a staggering amount of energy.
But I’m newly conscious of that infrastructure since touring a mammoth data center in Singapore a few weeks ago. If you’ve never been to one (I hadn’t) the experience is jarring. The facility I visited was 290,000 square feet—about the size of two Costcos. You can hear the hum of the cooling systems from hundreds of yards away.
Singapore is a tropical city-state with barely a quarter the land mass of Rhode Island. It may be the worst place on the planet to build data centers: land is scarce, energy is expensive, and the island is hot and humid all year round. And yet the city-state is home to more than 70 data centers, making it one of the world’s densest digital hubs.
Five years ago, alarmed by the realization that data centers were guzzling 7% of the city-state’s total electricity consumption, Singapore’s government imposed a moratorium on building new ones. But there was no stopping the surge in demand for the facilities. Operators rushed to build more across the straits in Malaysia. A few months ago, Singapore reversed tack, announcing a plan to increase data center capacity by as much as a third over the next several years.
Singapore’s data center dilemma highlights the challenges faced by small and medium-sized economies in the age of Big Data. In his 2018 book AI Superpowers venture investor Kai-Fu Lee argued that the world would evolve into Cold War-style digital power blocs, one led by the U.S. and the other by China. The rest of the world, he suggested, was doomed to become digital vassals of the two giants.
But Singapore is holding its own in the global AI arms race. In the new issue of Fortune, I argue that the city-state, home to only 5.6 million people, offers a case study in how small and medium-sized economies can unlock AI’s power—and suggests that in many ways small states can be uniquely agile in doing so.
Data centers are one example. My tour guide at the hulking facility I visited was Tim Rosenfield, an Australian engineer who is co-founder of Singapore-based cloud services provider Sustainable Metal Cloud. Tim and his colleagues have developed a technology that uses liquid immersion to cool GPUs instead of the energy-intensive air-cooling method used in traditional data centers, reducing space, power consumption and carbon emissions by up to 50%. Tim says he and his co-founders came to Singapore to launch their venture precisely because the island nation is so severely resource-constrained. His argument: Singapore benefits from the opposite of the “resource curse.” Incremental tweaking doesn’t cut it in a place like Singapore; the city-state has to embrace radical innovation when it comes to digital technology because it has no other choice.
Tim was among the many insightful speakers at Fortune’s inaugural Brainstorm AI summit in Singapore this week. Others included Tim Baldwin of the UAE’s Mohamed bin Zayed University of Artificial Intelligence, Quantum AI founder Seth Dobrin, Accenture’s Joon-Seong Lee, Microsoft Asia president Ahmed Mazhari, and Singapore Minister for Digital Development and Information Josephine Teo. You can find video of Tim’s presentation, and those of more than 60 other brilliant panelists on the Fortune Live Media site here.
More news below.
Clay Chandler
clay.chandler@fortune.com
Follow on LinkedIn
TOP NEWS
Inside Volkswagen’s EV battle
Volkswagen is making an aggressive pivot to EVs, aiming to have more than half the vehicles it sells globally be electric by the end of the decade. It’s not quite meeting expectations: production at VW’s EV factory in Zwickau, Germany is slipping from 247,000 cars in 2023 to a projected 220,000 cars this year. VW’s size could end up being a disadvantage against more nimble producers, like Tesla or China’s BYD: “The new producers can be like speedboats,” says Joe McCabe, CEO of industry analysts AutoForecast Solutions. Fortune
AI rhetoric
Last quarter, Meta, Microsoft and Google parent Alphabet spent a combined $40.5 billion on infrastructure, land and chips to power AI services. Fortune’s Alexei Oreskovic notes that figure is the only hard number that emerged from this week’s earnings calls, as tech companies give optimistic, yet vague, promises about how AI will transform their businesses. Meta CEO Mark Zuckerberg, for example, said the company was on track to have “the world’s most used AI assistant by the end of the year”—without giving much indication of what he was referring to. Fortune
Prisoner swap
The U.S. and Russia conducted the largest prisoner swap since the end of the Cold War, as Moscow released journalist Evan Gershkovich and fellow American Paul Whelan, as well as other journalists and dissidents. In return, the U.S. and its allies released several Russians suspected of serving as intelligence agents. The Associated Press
AROUND THE WATERCOOLER
Amazon finance chief blames weak sales forecast on shoppers distracted by Trump assassination attempt and Olympics by Jason Del Rey
The European Commission is banning staff from nabbing free tickets for the Olympics, Euros and Eurovision as they could defy ‘ethical behavior’ by Prarthana Prakash
Americans: If you want to move abroad because of the election, Spain is your best bet by Orianna Rosa Royle
Cities no longer need to be rich to be smart as AI ‘levels out the playing field,’ urban experts say by David Austin
Bosses want to use AI to boost productivity—but 77% of employees say it just creates more work by Emma Burleigh
This edition of CEO Daily was curated by Nicholas Gordon.
