Can Bessent’s Buyback Push Help Sketch a New US Financial Order?

1. Trump’s Iran-and-China Pressure Campaign Could Also Undercut Dollar Dominance

Oil prices were rising on Aug. 20, with West Texas Intermediate up 2.5% at $88 a barrel and Brent crude gaining 2% to $93.4. The move followed the Trump administration’s statement a day earlier that it would seek to economically isolate not only Iran, but also countries and companies that do business with Tehran.

That puts China at the center of the pressure campaign. China buys more than 90% of Iran’s crude exports.

Washington’s main goal is to cut off Iran’s funding. The US has already sanctioned some Chinese independent refiners, or teapot refiners, involved in trading Iranian oil. It has yet to directly target major Chinese banks that handle settlement for oil transactions between Iran and China.

The most powerful option would be to cut Chinese banks off from dollar payment channels. Chinese banks typically rely on correspondent accounts at US banks to transfer dollars internationally. If Washington restricts that access, those lenders could face severe limits on dollar payments, a core part of global trade. That would effectively force Chinese banks to choose between continuing business with Iran and preserving access to the dollar settlement system.

The fallout would grow if sanctions reached China’s biggest banks. Even if Chinese refiners want to buy Iranian crude, banks handling the transactions may refuse to process the payments for fear of losing access to dollar settlement. The broader US aim is to break the entire flow of funds: Iranian oil sales, purchases by Chinese companies, settlement by Chinese banks and the transfer of money back to Iran.

Using the dollar as a weapon could also backfire over time. China may conclude that heavy reliance on the dollar leaves it too vulnerable to US financial sanctions. That could accelerate efforts to expand yuan-based trade settlement and diversify reserves into gold and other currencies. The dollar’s share of global foreign-exchange reserves has already fallen to 57.8% in 2024 from 62.7% in 2017.

2. Bessent’s Financial-Order Vision Runs From Treasuries to Stablecoins

US Treasury Secretary Scott Bessent has become one of the most closely watched figures in finance. Treasury buybacks, the expansion of the Foreign and International Monetary Authorities repo facility, and the formalization of stablecoins may look like separate policies. Together, they can be read as an effort to manage the US government’s huge debt load and long-term borrowing costs while creating new sources of dollar demand.

The Treasury said on Aug. 19 that it would more than double the size of its buybacks of Treasuries with maturities of 10 to 30 years, raising the amount to at least $4 billion from a previous maximum of $2 billion per operation. The move came after the 30-year Treasury yield climbed above 5.3% this week, the highest level since 2007. After the announcement, the 30-year yield fell sharply and the 10-year yield dropped back into the 4.6% range.

Some investors view the move as a Treasury version of Operation Twist. If the department buys long-dated bonds while increasing issuance of short-term Treasury bills, the market gets less long-term supply and more short-term paper. It differs from quantitative easing, in which the Fed creates money to buy Treasuries, but it can still work to lower long-term yields. That strategy, however, requires enough demand to absorb the additional bill issuance.

That is where the GENIUS Act, a US stablecoin bill, fits in. Dollar stablecoin issuers must hold high-quality short-term reserve assets to maintain their pegs. As the stablecoin market grows, it could create a new source of demand for short-term US Treasuries used as reserves.

3. In AI Infrastructure, the Race Is Shifting From Securing GPUs to Turning Them On Quickly

Competition in AI infrastructure is changing. The key question used to be how many Nvidia GPUs a company could secure. Now the bigger challenge is how quickly those chips can be deployed with the power and data-center capacity needed to put them into use.

Nvidia is moving beyond GPU sales and acting as a matchmaker. CNBC reported on Aug. 20 that the company has recently connected firms in Northern Europe that have secured GPUs but need installation space with data-center operators that have power and space but need customers. The report said Nvidia has played a similar role in the US and Asia.

The company has a clear reason to focus on sites and electricity as well as chip shipments. Selling GPUs does not immediately start the AI computing that generates revenue. Even customers that secure billions of dollars of GPUs cannot run the equipment without data centers and power. The bottleneck in AI infrastructure has expanded beyond chips to include electricity, land and data-center capacity.

That is helping Northern Europe emerge as a new AI data-center hub. The region offers relatively easier access to power and large sites, and its cooler climate is an advantage for data-center cooling. Finland and Norway are seeing a wave of projects in the hundreds of megawatts. In Norway alone, data-center capacity waiting for grid connections totals 2.3 gigawatts.

Speed is especially important. Analysis by the Carnegie Endowment for International Peace found that if a 100-megawatt AI data center in the US starts operating one year late, its lifetime value drops 5.5%. A delay of 18 months raises that decline to 8.9%. By contrast, even if electricity prices double, the loss in value is only 4.5%. In other words, bringing a data center online a year late is more costly than paying twice as much for electricity.

The rapid pace of AI chip upgrades adds to the pressure. Even if a company buys the latest GPUs now, next-generation products could arrive while a delayed data center sits idle for more than a year. That leaves customers holding expensive hardware that produces no revenue even as its economic value declines.

The dynamic also works in Nvidia’s favor. The faster customers bring GPUs online, the faster they can launch AI services, generate revenue and improve returns on investment. That can lead to more GPU orders. Nvidia’s competitive edge is no longer limited to chip performance and CUDA. It is expanding into an ecosystem that connects power, land, data-center operators and customers.

Texas is also seeing a regulatory response to the spread of data centers. Governor Greg Abbott said as many as 1,800 data-center projects had been halted under his directive, adding that data centers should not consume water and electricity needed by local residents or push up power bills.

4. US-Listed ETFs Are on Track for $2 Trillion of Inflows This Year, With AI Money Piling Into Chips

The US ETF market is expanding at a record pace. Goldman Sachs Global Banking & Markets projects that inflows into US-listed ETFs will exceed $2 trillion this year, about 40% more than last year.

US-listed ETFs drew more than $1 trillion of net inflows in the first half alone. ETFs are no longer just vehicles that track indexes such as the S&P 500 or Nasdaq. They are increasingly used as core tools for portfolio construction and risk management.

Active ETFs are growing especially quickly. More than 35% of new ETF inflows this year have gone into active products. Active ETFs account for only about 13% of the roughly $16.1 trillion in assets managed by US-listed ETFs, yet they are taking more than 35% of new money.

The number of products is also surging. More than 1,100 ETFs were launched in the US last year, and this year is set to surpass that figure. Goldman Sachs expects the number of US-listed ETFs to top 6,000, more than the number of individual stocks listed on US exchanges.

Institutional investors are also making broader use of ETFs. Assets in model portfolios, which bundle multiple ETFs in advance, rose 46% over the past 12 months to $950 billion. That suggests asset managers and investment advisers are increasingly using ETFs to build entire portfolios instead of selecting individual stocks and bonds directly.

Within AI investing, the split in flows has been even starker. Semiconductor ETFs took in more than $19 billion of net inflows in June alone, the largest monthly total on record for the category. Software ETFs, by contrast, saw about $1.9 billion of outflows. Even amid the AI boom, investors are not buying technology stocks indiscriminately. They are pouring money into semiconductors while pulling it from software.

ETF trading now averages about $320 billion a day. During periods of higher volatility, ETFs can account for as much as 40% of total market trading. That is reinforcing the view that ETFs have become not just investment products but part of the market’s core investment infrastructure.

5. JPMorgan Says Nokia Is a Hidden AI Infrastructure Winner

JPMorgan has made a bullish call on Nokia. Its argument is that orders for the company’s networking equipment are rising quickly as AI data centers proliferate, but Wall Street earnings forecasts still do not fully reflect the benefit.

AI data centers need more than GPUs. They also require optical communications and IP networking gear capable of moving massive amounts of data quickly among GPUs and servers. Nokia is benefiting directly from that demand. Over the past two quarters, the company secured 3.8 billion euros of AI- and cloud-related orders. Of that total, 2.4 billion euros can be delivered within the next 12 months.

IP networking is emerging as a new growth driver. More than 40% of second-quarter orders, or more than 1.1 billion euros, were tied to IP networking. That shows AI data-center demand is rising quickly not only in Nokia’s traditional optical business, but also in IP networking.

The immediate constraint is supply. Component shortages mean Nokia cannot yet convert all incoming orders into revenue. JPMorgan expects that if supply improves in 2027 and 2028, the current backlog will begin feeding more meaningfully into sales and profit.

The bank also argues that current Wall Street estimates are too low. If the current pace of order growth continues, network infrastructure revenue growth in 2027 could be more than double market expectations. Even under a more conservative scenario in which order growth slows sharply, AI and cloud revenue in 2027 could still come in about 17% above consensus.

In a more optimistic case, operating profit could exceed market estimates by more than 28% in 2027 and more than 46% in 2028. JPMorgan maintained an overweight rating on Nokia and set a price target of 18 euros, or $21. That implied roughly 100% upside from the stock’s level at the time.

The bank’s logic is straightforward. AI data-center expansion is already driving a sharp increase in orders for Nokia’s networking gear. Those orders are piling up because of supply shortages. If production capacity improves in 2027 and 2028, the backlog could begin translating in earnest into revenue and profit. That is why JPMorgan sees Nokia as an overlooked beneficiary of the AI infrastructure buildout.

Park Shin-young, New York correspondent, Hankyung.com, nyusos@hankyung.com