Nomura’s Chief China Economist for China, Dr. Lu Ting, recently sat down with Tencent Finance’s interview program Economic Conversations (《经济大家谈》). During the interview, Dr. Lu discussed the evolving AI landscape, its macroeconomic implications, regional development, and the recent movements and future outlook for gold prices.

Given the intense international attention surrounding AI developments in both China and the United States —including the alleged transcript of an internal interview with DeepSeek founder Liang Wenfeng that went viral across China’s tech community on Thursday — I believe this interview is worth reading.

Tencent published a condensed version of the conversation on its WeChat account on July 21. I have translated it below. Please note that the views expressed are those of Lu Ting and do not necessarily represent my own. The notes were added by your Ginger River.

1. China Is Firmly the World’s No. 2 in AI, Trailing the U.S. by No More Than One Year

Economic Conversations: You have argued that the global AI landscape has effectively become a clear “G2.” Where exactly does China stand in this global AI supercycle?

Lu Ting: If I had to summarize it in the simplest possible way, China is in second place, behind the United States. We need to acknowledge that the U.S. is currently ahead in AI.

So today’s AI landscape is one where the United States leads, China follows closely behind, and each possesses its own comparative advantages. At the same time, the U.S. has imposed numerous restrictions on China’s AI development.

From the perspective of large language models, China has the world’s second-largest number of frontier models, exceeding Europe — the third-ranked region—by a wide margin. In terms of performance, I believe the gap between China and the United States is roughly six months, and certainly no more than one year.

At the same time, China has several unique strengths in AI.

The first is computing power and electricity.

China’s electricity supply is not only substantially larger than that of the United States but also considerably more stable. Looking ahead, China’s growth potential in this area is stronger, particularly because of its power grid infrastructure.

Initiatives such as the “East Data, West Computing” strategy demonstrate the country’s exceptional capabilities in large-scale infrastructure construction. In this respect, I believe the United States is at a disadvantage compared with China.

[Editor’s note: “East Data, West Computing” strategy is a national project aiming to channel data and computing demand from economically developed eastern regions to western areas with greater advantages in energy availability and land resources.]

China’s second advantage lies in integrating AI with manufacturing, including advanced robotics.

China possesses a more complete industrial supply chain in these sectors, giving it an edge over the United States in certain areas.

A third advantage is China’s long-term accumulation of AI talent.

Over many years, China has built a strong higher education system that produces large numbers of graduates in computer science, mathematics, statistics, and other STEM disciplines. At a minimum, China is second only to the United States in cultivating this talent.

If we look specifically at undergraduate graduates who ultimately become top-tier AI researchers, China may even produce substantially more than the United States.

Taken together, these factors give China exceptionally favorable conditions for AI development and explain why the world has effectively entered a clear G2 AI landscape.

2. Market Optimism Doesn’t Necessarily Mean There’s a Bubble; Economic Analysis Should Be Independent of the Stock Market

Economic Conversations: You’ve estimated China’s broad AI capital expenditure at roughly RMB 1.2 trillion (about 177 million U.S. dollars) in 2026, yet market sentiment appears far more optimistic than those figures suggest. Do you think expectations surrounding AI have become a bubble?

Lu Ting: Not necessarily.

Market sentiment is rarely driven by these kinds of underlying numbers. Instead, it tends to be shaped by factors such as stock market performance — particularly the rapid share-price appreciation of leading AI companies after they go public.

Investor sentiment and the kind of sober quantitative estimates that economists produce are two entirely different processes. There’s no contradiction between them.

The reason we conduct these calculations is that we need to think carefully about how much AI can realistically contribute to China’s economy over the short business cycle.

We should neither underestimate nor overestimate its impact.

Economic analysis should remain independent of sentiment in the stock market.

If we look at the United States, AI investment is indeed far larger than it is in China, and consequently its influence on both the U.S. stock market and global equity markets has also been much greater.

In China, both the relative scale and the absolute amount of AI investment remain considerably smaller than in the United States. As a result, AI’s influence on China’s stock market is, in my view, more limited.

That said, over the past several months there have been few other standout investment themes. AI has become the dominant global narrative, so naturally most market attention has gravitated toward it.

3. Soaring Chip Prices Have Worsened China’s Terms of Trade; Technological Self-Reliance Ultimately Depends on Breaking Through in Lithography

Economic Conversations: You’ve pointed out that China’s AI investment remains highly dependent on imported high-end chips, creating certain spillover effects. On the other hand, could this dependence also be accelerating China’s efforts to build an independent semiconductor supply chain? To what extent will progress in domestic chip substitution determine how much room China’s AI investment ultimately has to expand?

Lu Ting: I think this is an extremely important question.

First, from the perspective of the short-term macroeconomy, we need to remain objective.

The reason I emphasize this is that some observers have looked at China’s recent export growth — which exceeded 19% in May — and concluded that AI has dramatically boosted China’s economy through stronger external demand.

I believe we should be more cautious.

It is true that the 19% export growth is closely related to AI. One could even argue that roughly half of that increase was AI-related.

However, much of that contribution came through price changes, rather than volume.

Over the past year, global memory-chip prices have risen sharply.

In fact, China’s exports of memory chips are not directly tied to AI. Instead, many leading overseas semiconductor manufacturers shifted production capacity toward more advanced AI-related memory products, creating indirect demand for the types of memory chips produced by Chinese companies.

At the same time, when evaluating exports we also need to examine imports.

Exports make a positive contribution to China’s economy. Imports are different. When considering imports, we need to subtract China’s own AI investment from the equation.

China is the world’s largest importer of semiconductors and remains heavily dependent on other economies for high-end chips.

While export growth has attracted considerable attention, imports have also been expanding rapidly, with growth exceeding 25% over the past two months.

Taken together, from the perspective of international trade, the global AI-driven surge in semiconductor prices has actually worsened China’s terms of trade rather than delivering significant net benefits to the broader economy.

As for the implications for China, under the current G2 landscape, it is inevitable that the United States and some other developed economies will continue imposing restrictions on China’s AI development, primarily through controls on advanced chips.

I don’t expect any significant improvement over the next several years.

This makes China’s development path fundamentally different from that of many other countries.

It means China has to place greater emphasis on technological self-reliance.

Achieving that won’t be easy.

It will require a great deal of time, sustained effort from many people, and breakthroughs at critical technological bottlenecks.

At the same time, we cannot afford to become impatient.

Economic Conversations: In your view, what is the single most important breakthrough that China still needs?

Lu Ting: Assuming there isn’t a major technological paradigm shift, I think the most critical bottleneck remains lithography machines.

That means China needs to pursue two objectives simultaneously.

First, it should continue participating in the global industrial chain as much as possible.

Improving relations with other countries and importing more machinery, equipment, intermediate components, and materials remains both necessary and beneficial.

After all, other countries continue to innovate, and importing these technologies can shorten China’s own research and development process.

Second, China cannot become overly dependent on imports.

We ultimately need to achieve original, breakthrough innovations of our own — much like the “Two Bombs, One Satellite” program did decades ago.

[Editor’s note: “Two bomb, One satellite” refers to China’s nuclear bombs, missiles, and man-made satellites that were independently developed by Chinese scientists and engineers from the 1950s to the 1970s. “Two bombs and one satellite”, a well-known term in China, has become a valuable spiritual asset for the Chinese people, especially those working in the scientific and technological fields.

4. Lu Ting’s Biggest Concern: AI Could Deepen a “Double K-Shaped” Divergence

Economic Conversations: You’ve argued that the benefits of AI are currently concentrated in just a handful of cities, such as Beijing, Shanghai, Shenzhen, and Hangzhou. Such concentration seems fairly typical during the early stages of technological revolutions. As AI moves into broader commercial deployment, how do you think these gains could spread to more cities?

Lu Ting: This question of whether AI’s benefits will spread is actually one of my biggest concerns.

Historically, new technologies have generally generated spillover effects over time. But we shouldn’t assume that because previous technological revolutions unfolded this way, the AI era necessarily will as well.

Sometimes the world changes.

That’s why I’ve recently proposed what I call a “double K-shaped divergence” framework, one that also incorporates China’s property sector.

Why?

Because the downturn in real estate has already produced two distinct K-shaped patterns in China.

The first is the widening gap in income and wealth among households.

The second is the widening gap between large cities—particularly first-tier cities—and the rest of the country in terms of economic performance, fiscal revenue, wealth, and even housing prices.

I frame it this way because I believe the arrival of the AI era is likely to make both of these K-shaped divergences even more pronounced.

Moreover, the two trends reinforce each other.

AI is likely to widen income and wealth disparities across individuals while simultaneously increasing disparities between cities. And the cities that benefit most are almost exactly the same ones you just mentioned —Beijing, Shanghai, Hangzhou, and Shenzhen, where AI resources are already highly concentrated.

Why is that?

Because AI demands exceptionally high concentrations of talent, capital, and technology, causing these resources to become even more concentrated in a small number of leading cities.

But there’s another side to the story.

Once these resources are concentrated in those cities, the services they provide to the rest of the country may become substitutes, rather than complements, for local economic activity.

Take the internet era as an example.

Hangzhou had Alibaba, while Beijing had JD.com.

These platforms enabled countless businesses to sell products online, helping realize Alibaba’s vision of “making it easy to do business anywhere.”

As Alibaba grew, many smaller cities and rural areas also benefited from those platforms. The gains spilled over.

The AI era may look very different.

If foundation models developed in Beijing, Shanghai, Hangzhou, or Shenzhen provide services directly to lower-tier cities, those cities may purchase the services—but in doing so, they could also end up replacing local jobs.

I’m not trying to sound alarmist.

I’m simply pointing out that this is a possibility, and it’s something policymakers around the world are already worried about.

China deserves particular attention because the property downturn has already widened the gap between first-tier cities and the rest of the country.

AI could make that divergence even more severe.

Ultimately, what does this lead to?

The greater the divergence becomes, the more likely it is to weaken aggregate demand—whether consumption or investment.

That’s why I believe AI undoubtedly has positive effects on the economy.

But if it simultaneously amplifies K-shaped divergence and ultimately suppresses aggregate demand, then some of those positive effects will inevitably be offset.

[Editor’s note: On July 20, a commentary published by a People’s Daily-affiliated WeChat account argued that China’s ongoing K-shaped recovery should be viewed as a normal feature of economic restructuring rather than a cause for excessive concern.

It wrote that while upgrading the economy inevitably brings short-term divergence, “temporary fluctuations should not be overstated, nor should differences in people’s lived experiences be overinterpreted.” The commentary added that the transition from old to new growth drivers is “an iterative process rather than one of outright replacement,” with new industries creating fresh sources of growth while continuously revitalizing existing sectors.

Looking further ahead, the article expressed confidence in China’s long-term economic trajectory, arguing that continued industrial upgrading and the broader adoption of new technologies across industries would ultimately lead to “more balanced and more inclusive development,” benefiting both businesses and households.]

Economic Conversations: So overall, you’re relatively more concerned than optimistic about whether AI’s benefits will spread?

Lu Ting: Yes, I do have concerns.

If we compare AI with earlier waves of technological development — the internet era, the rise of e-commerce platforms, the real estate boom, or even the expansion of what China calls the “new three” industries (electric vehicles, batteries, and solar photovoltaics) — I believe AI’s spillover effects from first-tier cities to second-, third-, and fourth-tier cities are likely to be much weaker.

In fact, instead of generating positive spillovers, AI could in some cases produce what I would describe as a siphoning effect or a substitution effect.

5. Wuhan, Hefei, and Suzhou May Benefit from Earlier Investments in Chips—But Their Moats May Not Be Deep Enough

Economic Conversations: We’ve seen cities like Hefei establish an early foothold in semiconductors and robotics, while Wuhan has built strengths in optics and sensors. Compared with previous technology cycles, what is fundamentally different about the way AI is reshaping China’s regional economy? Can cities such as Hefei and Wuhan truly use this wave to move up the industrial value chain?

Lu Ting: The first beneficiaries will undoubtedly be the cities that are home to leading foundation models. Beijing, Shanghai, Hangzhou, and Shenzhen are the clear winners.

A second group of cities will benefit through their positions in the hardware supply chain.

Based on our research and data analysis, I’ve highlighted three cities in particular: Wuhan, Hefei, and Suzhou.

These cities stand to benefit from the AI supply chain largely because of investments they made years ago. In some cases, today’s gains are indirect consequences of those earlier strategic decisions.

Take memory chips as an example.

At present, China’s two major memory-chip manufacturers are located in Wuhan and Hefei.

Their gains from AI are largely indirect rather than direct.

Leading global semiconductor companies have redirected production capacity toward more advanced AI-related high-bandwidth memory (HBM) chips, creating additional demand for products made by Chinese memory-chip manufacturers. That has, in turn, benefited these two cities.

From this, I would draw two conclusions.

First, beyond Beijing, Shanghai, Hangzhou, and Shenzhen, there are indeed other cities that are benefiting from investments they made years ago in semiconductors and other AI-related hardware.

At the same time, however, we need to view these gains with caution.

It’s important to remain optimistic and recognize that some cities are genuinely benefiting.

But we should also recognize that the competitive moats protecting these advantages may not be particularly deep.

Cities shouldn’t assume that because they are benefiting today, they will continue benefiting throughout the AI era.

They need to keep moving forward and continue strengthening their capabilities.

After all, even cities like Wuhan and Hefei could themselves become subject to the same siphoning effects in certain areas, and some of what they manufacture today could eventually be replaced by competitors.

Economic Conversations: How can a city determine whether it truly possesses a strong competitive moat?

Lu Ting: The most important question is how difficult it is for others to enter that particular industry.

The experience of China’s “new three” industries over the past several years offers a useful example.

Take electric vehicles.

Eventually, cities across China all rushed into battery manufacturing. At one point, there were more than 60 companies nationwide producing power batteries.

Solar photovoltaics saw an even greater influx of investment.

These industries certainly had competitive barriers — but those barriers weren’t especially high.

What happened in the end?

After an initial period of prosperity, investment became excessive, production capacity became highly redundant, and prices eventually collapsed.

Solar panels and batteries fell by roughly 85% from their peak prices.

As a result, many investments that had once appeared highly profitable eventually turned into loss-making assets and financial burdens.

Those are lessons we need to learn from.

That said, I believe semiconductors — including memory chips — have significantly higher barriers to entry than the “new three” industries.

But that still doesn’t mean these businesses will never face competition or new challengers.

We need to remain realistic.

China must continue climbing the technological ladder.

At the same time, we shouldn’t mistakenly believe that China is already approaching the global technological frontier in semiconductor manufacturing or memory chips.

The gap remains substantial.

Economic Conversations: In the AI competition between China and the United States, the U.S. has largely pursued a strategy built around technological barriers, while China has increasingly emphasized open-source development. Which approach do you think will prove more sustainable? And when do you believe China might achieve a genuine breakthrough in cutting-edge foundation models?

Lu Ting: I think both approaches are likely to coexist for quite some time.

The key question isn’t which model is ultimately superior.

Rather, given the realities China faces, the important thing is how to make the best possible use of the path it has chosen.

It’s very difficult to say which approach will prove better in the long run.

In fact, there may ultimately be no need to compare them at all.

Each country is operating under a different set of constraints, and each is choosing the strategy that best fits its own circumstances.

6. Gold Has Entered an Era of High Volatility Without a Clear Anchor

Economic Conversations: Gold remained remarkably strong throughout 2025 despite high interest rates, leading many people to argue that the traditional pricing framework based on real interest rates had broken down. But in 2026, as U.S. real interest rates continued rising between March and May, gold prices retreated noticeably from their highs, and the negative correlation between the two appears to have re-emerged. How do you interpret this shift?

Lu Ting: I think the most important change didn’t occur this year. It happened over the past three years.

From the beginning of 2022 through 2025, the global economy underwent profound changes.

Following the pandemic came a period of elevated inflation, but beneath that was an even more important structural factor: persistently high fiscal deficits across developed economies.

Ordinarily, once the pandemic ended, one would have expected those exceptionally high fiscal deficits to return relatively quickly to more normal levels.

Take the United States as an example.

During the pandemic in 2020, the fiscal deficit reached about 15% of GDP. In 2021, when the pandemic had largely subsided, it still stood at around 13%. Since then, it has remained in the range of 5% to 6%.

That’s far above the roughly 3% average before the pandemic.

More importantly, there is now a broad consensus that the U.S. fiscal deficit is unlikely to return to 3%.

That represents a major structural shift.

The same is true elsewhere.

Europe, for example, has had to contend first with the Russia–Ukraine war and the resulting energy crisis, and now faces another challenge. As relations between the United States and Europe have become more strained, European governments are under growing pressure to increase defense spending.

As a result, fiscal deficits in Europe are also likely to remain elevated.

In other words, the United States now operates with deficits closer to 6% than 3%, and similar trends can be seen across other advanced economies.

Markets have therefore fundamentally reassessed this new reality.

During that reassessment, government bond yields rose significantly across major developed economies — including the United States, Europe, and Japan — with the sharpest increases occurring between 2022 and 2024.

It was during precisely that period that gold prices climbed rapidly.

Traditionally, we would have said that rising U.S. Treasury yields — whether nominal or real — are generally negative for gold.

But this time, investors increasingly concluded that persistently large fiscal deficits would eventually force governments to rely on higher inflation and currency debasement to manage their debt burdens.

That was the fundamental shift.

So if we want to explain gold’s performance, the key lies in what happened during those years.

That structural adjustment has now largely run its course.

It’s not that conditions have stopped changing altogether, but rather that they have reached a new equilibrium.

Gold has therefore entered a new price range.

Instead of trading between US$1,000 and US$2,000 per ounce, it may now remain above US$3,000, or even reach US$4,000 or US$5,000.

Exactly where it ultimately settles will largely depend on market psychology.

Some analysts have projected prices as high as US$6,000 or US$8,000, while others believe gold could fall back to US$3,000 or US$4,000.

All of those outcomes remain possible.

That’s why I believe gold has entered an era of high volatility.

Economic Conversations: What’s the single biggest reason behind this year’s decline in gold prices?

Lu Ting: The simplest explanation is that prices had risen too far.

Gold has effectively lost its anchor.

When gold traded between roughly US$1,000 and US$2,000, those levels were broadly anchored by production costs.

After the structural repricing that took place over the past several years, we’ve entered an environment where that anchor has largely disappeared.

Once you’re in such an environment, large price swings become entirely understandable.

There’s no need to search for overly complicated explanations.

We should also avoid a common mistake in reasoning.

When people are bullish on gold, they often assemble a long list of arguments to justify further gains.

Then, when prices fall, it suddenly seems as though all of those arguments have disappeared.

In reality, they don’t need to.

The real explanation remains the structural shift that occurred between 2022 and 2024, when developed economies — particularly the United States — fundamentally revised their expectations regarding future inflation and long-term fiscal deficits.

Economic Conversations: Against today’s global economic backdrop, how do you view the outlook for gold?

Lu Ting: For more than a year now, whenever people have asked me about gold, the first thing I’ve emphasized is volatility.

I don’t believe gold has a particularly clear long-term trend at this point.

Second, from an investment perspective—particularly for overseas investors—the appropriate benchmark should be U.S. Treasury yields.

Suppose the yield on the 10-year U.S. Treasury is around 4.5% to 4.6%.

Gold would need to appreciate by roughly that amount every year simply to match the return on holding Treasuries.

In other words, if gold rises by only about 4.5% annually, it hasn’t really generated excess returns—it has merely kept pace with what is generally regarded as one of the safest investments in the world.

The key is understanding your benchmark.

Using a simple rule of thumb, dividing 72 by 4.5 tells you it would take roughly 16 years for your investment to double.

Only returns above that threshold would represent a meaningful advantage over Treasuries.

Third, in this environment, investors should avoid using leverage when investing in gold.

We’re now living in a period of high volatility, and leverage can significantly magnify investment risk.

Economic Conversations: Given the combination of rapid AI development, questions surrounding the U.S. dollar’s credibility, and intensifying geopolitical competition, what do you see as the biggest upside and downside risks for gold over the next two years?

Lu Ting: If gold experiences another significant and sustained rally over the next two years, I believe the most likely catalyst would be a further deterioration in fiscal deficits across developed economies, including the United States.

If that doesn’t happen — if U.S. deficits remain broadly around the 5% to 6% level that markets currently expect — then we may not see another prolonged surge in gold prices.

The biggest downside risk is essentially the mirror image.

If, over the next several years, the United States performs exceptionally well — for example, if the AI revolution delivers substantial productivity gains and fiscal deficits fall toward 4%, 3%, or even lower — then gold prices could decline significantly, perhaps even experiencing the kind of major correction seen in previous historical cycles.

Ultimately, if you’re trying to understand the future trajectory of gold prices in U.S. dollars, the most important variables are the Federal Reserve’s policy stance, the U.S. federal government’s fiscal deficit, and above all, market expectations regarding future deficits, inflation, and monetary policy.

Subscribe now

Ginger River Review

Nomura's Chief China Economist on AI's…