Daily: Beijing limits H200 orders, despite Nvidia optimism
13.5 min read.
Happy 2026! This is the first newsletter of the year and as is often the case, there is a lot to catch up on.
Highlights
Beijing to limit H200 orders. Despite Jensen Huang saying in recent days that Chinese demand for the H200 chips were “high” and that production was ramping up, the Chinese government has issued an overnight missive to some Chinese firms telling it to stop orders for the H200 chips.
As I wrote last year, I was skeptical that Beijing would completely open up to Nvidia’s advanced AI chips, even if they were substantially better than domestic alternatives. It simply wasn’t consistent with a lot of the aggressive policy directions towards semiconductor self-sufficiency. It is still unclear what exactly Beijing’s order was or its extent, but it certainly casts some cold water on Jensen’s optimism for a large deal to China.
Nvidia’s Rubin set for 2H26. The Consumer Electronics Show (CES) played host to a series of announcements by semiconductor companies. Nvidia’s next generation of Rubin chips will begin shipping to customers in the second half of this year, as promised. AMD released a new MI440X chip for use in smaller data centres. Intel and Qualcomm each released new processors for computers.
Memory supercycle continues. Samsung, which had been struggling only a year ago, is back in full force, tripling its quarterly profit year-on-year. Its HBM4 is reportedly competitive, recovering from its lacklustre HBM3, and Samsung Foundry has inked major deals over the past year, including a new deal with Qualcomm for its 2nm node in the works. Above all, the AI boom has supercharged the memory industry.
Memory makers were some of the biggest business winners in 2025, with firms Sandisk, Seagate, and Western Digital making some of the largest stock gains last year. The memory crunch is expected to continue in 2026.
Thanks for reading.
1. Policy and Geopolitics
1.1
WSJ (01/06): Nvidia CEO Says Chinese Demand for Its AI Chips Is ‘Quite High’
Chinese demand for Nvidia’s H200 advanced artificial-intelligence processors is “quite high,” Nvidia Chief Executive Jensen Huang said, a month after the Trump administration made the controversial decision to approve the chips’ sale in China.
“We’ve fired up our supply chain, and H200s are flowing through the line,” Huang said Tuesday during a question-and-answer session at the Consumer Electronics Show here. “We’re getting the last details of the licensing finished with the U.S. government.”
1.2
Reuters (01/08): China asks tech firms to halt orders for Nvidia’s H200 chips, Information reports
Beijing has asked some Chinese tech companies to halt orders for Nvidia’s H200 chips this week, and is expected to mandate domestic artificial intelligence chip purchases, the Information reported on Wednesday, citing people familiar with the matter.
1.3
Reuters (12/31): China mandates 50% domestic equipment rule for chipmakers, sources say
China is requiring chipmakers to use at least 50% domestically made equipment for adding new capacity, three people familiar with the matter said, as Beijing pushes to build a self-sufficient semiconductor supply chain.
The rule is not publicly documented, but chipmakers seeking state approval to build or expand their plants have been told by authorities in recent months that they must prove through procurement tenders that at least half their equipment will be Chinese-made, the people told Reuters.
The mandate is one of the most significant measures Beijing has introduced to wean itself off reliance on foreign technology, a push that gathered pace after the U.S. tightened technology export restrictions in 2023, banning sales of advanced AI chips and semiconductor equipment to China.
1.4
Bloomberg (12/26): China Starts State-Backed Venture Funds to Support Tech Startups
China has officially launched a national venture capital fund and three large regional funds, together worth billions of dollars, as part of a push to foster home-grown tech champions and improve investment efficiency.
The National Startup Investment Guidance Fund, as well as the three vehicles covering the Beijing-Tianjin-Hebei region, the Yangtze River Delta and the Greater Bay Area, began operating on Friday, according to Xinhua News Agency.
The national fund is backed by 100 billion yuan ($14.3 billion) from the Ministry of Finance, financed through the issuance of ultra-long special sovereign bonds, Guo Fangming, a Finance Ministry official, said at a briefing on Friday.
1.5
Reuters (01/02): China AI chipmaker Biren soars in Hong Kong debut as IPO wave builds
Shares of Chinese AI chip designer Shanghai Biren Technology closed up 76% in their Hong Kong debut on Friday, the financial hub’s first listing of 2026.
The company’s shares opened at HK$35.70, hit an intraday high of HK$42.88 and closed at HK$34.46, up 76% from the offer price of HK$19.60.
1.6
Reuters (01/02): Baidu’s AI chip arm Kunlunxin files confidentially for Hong Kong listing
Chinese internet search giant Baidu said on Friday its AI chip unit Kunlunxin has confidentially filed a listing application with the Hong Kong stock exchange on January 1, paving the way for a spin-off and separate listing.
Reuters had earlier reported that Kunlunxin was planning for a Hong Kong initial public offering after completing a fundraising that valued it at 21 billion yuan ($3 billion).
1.7
Reuters (12/31): China’s CXMT eyes $4.2 billion Shanghai listing to fund DRAM expansion
China’s leading maker of DRAM chips, ChangXin Memory Technologies (CXMT) Corp, said on Tuesday it plans to raise 29.5 billion yuan ($4.22 billion) through an initial public offering of 10.6 billion shares in Shanghai.
The company, which directly challenged South Korean and U.S. rivals by unveiling its latest DDR5 DRAM chips last month, will use the listing proceeds to upgrade production lines and technologies, it said in a prospectus on Tuesday.
1.8
Reuters (12/26): South Korea charges 10 over alleged chip technology leak to China’s CXMT
South Korean prosecutors have indicted 10 people on suspicion of leaking memory chip manufacturing technology to Chinese chipmaker ChangXin Memory Technologies (CXMT), a case authorities say helped pave the way for China’s development of high-bandwidth memory (HBM), a critical component for artificial intelligence computing.
The Seoul Central District Prosecutors’ Office said on Tuesday five people, including a former Samsung Electronics executive and engineers, were charged and held in custody for violating South Korea’s industrial technology protection law, while another five were charged but allowed out on bail.
Prosecutors said a former Samsung Electronics researcher who was leaving his job to join CXMT copied out hundreds of steps of proprietary DRAM manufacturing processes by hand, recording detailed process recipes covering equipment specifications, sequencing and yield optimisation. The handwritten notes were later used to reconstruct the manufacturing flow at CXMT, they said.
1.9
Reuters (01/01): U.S. grants TSMC annual licence to import U.S. chipmaking tools into China
The U.S. government has granted an annual licence to Taiwan Semiconductor Manufacturing to import U.S. chip manufacturing equipment to its facilities in Nanjing, China, the chipmaker said on Thursday.
South Korea’s Samsung Electronics and SK Hynix have also received similar import licences.
2. Economy, Finance, and Business
2.1
FT (01/08): Samsung forecasts record profit and signals sustained AI boom
Samsung Electronics has forecast record quarterly earnings, marking a turnaround for the South Korean technology group as demand surges for chips during the global race to build out data centres powering artificial intelligence.
The world’s largest memory chipmaker on Thursday estimated that its fourth-quarter operating profit trebled to about Won20tn ($13.8bn) from a year earlier, surpassing analyst estimates of Won17tn. The company forecast a 23 per cent revenue increase to Won93tn.
The strong guidance highlights the continued AI boom with Samsung now seen as one of the biggest beneficiaries of the chip supply crunch sparked by the industry’s focus on high-bandwidth memory chips used in AI hardware.
2.2
Reuters (01/07): Qualcomm in talks with Samsung Electronics for contract manufacturing, South Korean newspaper says
Qualcomm is in talks with Samsung Electronics over contract manufacturing of two-nanometre chips, the U.S. company’s CEO Cristiano Amon was quoted by the Korea Economic Daily as saying on Wednesday.
The South Korean paper cited Amon as saying that Qualcomm was holding talks first with Samsung Electronics, among several semiconductor foundry firms, about contract manufacturing using the latest two-nanometre process, with design work already completed for commercialisation in the near future.
2.3
Bloomberg (01/07): Samsung Warns of Price Hikes as Rising Memory Costs Affect All
Samsung Electronics Co. expects memory chip supply shortages to raise prices across the electronics industry, including potentially among its own consumer products.
The South Korean company is the world’s largest memory maker, but even its portfolio is not immune to the surging cost of the component essential to everything from smartphones and laptops to connected home appliances and self-driving cars.
2.4
Bloomberg (01/07): Sandisk’s 1,000% Rally Gets Turbocharged by Nvidia CEO’s Remarks
Sandisk Corp. shares jumped 28% Tuesday, their best performance since February, after Nvidia Corp. Chief Executive Officer Jensen Huang highlighted the need for memory and storage in comments at the CES tech conference.
The stock is on a tear, leaping more than 47% in the first three trading sessions of 2026 and soaring 1,080% since hitting a low on April 22. It was the best performer in the S&P 500 Index Tuesday, followed by memory and storage companies Western Digital Corp. and Seagate Technology Holdings Plc, both of which posted double-digit percentage gains.
2.5
FT (01/01): Chip shortages threaten 20% rise in consumer electronics prices
Consumers should prepare for price increases this year, of as much as 20 per cent for smartphones, computers and home appliances, analysts and manufacturers have warned, as artificial intelligence demand drives up the cost of memory chips used in electronics.
Consumer electronics makers including Dell, Lenovo, Raspberry Pi and Xiaomi have warned that chip shortages were likely to add to cost pressures and force them to raise prices, with analysts forecasting increases of 5 to 20 per cent.
2.6
Bloomberg (01/08): Asia Bankers Fear US Scrutiny of Loans Involving Chips, China
Some bankers in Asia are starting to worry about financing deals that would grant Chinese end-users remote access to advanced US chips amid the escalating battle between the world’s two superpowers for artificial intelligence dominance.
At least nine bankers at global lenders that have financed or been pitched such deals have expressed concerns in private that their firms could be caught in the crossfire and face more US scrutiny, according to people familiar with the matter. Some transactions and pitches have sparked internal debate on whether the risk-reward is worth it anymore, the people said, asking not be identified discussing private matters.
2.7
Reuters (01/08): Arm launches ‘Physical AI’ unit, joining rush to robotics by tech and automakers
Chip technology company Arm Holdings has reorganized the company to create a Physical AI unit to expand its presence in the robotics market, company executives told Reuters at CES, where robots are a theme of the year.
The decision to create a unit that specializes in robotics arrives amid a flurry of announcements and activity at CES around humanoid robots. At the sprawling Las Vegas trade show, large and small companies demonstrated robots that could help build cars, clean toilets and deal games of poker - at a glacial pace.
3. Technology
3.1
WSJ (12/24): Nvidia Licenses Groq’s AI Technology as Demand for Cutting-Edge Chips Grows
Nvidia has forged a licensing deal with the chip startup Groq for its AI-inference technology, the companies said Wednesday, a sign of growing demand for cutting-edge AI chips.
Under the nonexclusive deal, Groq’s chief executive officer and founder, Jonathan Ross, will be joining Nvidia, along with its president and some of the startup’s staff. The company, founded in 2016, makes chips and software to run artificial-intelligence models.
Groq’s “language processing unit” chips are built for inference, the everyday process that occurs when consumers or businesses ask trained AI models to provide answers, make predictions or draw conclusions on new data. Ross has said that the company’s design, with embedded memory, means its chips can be produced and deployed faster and use less power than graphics-processing units, which typically consume a lot of energy and are more necessary for training models.
3.2
NYT (01/05): Nvidia Details New A.I. Chips and Autonomous Car Project With Mercedes
Mr. Huang, who spoke during CES, an annual tech conference in Las Vegas, also discussed Nvidia’s surprisingly ambitious work around autonomous vehicles. This year, Mercedes-Benz will begin shipping cars equipped with Nvidia self-driving technology comparable to Tesla’s Autopilot.
Nvidia’s new Rubin chips are being manufactured and will be shipped to customers, including Microsoft and Amazon, in the second half of the year, fulfilling a promise Mr. Huang made last March when he first described the chip at the company’s annual conference in San Jose, Calif.
Companies will be able to train A.I. models with one-quarter as many Rubin chips as its predecessor, the Blackwell. It can provide information for chatbots and other A.I. products for one-tenth of the cost. They will also be able to install the chips in data centers more quickly, courtesy of redesigned supercomputers that feature fewer cables.
3.3
Bloomberg (01/06): Data-Center Cooling Stocks Sink After Nvidia CEO’s CES Talk
Makers of cooling systems saw their stocks sell off Tuesday after comments by Nvidia Corp.’s chief executive raised fears about demand for their products from data centers.
Johnson Controls International Plc, which makes water-cooled chillers and similar systems, slid 6.2% in its worst day since July. Modine Manufacturing Co. closed down 7.5% after paring an earlier drop of as much as 21%. Trane Technologies Plc fell 2.5%.
It’s possible to cool racks with Nvidia’s next-generation Rubin chips using water at a temperature that does not require a water chiller, the company’s CEO, Jensen Huang, said at the CES trade show in Las Vegas. Chillers are a “predominant” piece of the equipment that companies like Johnson Controls and Trane provide to data centers, according to Bloomberg Intelligence.
3.4
Bloomberg (01/06): AMD Unveils New Chip for Corporate Data Centers, Talks Up Demand
Advanced Micro Devices Inc., aiming to make a dent in Nvidia Corp.’s stranglehold on the AI hardware market, announced a new chip for corporate data center use and talked up the attributes of a future generation of products for that market.
The company is adding a new model to its current lineup – one called the MI440X – for use in smaller corporate data centers where customers can deploy local hardware and keep data inside their own facilities. The announcement came as part of a keynote at the CES trade show, where Chief Executive Officer Lisa Su also touted AMD’s top-of-the-line MI455X, saying systems based on that chip are a leap forward in the capabilities on offer.
3.5
Bloomberg (01/06): Intel Shows Off New Computers That Are Central to Comeback Bid
Intel Corp. showed off laptop computers based on processors with a new design, part of the chipmaker’s effort to make its products competitive again.
Updated notebooks from the world’s biggest manufacturers will rely on Intel processors with the Panther Lake design and an improved manufacturing technique, the chipmaker said Monday during a presentation at the CES trade show in Las Vegas. Orders for consumer laptops with the new chips start Tuesday, the company said, with global availability beginning on Jan. 27.
The new laptops will offer consumers a leap in performance, particularly in running AI software, according to Jim Johnson, an Intel senior vice president. The company earlier introduced the Panther Lake technology at an event in October.
3.6
Bloomberg (01/06): Qualcomm Unveils New Cheaper Version of X2 Laptop Processor
Qualcomm Inc., the biggest maker of mobile phone chips, said it’s extending a push into the market for processors that are the heart of personal computers.
Its new X2 Plus processor is a scaled-back version of existing offerings designed to be the main component of more affordable notebook computers. Qualcomm is offering two versions, one with 10 computer cores and one with six, the chipmaker said Monday in a statement. Those cores are the newer Oryon third-generation design. Among their biggest selling points, according to Qualcomm, is the inclusion of a powerful neural processing unit, or NPU, that’ll speed up the responsiveness of AI software without rapidly depleting the battery.
3.7
Bloomberg (01/02): DeepSeek Touts New Training Method as China Pushes AI Efficiency
DeepSeek published a paper outlining a more efficient approach to developing AI, illustrating the Chinese artificial intelligence industry’s effort to compete with the likes of OpenAI despite a lack of free access to Nvidia Corp. chips.
The document, co-authored by founder Liang Wenfeng, introduces a framework it called Manifold-Constrained Hyper-Connections. It’s designed to improve scalability while reducing the computational and energy demands of training advanced AI systems, according to the authors.
Such publications from DeepSeek have foreshadowed the release of major models in the past. The Hangzhou-based startup stunned the industry with the R1 reasoning model a year ago, developed at a fraction of the cost of its Silicon Valley rivals. DeepSeek has since released several smaller platforms but anticipation is mounting for its next flagship system, widely dubbed the R2, expected around the Spring Festival in February.
–