This is an edited version of the opening remarks GIC CEO Lim Chow Kiat gave at GIC Insights 2026 on 17 November 2026 in San Francisco.

This year marks 40 years since GIC opened its San Francisco office. Our first investments here were not even in technology, but in California farmland and buildings. We soon moved into venture capital, and nearly 10 years ago, we set up a dedicated Technology Investment Group. At the time, we were hopeful of the ‘A, B, and C’ technologies, or AI, blockchain, and cloud computing. I cannot say we foresaw the AI explosion that followed, but staying close to new technologies has served us well.

This year’s theme is ‘What If’. Every investment comes with some. Let me start with one I suspect is top of mind:

What if intelligence became scalable and plentiful?

In many ways, we are well on the way. The frontier keeps advancing, and the cost of reaching a given level of performance from an AI model has been falling roughly tenfold a year1.

But getting there would require overcoming significant physical, financial, and regulatory challenges.

The supply of chips, power, grid connections, cooling, and skilled labour is constrained. Data centres’ electricity use could triple by 20352, much of it where grids are already congested.

Financial capital is another bottleneck. To put the scale of the build-out into perspective, the capital spending of just five technology companies, mainly on data centres, is larger than global investment in oil and gas production3. Operating cash flows are no longer enough, so firms are turning to credit markets to raise the trillions required. The largest technology companies have issued more than twice as much investment-grade debt this year as in all of last year4.

But few borrowers dominate the space and many lenders are already heavily exposed to the same names. To take on more, they want to be paid more, and credit spreads have widened meaningfully. More financing solutions are emerging, as the build-out increasingly competes with governments for the same pool of capital.

Regulation could pose yet another hurdle. Securing permits to build data centres requires careful site selection and negotiation, and the process varies by geography. For investors, delays could erode returns.

Assuming these challenges can be overcome, I see three consequences that matter for investors: we will use far more intelligence, friction will fall away, and the rules will change.

We will use far more intelligence

The first consequence is that we will use far more intelligence, as the Jevons paradox would suggest. The total addressable market is vast. Consumer AI spending tripled in the past year to US$40bn5, while enterprise AI spending is approaching US$200bn6. With much lower token costs and agentic AI, multiple S-curves could eventually emerge across industries. Governments are also likely to use more AI, driven in large part by national security and the desire to control critical AI capabilities.

For now, however, use is spreading unevenly within each segment. Our portfolio company Ramp, which tracks spending at more than 85,000 US businesses, found that among firms paying for AI, the top 1% spends more than US$10,000 per employee per month. The typical company spends about US$167.

Enterprises are climbing an adoption S-curve, and the gains typically come in stages: first productivity, then the harder work of redesigning how things are done, and new revenue last. Most companies are still at the start. After an initial burst of experimentation, the later stages take far more time and effort, as the supporting systems need to be put in place. A Deloitte survey found that only 30% of organisations were redesigning their processes around AI8.

AI can also create its own problems. As AI exposes cyber vulnerabilities, some of the savings, for example from coding, may go into fixing the security weaknesses that AI-written code introduces, leaving a low net return on investment.

The path of adoption is unlikely to be smooth, even if intelligence proves transformative. Investors need to watch closely as these S-curves develop, both to identify industries poised for lift-off and to spot winners and losers among individual businesses.

Friction falls away

The second consequence is that friction falls away. A great deal of economic value has traditionally relied on friction: the time it takes to find information, the cost of switching providers, work billed by the hour, differences in time zones, and layers of intermediaries. AI is removing much of it, but unevenly.

In February, software stocks fell sharply on fears that AI agents would replace much of what these companies sell. Share prices have since swung both ways—a reminder to separate market sentiment from business fundamentals. The longer-term disruption question remains and should be assessed company by company.

Parts of software have remained resilient. Every agent a company deploys needs reliable data, business context, and controls over what it can access and do, which drives demand for cloud platforms, data infrastructure, and cybersecurity. More broadly, software holds up best when it is deeply embedded in a customer’s operations, keeps its authoritative records, or runs regulated workflows. These are real moats. Even so, if competitors’ AI agents become the layer that orchestrates the work, margins could come under pressure.

Friction falls more slowly where the bottleneck is physical or regulatory. In healthcare, AI is speeding up and lowering the cost of drug discovery and pre-clinical research, but it has yet to show that it improves the odds of success in the later clinical trials, where most drug candidates fail.

Businesses and investors should stay alert to products and services that rely on friction. The intermediation risk could be very high.

The rules will change

The third consequence is that the rules will change around competition, jobs, and who is responsible when AI gets it wrong. When a technology reaches this deep into the economy, governments eventually respond. I do not know how, but I would watch three developments.

If a few companies end up controlling the most capable systems, will governments rely on competition to keep them in check, or regulate them like utilities? If AI lifts productivity but thins out entry-level jobs, who pays for retraining: employers, taxpayers, or the companies selling the technology? And when an AI agent makes a costly mistake, who is liable: the company that built the model, the one that deployed it, or the customer who relied on it? Each answer moves value from one group to another, and investors will have to price that in.

AI has become a national priority. Export controls, industrial policy, and where data centres are built are now strategic decisions. For decades, the world economy optimised for cost. Increasingly, companies and governments are paying for certainty: supply they can rely on, energy they can secure, and rules they can plan around. These choices will shape returns as much as the technology itself.

What if progress stalls?

All of this depends on the technology continuing to get better, cheaper, and more pervasive. But what if progress stalls? Many in Silicon Valley believe AI will transform the economy within a decade. History suggests it may take longer: electricity took about 40 years to translate into factory productivity. If that is the path, three things would follow.

First, equity markets. The ten largest companies in the S&P500, most of them closely tied to AI, now make up nearly 40% of its value9. Several invest more than their businesses generate in cash and fund the difference with bonds and new shares. Investors have gone along with this on the promise of faster growth. The sell-off of recent weeks has shown how quickly that confidence can be shaken10. If growth is delayed, valuations would be tested, and so would the many portfolios that own these companies.

Second, credit markets. New borrowing funds about a third of the largest cloud companies’ capital spending, up from under a tenth in 202411. Strain would likely show first among smaller, more indebted builders. And because many investors now own the same companies’ shares and bonds, bonds may cushion portfolios less than they used to.

Third, winners and losers would shift. Companies that reorganised around AI would find the savings smaller or slower than planned. Businesses that look most at risk of disruption today would gain breathing room.

So the question for investors is not just whether AI succeeds, but what is already in the price. A strong company can be a poor investment if its price assumes everything goes right. A threatened one can be a good investment if its price assumes the worst.

What an investor should do

What should a long-term investor do when both scenarios are plausible? At GIC, we rely on three principles: diversification, granularity, and agility.

To diversify, we seek out high-quality businesses outside AI, including in healthcare, consumer, and industrials, to give us other sources of earnings. But given how much of the market AI now accounts for, there is a limit to how far we can move away from it. That makes the quality of what we own all the more important.

Granularity means judging companies one at a time rather than buying the broad theme. We map AI across the value chain of enablers, monetisers, and adopters, and test each company on its moat, management, and momentum.

Our public and private equity teams apply the following questions to every company, not just technology firms: Does AI expand its demand and pricing power, or erode them? Does it raise barriers to scale, or lower them? Does it deepen its competitive advantage, or render it obsolete? And is management adopting AI with discipline and reinvesting the gains?

Agility requires keeping sufficient portfolio liquidity and watching the signals to know which scenarios are unfolding. Our teams in the US give us a front-row seat: how much AI is actually deployed, how fast the models improve, what customers gain, how fully data centres are used, and what lenders charge to finance the build-out.

Valuations will keep swinging between enthusiasm and caution. Our job is to stay focused on value, be selective in where we deploy capital, and stand ready to act when others hold back.