On September 16, 2025, the U.S. once again extended TikTok’s “sell or ban” deadline to December 16—this is the fourth extension in five years. With 170 million U.S. users and a platform that generates $5.8 billion in GMV annually (per TikTok Shop’s 2025.09 official report), TikTok remains in app stores. ByteDance has managed to hold onto the recommendation algorithm that was listed by the U.S. Commerce Department in China’s “Catalogue of Technologies Prohibited or Restricted for Export.” The story is still being written.
I. From “45 Days” to Four Extensions: How an Executive Order Was Blunted by Reality
On August 6, 2020, then‑President Trump declared aboard Air Force One, “TikTok must be sold to an American company, or it will be shut down in 45 days.” That executive order cut right to ByteDance’s most sensitive territory—the North American market, which at that time had contributed 200 million downloads and $1.2 billion in annual ad revenue—one quarter of its global ad business. Behind the edict lay a White House narrative of data risk to Chinese government access, along with lobbying from Facebook, which had lost ground in the same social media space.
But 45 days passed and TikTok did not change hands. Instead, it triggered a wave of users posting “Thank You, Trump” videos, demonstrating the organizing power of digital natives. The Washington team saw firsthand that banning TikTok would provoke backlash. Since then, the narrative has looped: each deadline approaches, a new order, a new extension, a new “baseline consensus.” When Biden came into office in 2021, he rescinded his predecessor’s order—but in 2022 inserted a clause banning TikTok on federal devices into the omnibus spending bill. In 2024, Congress passed the Protecting Americans from Foreign Adversary Controlled Applications Act, enshrining a 165-day divestiture requirement into statute, leading many to believe a “final judgment” was coming. Yet on January 19, 2025, the Supreme Court of nine justices upheld the law’s constitutionality citing “national security priority.” That night, TikTok blacked out for 12 hours. Then, Trump intervened with “Don’t issue the fines yet,” and TikTok came back online—miraculously. That “death act” left U.S. small merchants panicked, and turned what seemed like a rigid legal tool into a protracted tug-of-war.
By the time of the Madrid talks in 2025, Washington and Beijing packaged a “framework consensus” into a joint communiqué, and Trump signed another extension. Over five years, the edge of that executive ban has been blunted by employment numbers, GDP contributions, youth voters, and Christmas shopping season risks. In retrospect, that blade initially portrayed as decisive was never fully dropped. It has been reforged, re-sharpened in each countdown, to become a highly elastic rope—allowing Washington to posture tough on China while preventing Silicon Valley–Yiwu flows of goods and capital from snapping instantly.
II. The Evolution of the “Texas Plan”: How Data Localization Became an Algorithm Sovereignty Guardrail
In March 2022, TikTok CEO Shou Zi Chew publicly disclosed details of the “Texas Plan” at a Congressional hearing: U.S. user data would be stored only on Oracle servers in Texas; Oracle engineers could review the source code; a “Transparency Center” would allow third parties to observe algorithmic processes. At that moment, many saw it as a superficial data-localization measure—but one clause raised eyebrows: “review scope includes recommendation logic.” In other words, the U.S. side sought not only local storage but confirmation that the “black box formula” determining what Americans see couldn’t be remotely manipulated. Over the next three years the Texas Plan upgraded repeatedly. In 2023, independent auditor Deloitte was introduced to produce an annual “Algorithm Impact Report.” In 2024, a “U.S. Data Security Company” (USDS) was established, jointly held by Oracle, Walmart, and U.S. venture firms, with ByteDance retaining single-control equity. By 2025’s Madrid framework, USDS became the sole “custodian + authorizer.” ByteDance would license model weights to USDS, while maintaining the rights to update and iterate in Beijing.
This layered “entrusted operation” architecture mirrors Apple’s iCloud “Cloud in Guizhou” logic—putting the most sensitive part (data, audit interfaces) in a local entity, while locking away the valuable core engine (algorithm, model) in the parent country’s servers. The difference: in Apple’s case China demanded the encryption keys be kept in Guizhou; here, the U.S. demands TikTok hand over the “steering wheel” to Oracle, but allows the engine to remain in Beijing. Through “equity + contracts + technical” triple isolation, each side obtains the puzzle piece it wants: Washington gains real-time review of recommendation outcomes; Beijing retains model IP and iteration control. As one Silicon Valley lawyer involved in negotiation said: “This is not a sale, not a joint venture—it’s a ‘technological concession zone’—you remain the landlord, but the tenant can pitch tents in your yard.”
III. The $50 B Ecosystem’s “Heartbeat Freeze” Cost: Why No One Dares Let TikTok Actually Go Dark
That 12-hour “false death” on January 19, 2025 still haunts the U.S. e-commerce community. North American small-store GMV went to zero instantly. Advertisers rushed budgets to Meta, driving Facebook’s hourly CPM up 18%. Simultaneously, Xiaohongshu’s U.S. downloads soared tenfold, rocketing to first place on the App Store, but its servers crashed under cross-border load.
The Oxford Economics Institute later calculated: if TikTok were permanently removed, U.S. GDP would shrink by $28.4 billion over 12 months—roughly the size of Montana’s entire economy. Direct employment loss: 22,000 jobs. Indirect: 224,000 jobs. Federal and local tax revenue would decline by $7.4 billion. More insidious was the instant blockage of access for 5.5 million small merchants. On average, 42% of their repurchases come from TikTok video-driven discovery. Migrating to Meta would cost them 27% more per exposure, and ROI would drop 15%. For live streaming merchants relying on $10 average order size, that wipes out profit entirely. Beyond numbers, there is politics. In 2024’s election, 63% of voters aged 18–29 used TikTok for election information. Trump himself joined the platform in June—the most-watched video exceeded 150 million views. A ban would drive those young voters into the opposition’s arms. The political, economic, and social costs layered together make “extension” Washington’s rational middle ground—they can posture tough and yet avoid triggering a perfect storm of unemployment and inflation before Christmas.
IV. “Selling the Shell, Not the Core”: How Tech Export Controls Became China’s Hardest Trump Card
In August 2020, when Trump first demanded a sale, ByteDance considered selling both equity and code. But on August 28, China’s Ministry of Commerce and Ministry of Science & Technology revised the Catalogue of Prohibited or Restricted Export Technologies—adding “personalized content push based on data analysis” to the restricted list. It meant algorithms needed an “export permit.” Since then, ByteDance has used that regulation as an “immunity golden card” in every negotiation.
In 2023 the catalogue was expanded again to include “AI interactive interfaces” and “large-scale data mining & analysis technologies,” adding locks around model weights, training frameworks, and parameter optimization tools. That meant that even if the U.S. got the source code, it could not obtain the latest iteration weights developed in China. Even if reverse engineering re-creates the model, it would lack the domestic data scale to match performance. Crucially, approvals for technical exports rest firmly with the Ministry of Commerce—any “tech exit” must undergo provincial review, Science & Technology re-examination, and interagency approval, up to 45 working days, and can be vetoed at any time. Thanks to this firewall, the Madrid framework stipulated “licensed use” as an alternative—U.S. receives API access but every model upgrade must happen in China and then be mirrored to Texas servers. As a Tsinghua professor who helped revise the catalogue said: “An algorithm is not a mere commodity—it’s infrastructure of the digital era; who controls iteration rhythm holds next-generation discourse power.” By elevating “technical sovereignty” to a state tool, Beijing made Washington realize that forced divestment faces not only legal barriers but an administrative Great Wall.
V. From TikTok to “TikTok-ization”: The Enterprise Survival Playbook in the Digital Balkan Era
TikTok’s five-year survival saga is a geopolitical survival manual for Chinese tech companies aiming to go global. The first page reads: comply first, expand later. Since 2024, ByteDance has replicated this model in the Middle East, Southeast Asia, and Latin America: in Saudi Arabia, it formed a JV with sovereign fund PIF and localized data in Riyadh; in Indonesia, it brought in GoTo as a strategic shareholder following rules that social commerce can’t directly do payments; in Brazil, it mirrored the Texas model, storing local consumer data in a Brazil telecom-owned cloud and opening algorithm audit APIs to regulators.
Next, capital architecture must be “aquatic”: invite local cornerstone investors in consumer markets, create golden shares in the VIE layer to ensure critical decisions remain under Chinese control, and adopt modular licensing in B2B tech output—splitting recommendation engines into “user profiling,” “content understanding,” “cold start” subsystems and selectively exporting based on regulatory risk. Finally, brand narratives must be locally embedded: in Mexico, TikTok turned “help corn farmers live-stream sell their grain” into a public welfare project; in Thailand, it partnered with the tourism bureau on “One Village One Product,” letting algorithm amplify rural crafts. Through “data localization + equity decentralization + narrative localization,” ByteDance aims to upgrade TikTok from a “Chinese company” into “global infrastructure,” resilient in the age of digital Balkanization. As Foreign Affairs put it: “The Internet is shifting from a single net to many nets. A company’s only survival path is to become the router linking different net segments.”
VI. After December 16, the Story Is Just Beginning
Over five years: three presidential orders, four extensions, countless rounds of negotiation. TikTok still stands on a cliff’s edge, yet the world has glimpsed the underlayer of digital-era power: data is oil, algorithms are the grid, and whoever masters iteration rhythm controls the next-generation growth engine. Will December 16 be the end? Nobody dares say. What’s certain is that even if TikTok stays in North America via “entrusted operation,” that’s an annotation, not a period. The next stage will see EU’s Digital Services Act algorithm audits, India’s new “data mirror” rules, Saudi’s AI sovereignty fund ambitions—each queued to replay the TikTok script. For every Chinese enterprise planning to go abroad, TikTok’s five-year run proved this: technology may be borderless, but companies hold passports; algorithms can be neutral, but code always has a stance. True globalization no longer means “one ship traversing all seas,” but “building a local ship for each sea.” When the December 16 bell rings, the story won’t end—only the spotlight moves to the next actors: Shein, MiniMax, or any team tweaking models in a late-night office. History tells us: when technology and politics vibrate on the same chord, every swipe of a finger writes a footnote in the future map of global commerce.
[Disclaimer]: The above content reflects analysis of publicly available information, expert insights, and BCC research. It does not constitute investment advice. BCC is not responsible for any losses resulting from reliance on the views expressed herein. Investors should exercise caution.