(Artificial Intelligencer is published every Wednesday. Think your friend or colleague should know about us? Forward this newsletter to them. They can also subscribe here.)
Is the third time the charm for a TikTok deal in the U.S.?
As a reporter who has covered this saga since Trump’s first ban threat, this round feels closer than any before. U.S. and Chinese trade representatives reached a handshake deal in Madrid on Monday, with a formal sign ‑ off expected from Xi Jinping and Donald Trump this Friday .
It’s extraordinary that an app born from lip ‑ syncs and cat videos now sits at the center of global power bargaining. We’ve come a long way since 2020, when China tightened export controls on recommendation algorithms—rules that made it far harder to transfer ByteDance’s “secret sauce”—and when sources close to ByteDance told us the company would rather shut U.S. operations than sell .
The current framework largely mirrors the one negotiated this spring: ByteDance would retain just under 20% of TikTok U.S., with the remainder held by U.S. investors. That deal stalled in April after Trump imposed new tariffs on China and Beijing balked.
Five months later, it appears we’re closer than ever to a “yes” from Beijing, which would allow TikTok U.S. to continue to build on ByteDance’s algorithm, as I’ve detailed here . But nothing is guaranteed. The agreement is now a bargaining chip in broader U.S.–China negotiations, with the deadline extended to December 16.
What could also be on the bargaining table: U.S. AI champion Nvidia 
NVDA. In this week’s issue, we’ll dive into Nvidia’s fate to be caught up in “larger agendas” and what that means to its booming business, along with the latest data point we have on enterprise AI adoption. Scroll on.
OUR LATEST REPORTING ON TECH & AI
Meta to debut costlier smart glasses with display at annual Connect event
YouTube, new leader of US media industry, bets on AI as key for creating content
New US-China TikTok agreement similar to terms agreed this spring, sources say
Microsoft, OpenAI reach non-binding deal to allow OpenAI to restructure
Exclusive – Nvidia’s new RTX6000D chip for China finds little favour with major firms
Factbox – Hot or not? How recent high-profile US IPOs have performed
AI PAWNS
Jensen Huang and Zhang Yiming may have notes to compare about building companies that are becoming pawns in today’s digital wars: Nvidia and ByteDance.
Both are being squeezed on national security grounds in exchange for access to two lucrative markets. Beijing has effectively been steering the fate of the world’s largest social media company by revenue and, in a huge shift, is now saying it has secured a “win-win” deal that would allow a U.S.-owned TikTok to use ByteDance’s secret sauce—the AI algorithm trained on billions of interactions that uncannily predicts our interest.
Similarly, Nvidia is navigating a delicate dance between two superpowers. Just a month ago, everything looked promising. Jensen Huang was feted on a tour in China while notching some wins in D.C., as Trump engineered an unusual deal in which the U.S. would grant Nvidia licenses to sell H20 chips to China in exchange for a 15% cut of those sales. Investors cheered; Chinese companies couldn’t wait to regain access to top AI hardware.
But the sentiment quickly soured. Beijing warned major tech firms to hold off buying from Nvidia and to prioritize homegrown chips. Then came a preliminary antimonopoly probe accusing Nvidia of violating China’s competition law—a signature move by Beijing to send signals abroad.
“We can only be in service of a market if a country wants us to be,” Huang said on Wednesday in London. “I’m disappointed with what I see, but they have larger agendas to work out between China and the United States, and I’m patient about it.”
We know TikTok will come up on Friday’s Trump–Xi call agenda, with a final deal expected. We’re less sure Nvidia will secure White House backing to push for market access in China. Washington has been divided from the start about cutting any deal that bends export control rules.
It’s worth remembering that opening the Chinese market to U.S. firms was one of Trump’s goals since Trade War 1.0. Eight years on, the market shows few signs of being more open—especially for top AI firms. Meanwhile, the U.S. is borrowing from Beijing’s state-capitalist playbook: using rules and legislation to decide who can sell tech in China and what U.S. tech can be sold abroad.
CHART OF THE WEEK
This chart draws on the U.S. Census Bureau’s biweekly survey of 1.2 million firms to track real-world AI use. It asks whether, in the past two weeks, a business used AI tools—machine learning, natural language processing, virtual agents, or voice recognition—to produce goods or services. Since June, the rate of adoption has declined from 14% to 12% for companies with over 250 employees, although the survey also showed a slight increase in AI use among smaller companies, suggesting momentum at the top end has cooled even as the broader trend remains up.
That wobble matters. Large enterprises have the budgets to turn pilots into projects, so a pullback could mean choppier near-term spending on AI software and services. Are we entering a digestion phase, where tech debt and integration slow down AI’s usefulness? Ultimately, the massive infrastructure buildout will be judged on delivered productivity, not demos.
AI RESEARCH TO READ
What do 700 million people use ChatGPT for every week? A new OpenAI study of real user behavior confirms some assumptions and upends others. Non-work use now dominates—over 70% of consumer messages, up from about 53% a year ago—driven largely by existing users shifting how they engage.
Consumers use ChatGPT more as a search-plus adviser than as a pure content engine. “Practical Guidance” and “Seeking Information” together rival “Writing,” while “Asking” for advice accounts for about 49% of all messages, outpacing task execution. Meanwhile, the buzzy “AI friend” narrative is modest in the data: only 1.9% of messages are about relationships or personal reflection and 0.4% are games or roleplay.
For work, writing is the standout use case at roughly 40% of work messages, while coding is small at about 4.2% of all messages as more users turn to specialized coding tools . And most “writing” isn’t from scratch: roughly two thirds involve editing, critiquing, or translating.
Demographics have broadened fast—the early male skew flipped to parity and a slight female majority by mid-2025. Nearly half of adult messages come from 18–25-year-olds, and adoption is accelerating across low and middle-income countries.
