Alibaba Shares Slide after $10.2 Billion AI Share Sale Offered at Sharp Discount

FILE PHOTO: An Alibaba logo is displayed at the company's booth at China International Fair for Trade in Services (CIFTIS) in Beijing, China, September 10, 2025. REUTERS/Maxim Shemetov/File Photo
FILE PHOTO: An Alibaba logo is displayed at the company's booth at China International Fair for Trade in Services (CIFTIS) in Beijing, China, September 10, 2025. REUTERS/Maxim Shemetov/File Photo
TT

Alibaba Shares Slide after $10.2 Billion AI Share Sale Offered at Sharp Discount

FILE PHOTO: An Alibaba logo is displayed at the company's booth at China International Fair for Trade in Services (CIFTIS) in Beijing, China, September 10, 2025. REUTERS/Maxim Shemetov/File Photo
FILE PHOTO: An Alibaba logo is displayed at the company's booth at China International Fair for Trade in Services (CIFTIS) in Beijing, China, September 10, 2025. REUTERS/Maxim Shemetov/File Photo

China's Alibaba shares slumped in Hong Kong trade on Monday after it launched a $10.2 billion share sale at a steep discount to fund its AI ambitions, with investors focused on stock dilution and execution risks.

The e-commerce and cloud computing giant said it would sell HK$80 billion ($10.2 billion) of new shares at HK$112.70 each, an 8.4% discount to Friday's close, to fund chips, AI infrastructure and models.

AI has become Alibaba's biggest driver of revenue growth at a time when e-commerce growth is stagnating, and its Qwen AI models are some of the most popular in China. Even so, some investors have reservations about how successful it will be.

"Alibaba's DNA is in e-commerce, not advanced tech," said Yang Tingwu, vice general manager of asset manager Tongheng Investment.

"No matter how much it invests in AI hardware, it will likely be outmaneuvered by competitors in tech innovation."

Its Hong Kong shares fell as much as 10.5% but pared losses in the afternoon to trade in line with the discount offered.

The sale of 710 million ordinary shares is equivalent to 3.6% of enlarged total shares outstanding.

It drew strong demand, attracting $28 billion of orders, including $6 billion from long-only and sovereign investors, three people with knowledge of ⁠the matter said.

About ⁠40% of the book will go to long-only and sovereign investors, including major sovereign wealth funds in Europe, Asia and the Middle East, two of the people said.

Investors included the Qatar Investment Authority (QIA), Norway's Norges wealth fund and Hillhouse, according to one person.

Alibaba, Hillhouse, QIA and Norges did not immediately respond to Reuters requests for comment.

Alibaba chairman Joe Tsai bought 720,000 Hong Kong shares at an average price of HK$112 apiece, for about HK$80 million in aggregate, while Eddie Wu, the group's chief executive, bought 350,000 Hong Kong shares at an average price of HK$111.6 per share, totaling HK$40 million, according to the group's stock exchange disclosures later on Monday.

As the US and China vie for tech supremacy, investment in AI and related infrastructure such as data centers ⁠has reached dizzying heights.

The biggest Chinese AI names are, however, investing only a fraction of what their US counterparts are spending. Most fundraising globally is also conducted via heavy debt issuance — a trend that has begun to test the limits of investor demand. Japan's SoftBank on Monday announced it would issue $6.3 billion in bonds to retail investors — its biggest debt offering to date.

Alibaba's stock sale is the largest-ever follow-on offering of new shares by a Hong Kong-listed company and the third-largest globally this year after offerings of nearly $85 billion from Alphabet and $20 billion from Intel.

"Alibaba's placement — landing alongside massive capital raises by Alphabet and Intel in the US — proves that American and Chinese tech giants are operating off the exact same strategic playbook," said Winston Ma, an adjunct professor at NYU School of Law and former head of North America for sovereign wealth fund China Investment Corp.

"The global sovereign investors aren't blind to US-China tech friction — they are compartmentalizing it," Ma said, adding that they were more comfortable with compliance issues when investing in Chinese commercial cloud and open-weight AI plays over restricted semiconductor hardware.

Capital Group, one of the world's largest active investment managers, estimates that AI-related capital expenditure by the biggest US hyperscalers — Microsoft, Amazon, Alphabet, ⁠Meta and Oracle — reached $791 billion as of ⁠July 31. That compares with $118 billion for China's ByteDance, Alibaba, Tencent and Baidu.

Part of the reason for the more subdued Chinese spending has been a lack of access to Nvidia's most advanced AI chips due to US export controls. That in turn has pushed Chinese firms to develop more efficient AI models and infrastructure that require less computing power and capital.

The share placement comes a week after Alibaba reported quarterly net profit that tumbled 75% from a year earlier, primarily due to AI-related spending.

Underscoring how AI has leapt to become a key priority, Alibaba this year separated its AI operations from its cloud business, with the new unit to be led by CEO Eddie Wu.

In addition to positioning itself as a key AI partner for companies operating in China, it is preparing a listing of its chipmaking arm T-Head and developing AI agents linking services across its sprawling ecosystem, including shopping, food delivery, travel and entertainment.

Separately, Alibaba has helped train a large language model that Apple will sell in the Chinese market, sources have said.

At earnings, Alibaba said it had committed nearly half of its three-year capital expenditure plan of 380 billion yuan ($56.5 billion), but that AI computing investments have a "high certainty" of returns.

Wu said such investments are expected to break even within three years, possibly even 2.5 years, as margins improve and proprietary chips replace third-party hardware.



Taiwan Indicts Nine Over Alleged Illegal Export of AI Servers to China

People walk past a Taiwanese flag in New Taipei City on January 13, 2024. (AFP)
People walk past a Taiwanese flag in New Taipei City on January 13, 2024. (AFP)
TT

Taiwan Indicts Nine Over Alleged Illegal Export of AI Servers to China

People walk past a Taiwanese flag in New Taipei City on January 13, 2024. (AFP)
People walk past a Taiwanese flag in New Taipei City on January 13, 2024. (AFP)

Taiwan prosecutors said on Monday they indicted nine people, including employees of Nvidia and Super Micro, accused of illegal export of artificial intelligence servers to China.

Semiconductor powerhouse Taiwan is the world's largest producer of advanced chips used in AI applications. Prosecutors this year investigated the suspected ‌illegal export of ‌servers equipped with Nvidia ‌chips ⁠subject to US export ⁠controls.

Washington has imposed curbs since 2022 making it illegal for such semiconductors to be exported or sold in China.

In a statement, the prosecutors in the northern ⁠port city of Keelung said the ‌defendants, whose ‌full names they did not state, were "fully ‌aware" that both Nvidia and ‌Super Micro have "rigorous internal control procedures" regarding exports.

However, the defendants "colluded with one another at various levels for enormous profit, ‌illegally exporting high-end servers, increasing corporate compliance costs, and severely damaging our ⁠nation's ⁠international image", they added.

Neither Nvidia nor Super Micro immediately responded to requests for comment.

Taiwan has tightened export controls in recent years to keep advanced technology and know-how from reaching China, which claims the democratically governed island as its own territory despite the strong objections of the island's government.


TikTok Reaches $400 Million Settlement with US Justice Department over Children's Privacy

FILED - 24 August 2022, North Rhine-Westphalia, Cologne: The logo of Tik Tok is seen at Gamescom. Photo: Rolf Vennenbernd/dpa
FILED - 24 August 2022, North Rhine-Westphalia, Cologne: The logo of Tik Tok is seen at Gamescom. Photo: Rolf Vennenbernd/dpa
TT

TikTok Reaches $400 Million Settlement with US Justice Department over Children's Privacy

FILED - 24 August 2022, North Rhine-Westphalia, Cologne: The logo of Tik Tok is seen at Gamescom. Photo: Rolf Vennenbernd/dpa
FILED - 24 August 2022, North Rhine-Westphalia, Cologne: The logo of Tik Tok is seen at Gamescom. Photo: Rolf Vennenbernd/dpa

TikTok has reached a $400 million settlement with the US Department of Justice, ending a 2024 lawsuit alleging the company violated federal children's privacy laws.

The DOJ said Friday that TikTok will pay $300 million immediately and another $100 million after an order vacates an earlier consent decree against its predecessor company, Musical.ly.

“This settlement is a major victory for American children and parents,” said US Associate Attorney General Stanley E. Woodward Jr. in a statement. “The Department’s priority is ensuring that children are protected online and that companies entrusted with their personal information meet their legal obligations. This resolution secures a substantial recovery while reinforcing the protections that families expect and deserve.”

Since the DOJ's lawsuit in 2024, TikTok has undergone major changes, most notably in the ownership structure of its US arm. In January, the social video platform company signed agreements with major investors including Oracle, Silver Lake and MGX to form the new TikTok US joint venture.

Representatives for TikTok did not immediately respond to a message for comment Friday.

The latest lawsuit focused on allegations that TikTok and its China-based parent company ByteDance violated a federal law that requires kid-oriented apps and websites to get parental consent before collecting personal information of children under 13. It also says the companies failed to honor requests from parents who wanted their children’s accounts deleted, and chose not to delete accounts even when the firms knew they belonged to kids under 13.

The settlement comes as social media companies face an avalanche of lawsuits over children's safety and privacy and a growing number of countries are banning young kids and teens from social media apps. Instagram's parent company, Meta Platforms, is currently on trial in federal court in Oakland, California, over allegations it violated the 1998 Children’s Online Privacy Protection Act, or COPPA, along with various state statutes.


How AI Is Changing Event Discovery, Booking

Nadeem Bakhsh, chief executive and co-founder of webook.com (Company handout)
Nadeem Bakhsh, chief executive and co-founder of webook.com (Company handout)
TT

How AI Is Changing Event Discovery, Booking

Nadeem Bakhsh, chief executive and co-founder of webook.com (Company handout)
Nadeem Bakhsh, chief executive and co-founder of webook.com (Company handout)

A night out can start with a simple question: “What should we do tonight?”

That is where traditional search can struggle. It works best when users already know the event, date, or location they want. Booking platforms are now using artificial intelligence to tackle the opposite problem: first understanding vague intent, then turning it into options that can actually be booked.

Nadeem Bakhsh, chief executive and co-founder of webook.com, says the platform’s AI booking agent was not built because existing search tools were failing. It was designed to add another layer of discovery for users who do not yet know exactly what they want.

Speaking to Asharq Al-Awsat, Bakhsh said search and filters remain effective when users have a clear target. Open-ended questions, however, require the system to understand context before it can surface the right option.

From keywords to intent

In its first version, the agent uses natural language processing in Arabic and English and can handle informal requests that include location, preferences, group type or booking criteria.

Users do not need to know an event’s name or phrase their request like a search query, Bakhsh said. They can simply ask for a family activity, a nearby experience or something available in the evening.

The system then interprets the request, identifies the relevant details and matches them with options in the platform’s live catalog.

If key information is missing, the agent is designed to ask for clarification rather than produce a generic list that may not fit the user’s needs.

That makes accuracy central to the experience. The more discovery shifts from explicit search to conversation, the greater the risk that a question will be misunderstood or misinterpreted.

When the agent gets it wrong

Bakhsh acknowledges that mistakes are part of using AI systems.

The agent, he said, “should not pretend to know.” If a request is unclear or incomplete, it should ask for more information rather than force an unreliable recommendation.

The official booking page remains the final authority on availability, prices, fees and event terms. If information changes or tickets sell out before a purchase is completed, the details shown in the official booking flow take precedence over anything said earlier in the conversation.

When the agent cannot verify information, Bakhsh said, it should direct the user to official event details or human support.

A correction from the user may also help improve the next response, but Bakhsh does not present that as a guarantee that the same mistake will not happen again.

Recommendation or advertising?

Trust becomes a commercial issue when the agent is not only finding events, but ranking and recommending them.

Bakhsh said the system is meant to match user intent with options that are genuinely available to book, not present paid advertising as personal advice.

Recommendations are based on criteria supplied by the user, including location, timing, group type and activity preference, as well as live availability.

Any paid or sponsored placement, he said, should be clearly labeled so users can distinguish an organic recommendation from commercial content.

“Trust depends on making that distinction clear,” Bakhsh said.

One journey, several channels

The experience becomes more complicated when the conversation starts outside the booking platform itself, on services such as X, Instagram or WhatsApp.

Bakhsh said the agent uses information users choose to provide, such as location, timing, group type, interests and booking criteria, to make suggestions more relevant.

Public interactions remain limited to discovery. Once greater privacy is required, the conversation moves to a private channel and then into webook.com’s secure booking environment.

Card details and other sensitive personal information should not be requested through public posts, comments or direct messages on social media, he said.

Personal data remains subject to the platform’s privacy policy and to legal and regulatory retention requirements.

Bakhsh also said data is not treated as freely transferable from one channel to another. Any personalization depends on the permissions granted by the user and the purpose for which the data was collected.

What webook.com builds — and what it does not

The agent is not built entirely with in-house technology.

Bakhsh said webook.com develops and operates the core product layer and orchestration system. That includes understanding booking intent, linking it to the catalog, applying product logic, generating recommendations and directing users into the secure booking flow.

External infrastructure is used when needed for language-model processing, cloud services and access to channels such as X, Instagram and WhatsApp.

Those providers supply infrastructure or distribution channels, Bakhsh said, but they do not own the event inventory, booking logic or customer journey.

The company does not disclose vendor architecture or security-sensitive implementation details.

Bakhsh also stressed that the agent remains in beta and can make mistakes. Users should verify event details and final booking information on the official page before completing a transaction.

Discovery ends where the transaction begins

One of the main safeguards in the system is the separation between conversation and payment.

The agent may respond to a general request in a public space, but when privacy becomes necessary, it moves the interaction into a private channel and then directs the user to an official, identity-linked booking flow.

It does not request card details or sensitive information through social media posts, comments, or direct messages.

Bakhsh said this structure helps limit the impact of prompt manipulation, impersonation, fraudulent accounts, and fake booking links.

The conversational layer can help users find an event, but it cannot bypass the inventory, account, pricing, or payment controls on the official platform.

The agent does not set the price

Once a user moves from discovery to purchase, the agent’s role stops at a clear boundary.

Options come from the live catalog, but the official booking page remains the authoritative transaction interface.

Before confirming, users see current availability, the final price, booking fees and event-specific terms, including cancellation and refund policies.

The agent cannot change those terms or guarantee that a ticket will still be available while a user waits.

Age restrictions and accessibility requirements are shown only when they appear in the official event details. If the information is unclear, the system is not supposed to guess.

AI, in other words, changes how users reach an option. It does not change the contractual terms or controls governing the purchase.

Bakhsh also links the agent to a broader framework called TruFan, while drawing a distinction between the two.

The AI agent operates at the start of the journey, turning natural-language requests into bookable options.

TruFan addresses a different stage: integrity and access to high-demand events, by helping identify genuine, verified fans and giving them priority.