After Hours Summary: Major indices turn significantly lower following Trump tariff announcement; RH -23.7% sharply lower on earnings, W -14.3% down in sympathy
After Hours Gainers:
Companies trading higher in after hours in reaction to earnings/guidance: PENG +5.2%
Companies trading higher in after hours in reaction to news: SLND +6.2% (awarded two water resource projects), NGNE +5.2% (announces publication for NGN-401), FLR +1.2% (to provide EPCM services for a second multi-billion-dollar pharma facility in Indiana), NPCE +0.7% (announces refocusing of product portfolio; reaffirms guidance)
After Hours Losers:
Companies trading lower in after hours in reaction to earnings/guidance: RH -23.7%, FC -18.5%, LSTR -6.2%
Companies trading lower in after hours in reaction to news: W -14.3% (in sympathy with RH earnings and downbeat guidance), WSM -8.8% (in sympathy with RH earnings and downbeat guidance), PCAR -5.8% (CFO to retire), AAPL -5.8% (mega cap tech names lower on tariff news), AMZN -5.1% (mega cap tech names lower on tariff news), LZB -5% (in sympathy with RH earnings and downbeat guidance), TSLA -4.7% (mega cap tech names lower on tariff news), META -4.1% (mega cap tech names lower on tariff news), GCT -3.9% (increases share repurchase program by $16 mln), NVDA -3.2% (mega cap tech names lower on tariff news), LNW -3% (provides update on litigation; reaffirms 2025 targets), GOOG -2.7% (mega cap tech names lower on tariff news), LCID -2.1% ($1 bln convertible notes offering; also announces Q1 production and deliveries), HVT -2% (in sympathy with RH earnings and downbeat guidance), MSFT -1.9% (mega cap tech names lower on tariff news), SSTK -1.5% (GETY and SSTK receive DOJ request for additional info on merger), NAT -1.4% ($60 mln stock offering), PINE -1.3% (provides Q1 update), HRB -1% (names new chairman), UNIT -0.8% (stockholders approve merger with Windstream), UMH -0.5% (Q1 operations update), FORM -0.2% (doubles capacity at Taiwan service center), RLI -0.1% (partnership with TruckerCloud)
Trump to Outline TikTok Deal Proposal, With ByteDance Retaining Stake
President Donald Trump plans to announce later on Wednesday his plan for a new company called TikTok America, part of his efforts to broker a deal to keep TikTok alive in the U.S., according to a person close to the discussions.
It’s unclear whether the Trump administration has approval for the proposed structure from the Chinese government. Under Trump’s proposal, TikTok America would be roughly 50% owned by new U.S. investors and license TikTok’s algorithm from ByteDance, the person said. Existing investors in ByteDance would have a roughly one-third stake in the new company, while ByteDance would retain a 19.9% stake.
The Takeaway
The proposed structure would put ByteDance’s TikTok ownership just under the 20% limit required by a U.S. sell-or-ban law passed last year.
That structure would put ByteDance’s ownership just under the 20% threshold required in the U.S. law passed last year requiring TikTok to sever ties with its parent company or face a ban. With the new structure, Trump could deem a qualified divesture has happened under the law, which was passed by Congress last year and upheld by the Supreme Court.
The White House and TikTok did not immediately respond to requests for comment.
The planned announcement comes ahead of a Saturday deadline, which Trump set to delay enforcement of the divest-or-ban law. The Trump administration is expected to spell out a timeframe of roughly 90 to 120 days from Wednesday to finalize a deal, the person said. The proposed new company name is not finalized and could change, the person said.
It’s unclear which new investors will participate in the deal. In recent days, media outlets have reported that Andreessen Horowitz and private equity firm Blackstone have mulled taking a stake. Amazon also submitted a last-minute bid to acquire TikTok’s U.S. operations, according to a person familiar with the bid, though parties involved in the TikTok sale talks do not appear to be taking Amazon’s bid seriously. The New York Times earlier on Wednesday reported Amazon’s interest.
Nvidia Faces Dilemma After Chinese Firms Rush to Order $16 Billion in New AI Chips
The Takeaway
•Chinese firms are stockpiling Nvidia H20 chips ahead of a potential U.S. ban on their sale
•Nvidia hasn’t booked enough chipmaking capacity to fill the new orders
•Customers could request refunds if the U.S. restricts H20s before their delivery
Chinese companies including ByteDance, Alibaba Group and Tencent Holdings have placed at least $16 billion in orders for Nvidia’s H20 server chips for running artificial intelligence in the first three months of the year, ahead of a potential U.S. ban on sales of the chip in China, according to two people with direct knowledge of the transactions.
The orders represent a sizable windfall for Nvidia—but only if it can deliver the chips in advance of the new restrictions. Nvidia generated $17 billion from sales in China in the 12 months that ended Jan. 26, 13% of its total revenue.
The surge in demand presents a challenge, as Nvidia must decide whether to proceed with making the chips so it can pocket the revenue or risk wasting the effort if the ban materializes.
The H20 is the most powerful AI chip Nvidia is permitted to sell to China since U.S. export controls banned sale of the company’s most advanced chips in the country based on national security concerns. H20s are 15 times slower in making calculations than Nvidia’s Blackwell chips, which the company sells in most markets outside China.
However, Nvidia recently upgraded the H20 by combining it with the same powerful, high-bandwidth memory chips it uses for the Blackwell system, according to the two people with direct knowledge of the H20 orders. Most Chinese companies have ordered the upgraded version, which bolsters performance for training and running AI, these people said.
President Donald Trump’s administration has been considering banning the sale of H20s to China to further curb Beijing’s technological rise, according to two people with direct knowledge of the deliberations. The threat of a possible ban prompted China’s technology incumbents to increase their orders in recent weeks.
Nvidia declined to comment. ByteDance, Alibaba and Tencent did not reply to requests for comment. The White House didn’t immediately respond to a request for comment. Bloomberg earlier reported on the potential H20 restriction.
Chinese companies have been steadily increasing their orders for the H20 since the January release of an AI model from DeepSeek that operates more efficiently than some American-made models from firms such as OpenAI. This development made H20 chips more useful to the Chinese companies.
The breakthrough by DeepSeek, which is based in China, similarly sparked demand for Nvidia gaming chips, which Chinese firms have retrofitted to run AI. Businesses around the world have also been flocking to DeepSeek models to save on AI costs.
The surge in demand for H20s poses a dilemma for Nvidia, however.
Nvidia hasn’t booked enough chip production capacity at its manufacturing partner, Taiwan Semiconductor Manufacturing Co., to meet increasing demand, the two people said. Nvidia and TSMC can ramp up capacity, but producing so many H20 chips could take up to six months, meaning many of the orders won’t be ready for delivery until the fourth quarter, the people said.
If Nvidia decides to move forward with reserving additional chipmaking capacity for H20s with TSMC, that would run the risk of having to find new buyers when the ban comes into effect before it can deliver the orders. Given that the H20 is much inferior compared to the Blackwell, Nvidia might have to offer substantial price concessions in order to sell them to non-Chinese customers.
And if the U.S. moves to restrict H20s before Nvidia delivers the chips, customers could request refunds, and the company would have to find new buyers for the inventory, the two people said.
TSMC said it doesn’t comment on individual customer or market speculation.
‘Formidable’ Huawei
Nvidia CEO Jensen Huang has repeatedly cautioned against strict export controls, saying such measures could strengthen local Chinese competition that seeks to take the lead on supplying AI chips in the country. He previously called Huawei Technologies, which has been spearheading China’s efforts to design its own AI chips, the “single most formidable” tech company in the country.
The share of Nvidia’s revenue from China has been declining since the U.S. first introduced chip export restrictions in late 2022, a year in which the company generated 21% of revenue from that country. But in dollar terms, China sales rose 70% in the fiscal year that ended Jan. 26 as Nvidia saw a boom in demand for its chips in the post-ChatGPT era.
However, that may not reflect the totality of sales to Chinese firms. Singapore rose to become the second-largest market for Nvidia after the U.S., contributing 18% of its revenue during the same period. Yet Nvidia said it ships the vast majority of those sales outside Singapore, because its customers use the city-state to “centralize invoicing.” Singapore is a popular business hub for Chinese firms’s overseas operations.
Chinese firms such as ByteDance also can purchase or rent Nvidia’s most advanced chips as long as they use them outside China. For instance, ByteDance has been renting Nvidia chips in the U.S. from Oracle, as well as using data centers equipped with Nvidia chips in Singapore and other Southeast Asian countries.
It isn’t clear if the new H20 orders could increase Nvidia’s ballooning accounts receivable, the unpaid balances from products or services it has delivered.
The company’s accounts receivable surged to $23 billion as of the end of January from $10 billion a year earlier—far outpacing revenue growth. The trend prompted some investors to question whether Nvidia might have put some revenue on its books too soon.
The growth in accounts receivable could also be a result of Nvidia entering a new product cycle with Blackwell chips. It takes time for customers to get the new chips up and running, as they require a different cooling mechanism in data centers. Customers may not make payments until the chips are humming.
>>> Up
* ABB Raised to Buy at DNB Markets; PT 52.98 Swiss francs
* ABB Raised to Buy at DNB Markets; PT 52.98 Swiss francs
* Airbnb Raised to Buy at Punto Casa de Bolsa; PT $148.52
* AXA Raised to Buy at Goldman; PT 44 euros
* Barco Raised to Buy at ING; PT 17 euros
* Broadcom Raised to Overweight at Guotai Junan Sec; PT $254
* Bunzl Raised to Buy at Stifel; PT 3,500 pence (+)
* Cellavision Raised to Buy at Pareto Securities; PT 210 kronor
* Engie Raised to Buy at Jefferies; PT 20 euros
* Fiserv Raised to Buy at Goldman; PT $260 (++)
* Ipsen Raised to Buy at Deutsche Bank; PT 122 euros
* Rathbones Group Raised to Outperform at RBC; PT 2,000 pence
* Schibsted Raised to Buy at Pareto Securities; PT 330 kroner (+)
* Friedrich Vorwerk Group PT Raised to 63 euros at Berenberg (++)
>>> Down
>>> Down
* Abrdn Asia Focus Cut to Positive at Stifel
* Aedifica Cut to Neutral at JPMorgan; PT 69 euros
* Altria Cut to Hold at Deutsche Bank; PT $60
* ASML Cut to Neutral at Mizuho Securities; PT 650 euros
* Basic-Fit PT Cut to 21 euros from 26 euros at Berenberg (++)
* BioLine RX ADRs Cut to Hold at Jones
* BNP Paribas Cut to Reduce at Kepler Cheuvreux (+)
* DNO Cut to Hold at DNB Markets; PT 15 kroner
* EDP Renovaveis Cut to Hold at Jefferies; PT 8.70 euros
* Epiroc Cut to Hold at DNB Markets; PT 220 kronor
* Foresight Environmental Cut to Neutral at Stifel
* Lenzing PT Cut to 28 euros from 30 euros at Berenberg (++)
* Lundbeck Cut to Hold at Deutsche Bank; PT 44 kroner
* Lundbeck Cut to Hold at Deutsche Bank; PT 44 kroner
* On Holding PT Cut to $50 from $64 at Needham
* PORR Cut to Neutral at Oddo BHF; PT 29 euros (+)
* Redeia Cut to Hold at Bestinver; PT 18.50 euros
* SocGen Cut to Hold at Kepler Cheuvreux (+)
* Tryg Cut to Neutral at Citi; PT 152 kroner
>>> Initiation
* AB InBev Rated New Buy at Berenberg; PT 72 euros (++)
* AT&S Rated New Buy at Hauck & Aufhaeuser; PT 18.50 euros (+)
* Boston Beer Reinstated Hold at Berenberg; PT $218.10 (+)
* Carlsberg Reinstated Hold at Berenberg; PT 1,033 kroner (++)
* Coca-Cola HBC Rated New Buy at Berenberg; PT 4,068 pence (++)
* D'Ieteren Rated New Outperform at BNPP Exane (+)
* Diageo Reinstated Buy at Berenberg; PT 2,372 pence (++)
* Heineken Reinstated Buy at Berenberg; PT 99.20 euros (+)
* Mosman Oil & Gas Reinstated Speculative Buy at SP Angel (++)
* Molson Coors Reinstated Buy at Berenberg; PT $86.90 (+)
* Pernod Ricard Reinstated Buy at Berenberg; PT 114 euros (+)
* Remy Cointreau Reinstated Hold at Berenberg; PT 47.60 euros (+)
* Royal Unibrew Reinstated Hold at Berenberg; PT 597.10 kroner (+)
* Strategy Rated New Buy at Clear Street; PT $422
* YouGov Rated New Neutral at JPMorgan; PT 400 pence
>>> Call
* YouGov Rated New Neutral at JPMorgan; PT 400 pence
>>> Call
* Bluenord Falls as DNB Cuts PT, Sees ‘Tough Start’ to 2025 (++)
* Epiroc Falls as DNB Cuts to Hold as Macro Debates Dominate (++)
* Richemont, Swatch Price Target Cut at Vontobel; Shares Fall (++)
* Richemont, Swatch Price Target Cut at Vontobel; Shares Fall (++)
* Straumann Shares Fall as Goldman, HSBC Cut PT: Street Wrap (++)
AI Search Is New Arms Race for Retailers
The Takeaway
• Retailers overhaul product listings, websites to cater to AI searches
• Brands post more on Reddit in bid to show up in ChatGPT recommendations
• Perplexity tries to limit SEO in shopping searches
Earlier this year, employees at online diaper seller Coterie noticed customers arriving from an intriguing new source—ChatGPT.
Coterie, like most brands, asks its customers how they heard about the company after they make a purchase. The typical answer is word of mouth. But in recent months, some shoppers started crediting their purchase to OpenAI’s popular artificial intelligence chatbot, which added real-time search features last fall, making it a more important source of inspiration for shoppers—and of business for retailers.
“[Generative] AI is now its own marketing channel,” said Ankur Goyal, Coterie’s senior vice president of growth. “People are learning about us from ChatGPT. People ask, ‘What’s the best diaper?’”
Now Coterie is trying to learn how to influence the way it appears in AI-generated product recommendations, Goyal said, and find out whether it can use techniques similar to traditional search engine optimization, like choosing specific keywords to mirror common search phrases. The company is also weighing whether to buy software tools that claim to monitor AI search results.
Online retailers are racing to revamp their websites and product listings to cater to questions users are asking chatbots including ChatGPT, the Perplexity search engine and Amazon’s Rufus shopping assistant. They’re also trying to decipher the sources, such as Reddit or Wikipedia, these AI tools are using for answers to better understand how their products are represented in AI search results.
Chatbots, shopping assistants and other AI tools are funneling more potential customers to online merchants, with traffic to U.S. commerce sites soaring 1,200% in February compared with July 2024, according to a recent Adobe report.
To be sure, AI search is tiny compared to traditional search, and the Adobe report said only 39% of consumers have used an AI search engine. But rapid growth in AI search over the past two years points to a potential shift in shopper behavior that could upend how e-commerce sites attract customers, and in turn impact the roughly $800 billion a year retailers spend on digital ads across Google, Amazon, Meta Platforms and others.
Executives at online brands say they have been pleasantly surprised by customers showing up on their sites through AI searches, particularly after ChatGPT introduced the real-time search functions last fall.
But marketers haven’t made much progress in deciphering what makes an AI search engine tick, especially since ChatGPT does not always provide a link to a product’s page unless a user asks for one. And most AI chatbots and search engines don’t sell e-commerce search ads, so brands can’t pay to appear prominently in results or recommendations. (Amazon began testing ads on Rufus last fall but hasn’t rolled them out broadly.)
SpyGuy, a site that sells cameras and recording equipment, gets the majority of its traffic from Google Shopping, the search giant’s commerce-focused product comparison feature, according to Allen Walton, SpyGuy’s CEO and founder. But ChatGPT has also started showing up as a source of traffic in the brand’s conversion data, which shows what source or site directed a shopper to its site leading up to a purchase.
“Nobody knows what’s going in the training data for AI—nobody knows how to get your products ranked in AI,” Walton said. “There’s a lot of discussion—what are the things we need to be doing right now?”
Walton said he and other merchants he talks to have tried to influence AI search results, including posting frequently about their brands on Reddit, a popular forum site. Reddit, a major source of training data for AI companies, inked a content licensing deal last year with OpenAI. Reddit and OpenAI didn’t respond to requests for comment.
AI search optimization startups such as Profound, which launched less than a year ago with $3.5 million from investors including Khosla Ventures, are pitching services they say can help brands crack the AI search mystery.
Profound, whose customers include mattress cover brand Eight Sleep and corporate card startup Ramp, tries to figure out how AI search engines work by generating thousands of variations on a search prompt, such as “best dog food.” It runs those prompts through tools like ChatGPT, Perplexity and Microsoft Copilot, measures how often a given brand and its competitors appear in the answers, and keeps track of the sources for each result.
Once brands understand more about AI searches, Profound says, they can create additional pages or content on their site to cater to a specific AI search—for example, a landing page focused on a particular product or an FAQ section that addresses common questions like pricing or troubleshooting.
“Every brand on the planet has one new, very big, very important customer that’s called ChatGPT,” said James Cadwallader, Profound’s CEO and co-founder.
Similarly, consultants who work with brands that sell on Amazon are trying to boost client product appearances in Rufus, a shopping assistant and product comparison tool the e-commerce giant launched on its retail site last year.
Liran Hirschkorn, CEO of Incrementum Digital, a consulting firm that helps brands manage their Amazon storefronts, said he’s telling clients to beef up product listing pages to cater to Rufus, instead of prioritizing eye-catching images that would grab shoppers’ attention in normal Amazon search results. Sellers can come up with effective copy for their listings by looking at common question suggestions Rufus lists on product pages or by looking at similar listings from competitors, Hirschkorn said.
That can reduce the likelihood that Rufus will say it doesn’t have any information in response to a customer question or will give an incorrect answer, which could turn shoppers off. “Customers in general don’t read all the text—but Rufus is reading the text,” he said.
Still, there are limitations to how much insight AI optimization startups and consultants can provide. For example, Profound has discovered that while shoppers are using AI tools like ChatGPT for inspiration, many don’t click through from AI search results to retailers’ sites, instead opening another browser tab to complete a purchase.
And some AI search engines wish to avoid having searches as gameable as Google’s, saying they want to focus on giving users the most personalized and relevant suggestions possible. Of course, Google doesn’t give an explicit look under the hood either, but marketers have had decades to test strategies and learn what works.
Perplexity, which had more than 15 million active users as of last year and generates revenue mainly from selling premium search subscriptions, doesn’t disclose to brands how it determines which products are featured in answers to shopping-related searches or queries about the best products.
“We don’t have any way for someone to change what an answer is,” said Dmitry Shevelenko, Perplexity’s chief business officer. “My answer to brands is build the best product and have that be reflected in the reviews and what others say about it, and then it will naturally rise to the top.”
He concluded, “Our entire way of operating is so that SEO doesn’t emerge in this new category.”