What's up: Oracle co-CEOs take the stage; Goldman Sachs plans layoffs, more AI; Self-driving taxis are coming to London.

Good morning. I rewrote my new column on the economics of AI more times than I can recall, because the story keeps evolving by the day. It’s hard enough to comprehend the sheer scale of the investments going into AI, let alone grasp the contours of the technology itself. The pace of change, on the financial and technological fronts, makes the whole thing a rapidly moving target. (The only aspect of AI that isn’t moving quickly? Regulation and policy at the federal level, but that’s another story for another day.) This is where I came down on AI economics. For now.

AI economics are brutal. Demand is the variable to watch. It’s hard to predict how the artificial-intelligence bubble will play out, but here’s a clue. Keep an eye on usage of AI, measured in units known as ‘tokens.’ It’s soaring.

Usage, which everybody expected to increase as the unit price of AI computing fell, went absolutely parabolic.

“This time last year, we were processing 9.7 trillion tokens a month across our products and APIs [application programming interfaces],” Alphabet’s Google said in a May blog post, referring to units of AI usage. “Now, we’re processing over 480 trillion—50 times more.”

Google was just getting warmed up. “Since then we have doubled that number, now processing over 980 trillion monthly tokens, a remarkable increase,” the tech giant said in July during its second-quarter earnings call.

This month, Google said the figure had reached 1.3 quadrillion. (Per Nvidia, a token is a unit of data processed by AI.)

The latest models generate far more tokens to deliver better answers. That is generating losses given that startups and VCs are pricing AI to build a market, much as Uber once subsidized rides, Heath Terry, global sector lead for technology and communications research at Citi told me.

Better models, more data center capacity and the ability of AI to help improve itself should turn the economics around, Terry argues. At that point, soaring demand would generate a profit.

Even so, returns may be highly concentrated. “In venture capital, 6% of investments result in 60% of returns,” venture capitalist Vinod Khosla told me. “In AI, I think it will be half that percentage resulting in more than 60% of returns.”

One thing is clear already. There are no half-measures here. Going “halfsies” on model development or infrastructure build-out won’t be enough to turn around the brutal economics of this tech. It’s an all-or-nothing bet on a trillion-dollar scale, and investors are all in.

Oracle co-CEOs defend massive data-center expansion, plan to offer AI ecosystem

Freshly appointed dual chief executives Clay Magouyrk and Mike Sicilia are preparing to make their case to investors this week for how Oracle’s data-center expansion will provide a clear path to profitability.

“We’re really in a unique situation to deliver what we call applied AI,” Sicilia told the WSJ Leadership Institute's Belle Lin in a recent interview. That includes infrastructure, analytics and applications, he said 

Magouyrk told the WSJLI that Oracle has a huge opportunity to help clients “do their inferencing right alongside their data with the best models.” The company’s new AI Data Platform will allow for that, he said, predicting that customers’ AI usage will increase as much as one thousand times once they start using the platform.

Investors and tech analysts in the last month have expressed concerns about the extent to which the new build-out depends on OpenAI, which won’t generate a profit until 2029, according to the ChatGPT maker’s chief executive, Sam Altman.

Earlier this month, Oracle shares fell as much as 7.1%—though they later rebounded—following a report suggesting that its margins on renting specialized Nvidia chips were razor thin.

Oracle, AMD partner on new AI chip deal. 

Starting in the third quarter of next year, AMD plans to deploy 50,000 of its MI450 GPUs inside data centers run by Oracle. The two companies say they plan to expand their partnership in 2027 and beyond. The companies didn’t disclose the financial terms of the deal.

Both companies have been on dealmaking sprees. AMD recently struck a five-year partnership with OpenAI expected to produce tens of billions of dollars in revenue. Oracle signed a huge deal, also with OpenAI, under which the startup will buy $300 billion in cloud services over roughly five years, The Wall Street Journal reported last month.

More AI deals, partnerships and investments. 

Walmart enters the chat. Weeks after launching Instant Checkout, a service which lets shoppers buy products directly within ChatGPT, OpenAI appears to have landed the monster of retail partners: Walmart, WSJ reports.

Salesforce announced new expanded partnerships with OpenAI and Anthropic, bringing their AI models to its Agentforce 360 software, CNBC reports.

Workday plans to invest $202.5 million to set up an AI-focused research center in Dublin, WSJ reports. The investment aims at pioneer what Graham Abell, Workday’s vice president for software engineering, described as "the next generation of ERP, built for the AI era,” and work to ensure Workday’s use of AI complies with the European Union’s AI Act,

Also this....

In a post on X Tuesday, OpenAI CEO Sam Altman said that ChatGPT will allow "erotica for verified adults" when it rolls out age-gating in December.

Goldman Sachs plans layoffs, more AI

The bank on Tuesday also reported that is was on pace for its best year ever in its main investment banking and markets division

Self-driving taxis are coming to London. Cue freakout.

London's famed black cabs are getting scarcer. There were just 14,069 of the boxy people carriers earlier this month, according to licensing data, down from roughly 22,000 a decade ago. Vehicles regulated for “private hire,” including for services like Uber,  are gaining popularity and now comes a new challenge: Robotaxis.

Waymo said Wednesday it would launch a fully autonomous ride-hailing service in London next year, marking the Alphabet-owned company’s European debut. Uber is also planning to roll out a London robotaxi service after striking a partnership with local autonomous-driving startup Wayve.

Don't count cabbies out yet. London’s mix of Medieval alleyways, Victorian terraces and modern roundabouts are set to pose a new challenge, the WSJ reports. Also not to be taken for granted: The cabbies themselves. Drivers have to memorize London’s labyrinth of streets and pass a test known as “the Knowledge.” 

“I’ve got no confidence in the software being as smart as it needs to be to work in a city like London,” Steve McNamara, general secretary of the Licensed Taxi Drivers Association, tells the Journal. 

Criminal organizations operating out of China are working with gig workers recruited locally in the U.S. to flood phones with text messages seeking payment for unpaid tolls or postage fees. The scheme has become a billion-dollar business, according to the Department of Homeland Security. 

Dutch chip-making tool company ASML reported orders of 5.40 billion euros ($6.27 billion) for the quarter, up from 2.63 billion euros a year earlier and above analysts’ forecasts, the Journal reports. ASML said it expected sales next year shouldn’t be below the 2025 figure.

Meta's Instagram is rolling out new protections for teenagers’ accounts and introducing a system that limits what young users can see and interact with, including AI bots, based on PG-13 movie ratings.

👉 Meta’s ‘Digital Companions’ Will Talk Sex With Users—Even Children

The rapidly changing tariff landscape is prompting a flood of AI tools to help importers, exporters and freight companies. WSJ Logistics reports that San Francisco-based freight forwarder Flexport is the latest logistics-technology provider to enter the chat with its own AI tools.  “It’s become very complicated to calculate tariffs,” said Flexport founder and Chief Executive Ryan Petersen.

Inflation in the grocery aisle is picking up, and stinging consumers. Consumers said they are cutting back on purchases, stockpiling certain foods or exploring more-affordable stores. (WSJ)

In its trade standoff with Washington, Beijing thinks it has found America’s Achilles’ heel: President Trump’s fixation on the stock market. (WSJ)

D’Angelo, the soul savant whose sensual hit “Untitled (How Does It Feel)” was both his greatest triumph and his biggest frustration, has died. He was 51 years old. (WSJ)

The WSJ CIO Journal Team is Steven Rosenbush, Isabelle Bousquette and Belle Lin.

The editor, Tom Loftus, can be reached at thomas.loftus@wsj.com.