It’s important to understand this because the majority of Berkshire Hathaway’s assets are invested not in publicly traded securities, but in operating subsidiaries like Burlington Northern Santa Fe Railroad, Berkshire Hathaway Energy, and Geico. While it’s true that Buffett invested aggressively in wind energy, that was largely because of government tax incentives. In the main, he preferred to milk his operating subsidiaries for cash rather than reinvest in them for the digital age. Exhibit A is Geico, which thanks to a lack of IT investment has fallen behind Progressive as the nation’s leading for-profit auto insurer.

Buffett has called Geico his favorite child, and for good reason. Since it began in the 1930s, the auto insurer has used a direct-sales model to keep operating costs the lowest in the industry. In a commodity business like insurance, that’s a major competitive advantage. In the 1990s, after he bought all of Geico, Buffett found a second moat when he began to brand Geico as a trusted, even beloved American company. The gecko, the caveman, the camel who celebrated hump day—all these were marketing masterstrokes, ones directly derived from Buffett’s deep understanding of the mass brand-mass media industrial complex. The mascots also highlight how, while Buffett was comfortable investing in marketing, he was deeply uncomfortable with, and therefore didn’t understand, investing in tech.

When Buffett took control of Geico in 1996, he octupled its marketing budget. This wiped out almost all of Geico’s profits from a GAAP accounting standpoint, but Buffett was confident that increasing advertising outlays today would lead to more profitable customers tomorrow. And so it was: Under Buffett’s leadership, Geico’s market share grew from under 3% in 1996 to 12% in 2020, and it went from the No. 7 auto insurer to the #2 auto insurer, behind only State Farm.

So far, so good—but while Geico was investing in marketing, its rival Progressive was investing in technology. Founded only a year after Geico, Progressive began to upgrade its IT systems as early as the late 1970s. In the 1980s, it bought its agents computers and sent them floppy discs so they could better match price with risk. In 1996, Progressive became the first auto insurer to allow consumers to buy insurance online, and it continually streamlined its backend systems so that it could accurately quote new business. Today, Progressive brags that it has tens of billions of price points and that its tech stack allows the company to adjust its rates much faster than its competition—nearly once every business day. “We are a tech company that happens to sell insurance,” is one of Progressive’s internal mantras.

Driving the company’s tech investment was an insight that was perhaps even more astute than Buffett’s marketing insight. Thanks to its no-agent, no-commission model, Geico enjoyed a six-percentage-point cost advantage vs. Progressive in its operating costs. Because half of its business is through insurance agents, Progressive is unlikely ever to catch up here. But Progressive CEO Peter Lewis, who led the company from 1965 to 2000, understood that an auto insurer’s biggest cost center is the claims it must pay policyholders—four to five times bigger, in fact, than its administrative and advertising costs. If Progressive could manage these “loss costs” better than the competition, Lewis reasoned, then it could become the de facto low-cost auto insurer. 

The key to managing loss costs was technology in all its glorious variety. Back-end systems at headquarters that could parse price and risk for each driver were important, but so were front line innovations like Snapshot, a shoebox-sized device that in the 1990s Progressive began installing into the cars of willing customers. Snapshot, now an app on your mobile phone, tracks a customer’s driving behavior; more than one in three Progressive customers buying insurance directly from the company opts in for “usage-based” premiums. Thanks to Snapshot and other innovations, Progressive simply knows more about its drivers than any other insurer, and this creates a virtuous circle in which the company knows which to reward with discounts, which to punish with surcharges, and which to purge altogether. 

Thus, while Progressive’s operating costs have historically been six points worse than Geico, its loss costs have been 11 points better, which means that Geico’s low-cost moat has been breached by tech. In contrast to Progressive’s streamlined system, Geico has more than 600 legacy IT systems. It didn’t start working on a Snapshot-like product until 2019, twenty years after Progressive began. 

Buffett liked to say that when the tide goes out, you see who’s swimming naked, and COVID was the perfect storm to reveal how little Geico had paid attention to its digital wardrobe. During COVID, people suddenly stopped driving, and then, when the pandemic ended, they drove more than ever and more recklessly than ever. At the same time, the worst inflation in forty years hit all sectors of the economy, including auto-repair shops. Such rapidly changing conditions favored insurers with robust tracking tools, like Progressive, and punished insurers without them, like Geico. Since 2020, Progressive has almost doubled its personal auto policy count—but Geico has lost nearly 15% of its personal insurance base. Progressive, not Geico, is now the nation’s number two auto insurer.

It turns out that while the branding of the gecko was important, it wasn’t nearly as powerful as employing sophisticated digital tools. Geico is a good example of what happens when a company, even a powerful one, fails to reinvest in its future. Rather than a virtuous cycle—tech investment leading to better pricing and better products, which drives more profits, which can then be reinvested to drive the cycle on—Geico seems caught in the same vicious cycle that afflicts General Motors, Macy’s and other legacy companies. 

Yet despite deepening trade relations between the U.S. and ASEAN nations like Thailand, Malaysia and Vietnam, Singapore remains the agreement’s only Southeast Asian signatory. That decision comes even as ASEAN nations like Malaysia are investing in their own AI industries, like semiconductors and data centers.

Singapore is “precisely the kind of ‘trusted node’ the U.S. is seeking to anchor AI-era supply chains,” says Ruben Durante, a professor of economics and Provost’s Chair at the National University of Singapore (NUS). Singapore “offers strong governance, regulatory credibility, capital markets, logistics, and advanced data center and connectivity infrastructure.”

The country has a long history with chips. U.S.-based National Semiconductor set up a plant there in 1968, followed by the government’s creation of Chartered Semiconductor Manufacturing in 1987. Singapore now accounts for around 10% of all chip production. 

More recently, Singapore has strived to become an “AI nation,” investing in skilling programs to train its workforce and encouraging local AI development. The country has also attracted billions of dollars’ worth in cloud computing and data centers, including from Big Tech companies like Amazon and Google. 

While the U.S. is trying to shore up its AI supply chain, Singapore might also benefit from being part of Pax Silica, Atreyi Kankanhalli, a computing professor from NUS, suggests. Being part of Pax Silica gives the country—which has less land area than New York City—a seat at the table when the U.S. discusses joint ventures in chip production and logistics. It also gives the resource-poor city-state a safety net to ward off future supply shocks, while enabling access to the latest AI technologies. 

Both the U.S. and China are trying to leverage their dominance in particular industries against each other. 

Washington has blocked the sale of advanced processors, key to training and running AI models, to China since 2022. Beijing, in turn, has slapped export controls on rare earth minerals, a crucial component used for semiconductors and magnets in the AI supply chain. (China has a stranglehold on rare earths, supplying 90% of the world’s processed rare earths and rare earth magnets.)

“The AI race is often framed as a battle over data or models, but the real constraints are increasingly physical—chips, energy and supply chains,” says Simon Chesterman, a law professor from NUS and the senior director of AI governance at research institute AI Singapore.

In addition to Singapore, the U.S. included several close allies in the Pax Silica agreement: Japan, South Korea, Australia, the U.K. and Israel.

Japan and South Korea were chosen as they anchor advanced semiconductor manufacturing, says Durante of NUS. Additionally, Australia is central for critical minerals, the U.K. contributes standards-setting and intelligence alignment, and Israel brings high-end AI and defense-related innovation.

Experts think that the U.S.’s inner circle on AI will soon expand. Durante, from NUS, argues that a small founding group will facilitate early coordination on sensitive issues. Several non-signatories, like the Netherlands and the United Arab Emirates, were involved in initial discussions of the Pax Silica, which Durante sees as an “outer ring” of contributors, even if they’re not yet fully aligned with the U.S

“Expansion will depend on whether Pax Silica develops concrete mechanisms, such as financing, standards, or procurement coordination,” he says, adding that countries which combine industrial relevance with willingness to align on economic-security priorities are the most likely candidates for addition. 

While other Southeast Asia countries could eventually become important nodes in the AI supply chain, they still face constraints like a lack of infrastructure and dispersed talent, explains Anant Shivraj, a managing director and partner at Boston Consulting Group (BCG). 

Yet this could soon change, as Vietnam and Malaysia strive to become key hubs in the region, particularly in semiconductors and data centers.

“Pax Silica’s first wave is more focused on countries that can anchor long-term control, governance, and security across the AI stack,” says Shivraj. “Many countries play essential roles, and even if they are not part of the inner circle yet, that circle may well expand.”