DATA CENTER WATCH

Microsoft presses Trump to scuttle ban on Chinese data center tech

Microsoft and other companies and trade groups have pressed the Trump administration to halt plans to bar imports from China of a key component used ‌in AI data centers, three people familiar with the situation said.

Reuters reported in August that the Trump administration wants to ban US imports of new models of Chinese optical transceivers, which allow data to travel over fiber-optic cables at the speed of light within data centers. The concern is that they could be used by Beijing to steal data, install malware or disrupt service at US data centers.

Microsoft has 35 data centers under construction in the US ​and another 134 planned, according to Cleanview, a market intelligence company which tracks data center development. It is one of a group of companies and trade groups making the case ​to the Trump administration that such a ban could irreparably damage the US edge over China in the AI race by further slowing the ⁠data center rollout, the sources said. (…)

“It really threatens to arguably grind (the data center rollout) ​to a halt,” said John Miller, Executive Vice President of Policy and General Counsel of the Information Technology Industry Council (ITI), a tech lobbying group, which has publicly urged the administration not to move forward with a ban.

He said many cloud companies have already purchased the AI chips needed for the buildout. “Those chips, without optical transceivers literally could just sit there waiting around un-utilized until you actually have the transceivers to essentially connect them together,” he added.

Some ​data centers need millions of optical transceivers to connect servers, routers and switches.

The Federal Communications Commission, which oversees the US telecom industry, is working on the ban. (…)

Beijing “firmly opposes the U.S. overstretching the concept of national security and abusing state power to unjustifiably suppress Chinese companies,” the Chinese embassy in Washington said in a statement. “China will continue to firmly safeguard the lawful rights and interests of Chinese companies.”

China boasts two-thirds of the global market for optical transceivers, according to Counterpoint Research. Chinese firms like Eoptolink and Zhongji Innolight, which was added to the Pentagon’s list of Chinese military-backed companies, dominate global supply, counting top American hyperscalers among their ​customers.

US-based Coherent and Lumentum produce technology that competes ​with them but lack the scale to ⁠replace Chinese vendors, according to a report by the Foundation for American Innovation.

“A Pentagon-listed Chinese military company, including Zhongji Innolight, should not be wiring America’s data centers,” the House Select Committee on China, led by Republican John Moolenaar, said in a social media post following the report. The FCC should “move ​forward and keep new (China)-linked optical transceivers out of our networks,” the committee added.

ITI sent a letter to ⁠the FCC on August 13 calling on the telecoms watchdog to “refrain from adding optical transceivers” to a trade blacklist, emphasizing the “potential effects that supply disruptions could have on data center deployment and US AI leadership.”

Morgan Stanley said in an October client note that the Trump administration was considering barring Chinese optical transceivers capable of handling 3.2 terabits of data per second, whose production is not expected to ramp up until ⁠2028, with widespread ​installation expected in 2029.

Such a threshold could avoid antagonizing US companies and Beijing while giving US optical companies time to ​expand production.

The FCC has not addressed the optical transceiver issue publicly, but two of the sources said the 3.2 terabit per second threshold is under consideration.

Lawmakers may not want to wait. In September, a bipartisan group of US lawmakers introduced legislation ​to bar the federal government from equipping sensitive government systems with Chinese-made optical transceivers.

 

 

(JP Morgan Asset Management)

AI Boom Helps Fuel China Exports, Pushing Trade Surplus Above $1 Trillion

Through the first 10 months of the year, China’s surplus is nearly 9% larger than it was during the same period of 2025, when it posted a record $1.2 trillion (…).

Chips and computer parts accounted for 40% of the year-on-year increase in Chinese export value so far, reflecting soaring prices as a result of a building frenzy around data centers.

In volume terms, growth came from a broad range of high-tech products, ranging from the “New Three” sectors promoted by China’s leadership — solar panels, electric vehicles and batteries — to industrial robots, according to a Goldman Sachs analysis.

Other signs emerged today that the AI boom is continuing to fuel exports from Asia. TSMC, the Taiwanese chipmaker that’s an important supplier to companies like Nvidia and Apple, posted a 60% increase in monthly sales. (…)

The main suppliers of AI-related goods are emerging markets, with Taiwan, Mexico, Vietnam and Thailand accounting for more than 80% of US imports of AI-related products, according to a report this week from Oxford Economics. (…)

Meanwhile, US officials in Washington are voicing concerns again over Beijing’s curbs on a different group of key exports — critical minerals. Those restrictions and added bureaucracy are limiting the availability of products used in everything from aerospace and defense to semiconductors and magnets for electric vehicles. (…)

Prices are reflecting the scarcity of some strategic minerals outside China. Yttrium, used in electronics and turbines, is a staggering 125 times more expensive in Europe than in China, according to price reporting agency Argus Media.

A ton of the metal costs about $1.15 million compared with just $9,144 in China, where supplies remain ample.

China’s total exports of yttrium oxide are running at about two-fifths of the average level for 2023-2024. Almost none has been exported to Japan this year, and shipments to the US are about 70% lower than before 2025.

CAPITAL IS CAPITAL

Foreign demand for US Treasury auctions sees biggest drop since 2016 in benchmark 10-year

A day after Monday’s 3-year Treasury auction tailed (yield above expectations), today’s benchmark $42 billion 10-year sale sees biggest tail since February, as indirect bidders (which include foreign buyers) posted their biggest drop since 2016. (…)

The 5% coupon is the highest since 2006, per ZeroHedge.

Bid-to-cover: 2.46, lowest since May, vs. 2.77 in October and a 2.60 six-month average. (…)

ZeroHedge called it “downright ugly,” with foreign buyers staging “a bit of a strike.” (…)

Oaktree Capital’s Howard Marks writes (my emphasis):

AI-related debt issuance has totaled nearly $600 billion globally so far this year, according to Goldman Sachs, comprising debt issued by the hyperscalers, in data-center/project financings, and by other AI-linked borrowers. Around $400 billion of this went to the U.S. investment grade debt market, representing over one-fifth of total IG supply and taking the share of AI-related bonds in the Bloomberg USD IG Corporate index to around 12%.

For the U.S. high yield bond market – into which the hyperscalers don’t directly issue – the weighting is more like 6%.

Clearly, AI-related debt issuance is one of this year’s major capital-market developments.

Going forward, AI seems likely to require an average of $1 trillion of new capital per year in the period ahead, and hundreds of billions of that are likely to continue to be raised each year through the issuance of debt. What will be the impact on interest rates of AI’s demand for debt capital?

Since expanding capacity and developing new products are imperatives for the hyperscalers and frontier labs – and since they don’t want to see competition crop up – these companies likely can’t afford to slow their investment in AI just because interest rates may be higher than they expected. That means they have to be “takers” of financing, even if it is in short supply and increases in cost.

In the past, economists used to talk about “crowding out,” in which aggressive government borrowing makes it hard for the private sector to find the debt capital it needs. At present, it feels to me that the demand for debt from rate-insensitive hyperscalers willing to pay whatever it takes has the potential to elevate the interest rates the government must pay. Hasn’t that contributed to the rate rise we’ve seen to date?

In the same vein, will the equity markets be capable of providing the hundreds of billions of new equity investment required each year, supplementing the companies’ cash flow? The amount raised in traditional U.S. IPOs was below $50 billion in eight of the last ten years (the exceptions being almost $80 billion in 2020 and $140 billion in the boom year of 2021).

I suspect there’s just so much new capital available in a given year; won’t AI’s appetites for both debt and equity ultimately compete against each other for it?

In the past, the highly profitable non-AI businesses of the hyperscalers produced massive free cash flow that rendered them debt-free, cash-rich, and highly rated. Things are different now, and these companies are spending their cash hoards and current cash flow, and they’re borrowing more. Their strong investment-grade ratings enable them to easily finance their expansion in AI.

But as this process wears on, might their creditworthiness and/or ratings change? What are the possible implications for those who lend against the high ratings today and watch as the companies increase their leverage?

Some of the data-center financing taking place is facilitated by the investment-grade ratings of the hyperscaler tenants. Other debt is collateralized by chips. But investors have to be very conscious of “out clauses” in some data-center leases that can enable tenants or guarantors to terminate their obligation under certain conditions – such as the failure of the center to operate at 99.9% reliability – and of the potential risk that collateral in the form of chips won’t last as long as the debt.

What are the implications of the industry’s so-called “circular transactions,” in which companies finance purchases of their products? We see frequent instances of lending and related buying between chipmakers, hyperscalers, frontier labs, and data-center builders. Sometimes money goes from A to B and back to A. Sometimes it goes from A to B, and then to financing for C, which spends it with A.

Are these transactions the parties would enter into on an arm’s length basis, or are vendors subsidizing buyers to support demand and protect market share for their products? What are the risks associated with this interdependence?

And what is the impact on the hyperscalers’ creditworthiness of borrowing on the part of special-purpose vehicles that keep debt off balance sheets – is the parent behind the debt or isn’t it?

Finally, what are the effects of borrowing in a time of exuberance surrounding an exciting technological advance? In the end, we have the confluence of several factors:

• a massive desire to borrow on the part of AI companies,
• investor excitement surrounding AI, and thus a willingness to lend,
• heavy concentration on AI in the capital markets, as described above, and
• so-called “circular transactions” between participants in the AI industry.

Lenders to AI face a new era, marked by debt concentrated in a single industry, rapid change in industry fundamentals, and a high degree of reliance on the level of future revenues from sales of compute. Lenders will have to be exceptionally vigilant regarding the creditworthiness of the borrowers they back and the structure/terms through which they do it.

Howard then wonders if AI is a bubble:

First, we need a definition of a bubble, and I want to supply mine: irrational exuberance leading to imprudent behavior, excessive investment, and too-high prices for participation. I think an AI bubble, like any possible bubble surrounding a technological advance, has to be considered in at least three ways.

• Is the technology itself overhyped and likely to disappoint?
• Are the amounts companies are spending excessive and fated to produce inadequate returns?
• And are investors taking too much risk and paying too much for stakes in this field?

I still think AI itself will prove powerful and consequential and change corporations, the economy, and life speaking broadly in ways we can’t even imagine. So no, the technology isn’t an illusion.
A great deal of discussion is taking place regarding the revenues that will have to be generated in the future to justify the capital expenditures. But until we know something about how AI will be used and who will profit, we can’t assess the wisdom of the companies’ spending.
And until knowledge of these things enables us to reasonably estimate the companies’ earnings and the profitability of the infrastructure, we can’t judge the appropriateness of the prices being paid for stocks or the yields being offered on debt.

(…) the key boils down to this rather complex one: will AI provide sufficient value to users to warrant sales of compute and prices for compute high enough to result in adequate returns on the trillions of dollars being invested in it (and for which companies)?

The conclusion I draw from my conversations with people I consider knowledgeable is that, at this point, answering this question is largely a matter of faith.

That reminds me of a quote attributed to American economist and writer Stuart Chase:

For those who believe, no proof is necessary. For those who don’t believe, no proof is possible.

Try as I may, I can’t come up with a more appropriate resolution to the questions surrounding AI than this.

 

(JP Morgan Asset Management)

From Menlo Ventures:

More Money Goes into Using AI Than Building It

For every $1 the industry spends building AI models, enterprises spend about $1.70 putting them to work. As of September 2026, we estimate $176.8B in enterprise spending on using AI (up 5.4x) versus $102.5B on building models (up 5.3x).

That demand underwrites the buildout. We estimate that Anthropic and OpenAI alone account for 25%+ of AWS, Azure, and Google Cloud revenue, with much of that spend going to training. That investment only pays off if enterprises keep putting models to work, which agents are now driving. (…)

Open-source usage is rising and, in recent weeks, Anthropic and OpenAI have started discounting even their flagship models. The result is sharper competition at every intelligence tier, but our data show buyers continue to prioritize
capability over cost in model selection.

• Frontier models are gaining share. Frontier models more than doubled their share of tokens, from 18% to 42%, outpacing even open-source growth. Their share of spending also climbed, from 35% to 63%.
• Buyers are trading up from less expensive models. A year ago, the most popular models were Claude Sonnet 4.5 ($3 per million input tokens/$15 per million output tokens) and GPT-5 ($1.25/$10). Today, they are Claude Opus
5.5 ($4/$20) and GPT-6 Astra ($10/$60). The mix shift has kept most of the value with the most capable models, even as individual model families are discounted.

AI Spend Concentrates Where Agents Work Best

The total AI application market is now $77.3B, up from $19B last year. But just two markets account for 76% of all app spend: coding ($34.0B, up 7.7x) and enterprise chatbots ($25.0B, up 3.0x). To put that in perspective, healthcare—one of the world’s largest industries—ranks a distant third at just $3.7B.

Agents Are the Biggest New Consumers of Infrastructure

AI agents are becoming the standard way to build AI applications. Today, 74% of builders use agents to manage workflows and execute tasks, compared with 27% of startup deployments in 2025. Nearly a third (32%) are already using multiple agents working together.

This shift is reshaping infrastructure demand. Traditional software infrastructure was built primarily for applications used by people—human-paced, one session at a time, provisioned once and left running. Agents operate differently:
They work at machine speed, run in parallel, spinning up and discarding resources as needed.

Time To Sound The All Clear Signal?

(…) Our [Ed Yardeni] worry list is getting shorter:

(1) Geopolitics. The feared post-midterm re-escalation of the Middle East war did not materialize. Instead, President Donald Trump adopted a more conciliatory tone after Republican election losses, which some Republicans partly blamed on the unpopular war. Correspondingly, oil prices have retreated recently. The White House now seems determined to maintain the status quo by continuing to blockade Iran, while seeking a diplomatic solution.

(2) Bond yield. The 10-year Treasury yield is back in our old normal range of 4.00%-5.00%. We view this range as consistent with solid economic growth. It rebounded to the top of the range yesterday after October’s hotter-than-expected CPI, which increased the odds of a 25bps Fed rate hike early next month. We’ve expected a 50bps hike ever since the last FOMC meeting, which voted against another hike.

(3) Domestic policies. The midterms are done, removing one key source of uncertainty. The market’s recent rebound suggests investors are adjusting to the new political landscape, though fiscal standoffs over the federal budget and debt ceiling remain a risk. In our opinion, the Blue Wave in Congress and state governorships might slow data center construction, but it won’t stop the AI buildout.

(4) AI investment boom. Investors seem to be coming around to our view that financial, political, and physical constraints should be speed bumps, not roadblocks, to the AI buildout. That’s principally because demand for computing power remains exceptionally strong. CoreWeave’s earnings report yesterday provides further evidence, with Q3 revenue surging 168% y/y to $3.7 billion and backlog reaching $132.5 billion. Such robust demand not only justifies continued investment in computing capacity but also helps ease financing concerns. Indeed, the company expects roughly $9 billion in customer prepayments and deposits, reducing its reliance on expensive borrowing.

But is this exceptionally strong demand for computing power sustainable? The latest results from OpenAI and Anthropic suggest that it is. OpenAI’s Q3 revenue surged 160% y/y to $10 billion, while Anthropic’s soared 1,127% to almost $17 billion. This exponential revenue growth provides compelling evidence of rapidly expanding demand for AI services, which will continue to fuel compute demand. Meanwhile, reports that Anthropic’s IPO is already multiple times oversubscribed suggest that investors remain eager to finance the boom.

Fed Forced to Confront Case for Bigger and Faster Rate Rises

Federal Reserve officials spent October talking markets down from bets on a quick string of rate increases. Inflation data haven’t cooperated.

Core consumer prices, which exclude food and energy, rose at a 3.7% annualized rate over the three months through October, a broad acceleration from the 1.6% rate for the three months through July, according to a Labor Department report released Tuesday.

The figures left little doubt that the Fed will raise rates at its Dec. 8-9 meeting. The bigger questions are how quickly the central bank should move and whether it is still making the modest adjustment of two or three increases that many officials have envisioned or is instead at the start of a longer campaign to slow the economy and bring down inflation.

On Tuesday, two officials who supported the decision to hold rates steady last month suggested the Fed might now need to pick up its pace. Fed governor Christopher Waller, who spoke in London hours before the inflation data were released, said the Fed should consider a half-point increase next month if core prices rose around 0.3% in October from the previous month, which they did. And after the report, Dallas Fed President Lorie Logan hinted that a faster pace of hikes might be warranted given the current inflationary pressures hitting the economy. (…)

The strength in core prices last month showed up in categories that might be expected to reflect the pass-through of higher costs from tariffs, energy or shipping—and in categories less exposed to those pressures, such as housing, recreation and other non-transportation services, said Tiffany Wilding, an economist at Pimco.

Prices for retail goods also rose despite early-October discounting from Amazon’s Prime Day, which usually holds those figures down, she said. Surging artificial intelligence investment, meanwhile, has pushed up costs for electronic components. Prices of televisions and smartphones rose briskly last month. (…)

Officials last year were willing to look past one-off price hikes from tariffs, and in the spring, they did so for energy. But a steady run of new increases has led more officials to worry about an economy running hot enough that businesses can keep raising prices. (…)

Also in the WSJ:

Analysts described the October inflation report as concerning because price increases were broad based, ranging from a 0.4% month-over-month increase in airfares to a 0.7% increase in auto insurance. It came after data also showed larger-than-expected increases in September, creating a frustrating sense that inflation pressures aren’t going away.

“The details of today’s data are interesting largely because they are not interesting,” Eric Winograd, an economist at AllianceBernstein, wrote in a note to clients. “There was nothing obviously fluky about the data—just a generalized sense that inflation is not moving lower.” (…)

The latest inflation data provided strong evidence that investors will need to wrestle with competing threats simultaneously. Wall Street has grown increasingly wary that demand hiccups or a public backlash against data centers might trigger a slowdown in the historic boom in AI infrastructure investment. At the same time, that boom is helping to keep inflation elevated in the here and now—raising the risk that the Fed will raise rates not just once more in December but at least a few more times after that. (…)

A big contributor to inflation right now is the rise in oil prices spurred by the U.S.-Iran conflict. But those prices have already come down since mid-October and could fall further if the two sides can reach a peace deal, Arone said.

The reality is that there won’t be a “data center slowdown” until late 2028, if any, since approved projects will go ahead and that time to market is paramount.

The October CPI was bad, confirming that August and September were “real”.

CPI-Services have been rising at a steady 4.2% annualized rate for 3 months, with wage inflation (black above) picking up after being quiet between March and July. Even Kevin Warsh’s preferred measure has spiked up; trimmed-mean CPI was up 4.5% a.r. in October after +3.5% in September and +2.7% in August.

It might take more than falling oil prices, like, maybe, more productivity from that Trump-made declining labor force. Some law of economics are pretty stubborn.

• Fed Likely To Hike By 50bps In December (Ed Yardeni)

(…) Echoing the CPI report’s fire alarm, October’s NFIB survey showed that the net share of small business owners planning to raise selling prices jumped to 37%, the highest since July 2022, while 24% planned to raise worker compensation, the highest since December 2025. (…)

The FOMC focuses on PCE inflation which will be out at the end of the month. Goldman Sachs, most surprised by October’s core CPI coming in at +0.30% vs its forecast of +0.16%, acknowledges that

the upside surprise was driven by firmer-than-expected readings in a broad set of categories. Based on details in the CPI report, we estimate that the core PCE price index rose 0.35% in October (vs. our expectation of 0.23% before the CPI release), corresponding to a year-over-year rate of +3.30%.

That would come after +0.32% in September for a 2-m growth rate of 4.1% a.r., double the target that Warsh has committed to achieve.

• Trump’s Plan to Import Russian Diesel Stalls on the High Seas Tankers initially bound for U.S. change course after opposition to deal with Putin makes buyers nervous

President Trump’s efforts to supply the U.S. with Russian diesel have yet to result in any new imports into the country, as several fuel tankers bound for America shifted course following difficulties finding buyers willing to accept the controversial cargoes. (…)

Critics, including senior Democrats and Republicans in Congress, said that Trump’s carve-out of diesel from U.S. sanctions on Russia was illegal. Another concern was that accepting the diesel could cause companies to run afoul of international sanctions: fuel imports from Russia are still banned by the U.K., European Union and Switzerland. (…)

U.S. oil companies aren’t short of diesel: domestic refiners already make more than enough diesel to sate the country’s consumption and export more than 1.5 million barrels a day to overseas buyers. Despite this, the U.S. average for a gallon of diesel has climbed to $6.16 as of Tuesday, as global supplies of the fuel have tightened following the Middle East conflict. It is up from $3.75 before the war. (…)

Under its deal with Trump, Russia planned to restore its global diesel exports this year to less than one-fifth of what it was shipping before Ukraine’s drone-strike campaign, based on figures released by the president.

Those volumes represented a drop in the bucket compared with the global supply shortfall, analysts said.

“Combined diesel exports from the Mideast and Russia are still very low, with little movement there,” said Martijn Rats, chief commodities strategist at Morgan Stanley.

Why not a presidential decree capping US diesel prices… After all, these oil majors are making a windfall from the war against Iran.

• Masayoshi Son’s Huge Bet on Electrical Turbines Backfires

(…) Like an overeager miner who tries to corner the market for picks and shovels in a gold rush, Son’s SoftBank Group ordered so many turbines that even after fully supplying the world’s largest gas power plant, it was left with reservations for far more than it could possibly use.

The global tech investor switched course on Tuesday, reporting a $3.7 billion hit to its earnings due largely to canceled reservations on power generation equipment. (…)

Seeing that the rapid growth in data centers would require a historic surge of electricity production, between late 2025 and early 2026, SoftBank began placing reservations for the massive electricity generating machines, largely from GE Vernova, according to people familiar with the orders.

Turbines—the engines of gas plants that convert natural gas to electricity—require orders years in advance. (…)

The orders involved a substantial, nonrefundable reservation fee. Turbine manufacturers typically charge tens of millions of dollars per gigawatt for a reservation. (…)

The reservations strengthened SoftBank’s hand when it engaged in negotiations with the Trump administration to build a massive gas-fired power plant in Ohio tied to the larger Japan-U.S. trade agreement.

The final deal, announced in March, calls for the Japanese government to fund a $33 billion, 9.2-gigawatt gas plant on a former uranium enrichment site in southern Ohio. Construction is scheduled to start early next year, and the plant is to be owned by a joint venture of the U.S. and Japan.

SoftBank was named builder and operator of the plant, and it plans to use the electricity for a neighboring 8-gigawatt data center, which has been leased to OpenAI in what would be the world’s largest data center.

But even after supplying the plant and other smaller plants SoftBank plans, the company still had ordered too many turbines.

Its initial plan had been to use the turbines for future gas-powered plants, but in recent months, as additional payments for the turbine orders loomed, SoftBank made a change. The company opted to cancel its excess reservations, people familiar with the turbine deals said.

It also unwound contracts for some turbines that were further along in the order process, one of the people said.

A spokesperson for GE Vernova, the turbine supplier, said there was “no negative or positive impact” to the company from SoftBank’s moves. (…)

Some relief for real data center promoters.

• CoreWeave Meets Wall Street Expectations, Shares Rise

The so-called neocloud company—which buys advanced artificial-intelligence chips known as graphics processing units from Nvidia, installs them in data centers and leases them to companies such as Microsoft and OpenAI—reported revenue of $3.7 billion for the quarter that ended in September. And it said its backlog of business hit $133 billion. Analysts polled by S&P Global had expected quarterly revenue of $3.6 billion and a backlog of $131 billion. (…)

CoreWeave’s chief executive, Michael Intrator, sought to tamp down concerns about the company’s financing costs.

“Our growth is supported by multiple sources of funding, including financing backed by customer contracts, operating cash flow, and approximately $9 billion in customer prepayments and deposits expected in the next three months,” he said. “That financial flexibility positions us to keep delivering the infrastructure and software customers need to build, run, and scale AI.”

At the same time, the company said it launched a new program to sell stock to raise money. That will dilute existing shareholders and suggests the company can’t raise debt financing at attractive rates. (…)

CFO Nitin Agrawal said that CoreWeave is charging between $20 million and $25 million per megawatt for its longer-dated contracts, and up to $40 million per megawatt for short-term commitments. Shorter-duration contracts can pay back the initial investment in anywhere from six to 18 months.

• TSMC’s Revenue Rises 60% in a Sign of Resilient AI Chip Demand

The go-to chipmaker for Nvidia Corp. and Apple Inc. reported October revenue of NT$586.9 billion ($17.9 billion). That pace was up from around 51% for the September quarter. Analysts on average are projecting 63% growth for the year’s final three months. (…)

TSMC — a key barometer for the industry because it makes chips for a broad swathe of tech players — is betting on a fundamental shift. It’s trying to build and furnish about 20 factories at home and overseas, an unprecedented pace for a company that typically works on four or five new buildings simultaneously. (…)

The company has so far managed to accelerate sales growth and ramped up capacity expansion. The company reiterated its capital expenditures will reach as much as $64 billion, a record, this year. (…)

Even with its $250 billion build-out, leading-edge capacity stays effectively sold out through 2028, supporting price increases.

OpenAI Continues to Suffer Heavy Losses as Revenue Rises to $10 Billion The company improved its adjusted operating margin but remains far off from profitability

OpenAI grew revenue by roughly 50% to $10 billion from the second to the third quarter, but continued to report heavy operating losses, showing mixed results in its effort to catch up with industry leader Anthropic.

(…) The maker of Claude recently disclosed that its revenue rose to $17 billion in the third quarter while reporting a smaller operating loss of $567 million during that period.

The numbers show that OpenAI is accelerating revenue growth but still has a long way to go to match its rival’s sales and operating efficiency. The company grew revenue by only 18% in the second quarter compared with the first, The Wall Street Journal previously reported.

OpenAI told investors Tuesday that its adjusted operating margin improved from minus 150% to minus 109% from the second to the third quarter, according to documents viewed by The Wall Street Journal. Even so, the company’s adjusted operating losses ticked up slightly from $10.1 billion to $10.9 billion, based on those figures.

The calculations for the losses exclude stock-based compensation, which ticked up to $2.6 billion in the third quarter, according to a person familiar with the matter. Anthropic’s operating loss includes stock-based compensation. (…)

OpenAI also told investors that the number of paying users on ChatGPT grew to roughly 100 million, or 8.5% of total weekly active users. That percentage is slightly higher than what OpenAI reported for the previous quarter, suggesting progress in its effort to increase paying subscribers for the world’s most popular consumer chatbot. (…)

OpenAI has a large consumer business to support and maintains an array of research programs in areas from robotics to image generation. Anthropic has taken a more focused approach on building for enterprises, though its revenue is more concentrated on those customers as a result. (…)

US Loses Grip on $121 Billion Mobile Gaming Sector to China American studios are losing ground to Chinese rivals that can make games quicker, cheaper and more profitably.

(…) “It’s extraordinarily hard for startups like ours – despite having an incredible amount of capital – to find individuals to compete on the global stage, which gaming has become in the past 10 years; and one in which the US is no longer the dominant force,” Liu said.

China’s advance has been rapid. As recently as early 2024, the US and China were virtually neck-and-neck in terms of revenue share from the world’s top 100 mobile games, according to data from Sensor Tower. But by mid-2026, China held 40% of the global market, compared with just 16.5% for the US, as the Asian nation dominated the charts with hits like Honor of Kings and Whiteout Survival.

In a cruel twist, China’s biggest overseas market last year was the US itself. Chinese titles account for over 18% of revenue from the top-100 games there, up from 8% at the start of the decade. At the same time, the share held by US publishers has shrunk from 40% to less than 30%. (…)

Mobile games’ ubiquity — more than 3 billion people will play them this year — means they’re an understated but potent channel of soft power. The highly stylized fantasy and anime characters that dominate top Chinese titles now anchor the daily digital feeds of Gen Z and Gen Alpha, underscoring the broader departure from American cultural hegemony. (…)

In a nod to the structural shift at play, some top American developers like Microsoft Corp.’s Activision Blizzard and Take-Two Interactive Software Inc.’s Zynga are using Chinese studios to make their games.

It’s not just China taking share from the US. Studios in Singapore have produced Garena Free Fire. Turkish startup Dream Games has the popular Royal Match puzzle game and a swell of M&A activity has underpinned rapid growth in recent years. In the second quarter, publishers from those two countries combined accounted for 18% of revenue from the top 100 games in the US, up from 1.2% in the first quarter of 2020, according to Sensor Tower.

Meanwhile, Finland is a global gaming hub that’s produced Angry Birds, as well as some of the world’s current top titles like Supercell’s Clash of Clans and Clash Royale. Still, there’s no escaping China’s reach; Supercell has been majority owned by Tencent Holdings Ltd. since 2016. (…)

American firms looking at overseas markets face hurdles in China. A tricky regulatory environment effectively boxes them out of the market that’s expected to generate $41 billion by 2030, according to Niko Partners. In the first half of the year, the National Press and Publication Administration approved 1,110 domestic video games for Chinese consumers and just 37 from international game developers.

That lopsided access has been a longtime bugbear of Western countries, which after years of relative openness are tightening access to Chinese technology. The US has investigated Tencent over its stakes in Riot Games and Epic Games and the company has been designated as a Chinese military company — a classification it denies.

But talent is where China really has the edge.

Salaries are the biggest expense for game companies and game makers. They pay, on average, $142,000 for every developer in the US, according to a survey by the Game Developers Conference. In Shanghai, where salaries tend to be higher, the average game developer earns about $50,000 a year, according to data from Glassdoor. (…)

And now, AI is increasingly contributing.