What matters in AI this week
(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.)
While the who's who of American tech and finance gathered in Pittsburgh for a major energy and innovation summit this week, the most important figure in the industry was notably absent. Jensen Huang, CEO of Nvidia—the world’s most valuable company—wasn't listening to President Trump announce a $90 billion AI investment. Instead, he was in Beijing.
On his third visit this year, Huang confirmed that Nvidia would once again be able to sell its popular H20 chips in China as U.S. restrictions eased. He traded his signature leather jacket for a classic black Chinese Tang suit, opened his keynote in Mandarin, sat down for an interview with Chinese state TV, mingled with local tech moguls, and enthusiastically praised homegrown AI models like DeepSeek.
Huang’s move comes amid ongoing debate in Washington over whether restricting advanced chips to China could simply accelerate the rise of domestic competitors like Huawei. U.S. officials now view export restrictions as a key part of broader negotiations with China, including discussions on access to critical rare earth materials .
Chips aren’t the only flashpoint in the U.S.-China tech race. TikTok, the video-sharing app used by over a third of Americans, continues to grow despite geopolitical headwinds. We’ve learned from sources that TikTok’s Chinese parent company, ByteDance, has quietly surpassed Meta in quarterly revenue for the first time. Scroll down for what ByteDance’s meteoric rise signals for the future of social media and more.
Think your friend or colleague should know about us? Forward this newsletter to them. They can also subscribe here . Email me here or follow me on LinkedIn to share any feedback, and what you want to read about next in AI.
OUR LATEST REPORTING IN TECH & AI:
Google hires Windsurf execs in $2.4 billion deal to advance AI coding ambitions
Cognition AI to buy Windsurf, doubling down on AI coding
Musk suggests Tesla investor vote on xAI investment, rules out merger
Google inks $3 billion US hydropower deal in largest clean energy agreement of its kind
Citigroup considers issuing its own stablecoin
TIKTOK'S PARENT BYTEDANCE BIGGER THAN META
ByteDance has quietly become as the world’s largest social media company by revenue in the first quarter of this year, sources tell me and my colleague Kane Wu.
The Chinese parent of TikTok and other popular apps has raked in over $43 billion in revenue in the first three months this year—surpassing Meta’s $42.31 billion for the same period.
ByteDance has been growing faster than Meta. For all of 2024, TikTok brought in $155 billion, while Meta generated $165 billion. Both companies were able to maintain revenue growth at over 20%, driven by advertising. ByteDance didn’t respond to a request for comment.
What’s interesting is how the market values these two giants. Meta’s market cap stands at $1.79 trillion, while ByteDance, still privately held, recently marked its valuation at $315 billion in a share buyback—less than 20% of Meta’s value.
Despite this gap, Meta doesn’t appear wildly overvalued: its price-to-earnings ratio is about 27, even slightly below its historical average. Meta still enjoys higher margins and dominates lucrative markets like the U.S., while ByteDance’s growth is more global, including in its home turf China, which Meta doesn't have access to.
Some of the valuation gaps likely comes down to political risk. U.S. regulatory concerns are the main hurdle for TikTok, as President Donald Trump is orchestrating a sale on Chinese-owned app to avoid being banned in the U.S.
Both ByteDance and Meta are laser-focused on doubling down on building frontier AI. ByteDance founder Yiming Zhang is personally involved in developing AI and large language models (LLMs). In 2023, the company launched a team called Seed to focus on general intelligence and frontier AI research.
Meanwhile, Mark Zuckerberg is busy assembling the new Superintelligence team, including calling researchers himself. He announced this week that Meta will spend hundreds of billions of dollars to build massive AI data centers , trying to gather the recipe for breakthrough success in AI: world-class talent, access to vast computing power, and deep data.
“We have the capital from our business to do this,” Zuckerberg said this week, highlighting Meta’s robust ad-driven business model, which also stands to benefit as AI improves recommendations and ad targeting.
Yiming could say the same about ByteDance’s ambitions.
WHAT AI RESEARCHERS ARE READING
By Anna Tong, Tech Correspondent
Code generation has been undeniably one of the fastest-growing use cases of AI. Its impact is undeniable: developers swear by its utility, tech CEOs highlight major productivity gains, and startups in the sector are commanding high valuations . A study from AI research nonprofit METR, however, challenges the belief that AI always makes expensive human engineers much more productive.
Earlier this year, METR asked a group of seasoned developers to use Cursor, a popular AI coding assistant, to help them complete tasks in familiar, open-source projects.
Before the study, the open-source developers believed using AI would speed them up, estimating it would decrease task completion time by 24%. Even after completing the tasks with AI, the developers believed that they had decreased task times by 20%. But the study found that using AI did the opposite: it increased task completion time by 19%.
The study has drawn some criticism. Onetime OpenAI CEO Emmett Shear called the experiment “ wildly misleading ” because only one of the developers had more than one week of experience using AI coding assistants, and that developer ended up being 20% faster using Cursor, instead of slower.
The debate underscores a more complex reality: the true value of AI in coding may hinge less on the technology itself and more on the experience of the developer using it.
