Markets
markets
Luke Kawa

Nvidia reportedly halts H20 production after Chinese security decree clouds demand outlook

Shares of Nvidia are down in premarket trading after The Information reported that the chip designer has told two suppliers that put the finishing touches on its H20 processors — the chips it recently received licenses to sell to China once again — to suspend production work.

This news follows a report earlier this month that China’s internet regulator told major domestic tech giants like ByteDance, Alibaba, and Tencent not to purchase these chips because of data security concerns. Per The New York Times, Nvidia CEO Jensen Huang said he already made it “very clear” to Chinese regulators that their worries about backdoor access to these chips are unfounded.

The H20 has been a giant, multibillion-dollar headache for Nvidia and a flashpoint for the confusing geopolitical, commercial, and technological crosscurrents in the US-China relationship this year.

This nerfed version of Nvidia’s H100 chip was developed specifically for sale to China in response to export controls introduced by the Biden administration. Near the height of trade tensions with China in April, the Trump administration enacted fresh export restrictions on the sale of these chips. Nvidia took a $4.5 billion impairment charge in its Q1 earnings tied to this export ban, and said that its Q2 sales guidance would have been $8 billion higher if not for this change to trade policy.

After an intense public and private lobbying campaign, Nvidia (and Advanced Micro Devices) managed to receive assurances that they would be able to sell their tailor-made AI chips to China once again in mid-July. But the chip designers formally received those export licenses only after striking a novel deal to send 15% of revenues from those sales to the US government.

Nvidia had planned to sell down only its existing H20 inventory to China after it got the initial all-clear, but then reportedly elected to order more H20 chips from TSMC because demand for these processors was so hot — only to then see it seemingly doused by Chinese regulators.

Who knows what the twists and turns for the H20 mean for its successor model that’s in development, as China’s data security concerns surrounding the US chip designer’s products may be also colored by a desire to help promote domestic champion Huawei’s offerings.

More Markets

See all Markets
markets
Helena Cheng

China reportedly planning $295 billion data center network to power AI buildout

Beijing may spend roughly $295 billion (2 trillion yuan) over the next five years to build a nationwide network of AI-focused computing hubs according to a Bloomberg report.

The blueprint would connect data centers across the country into a unified computing network while prioritizing domestic suppliers such as Huawei for much of the underlying technology. State-owned telecom giants including China Mobile and China Telecom would operate much of the infrastructure, according to the report.

The proposal, still under discussion, would mark one of China's most aggressive efforts yet to build an AI infrastructure stack largely independent of US technology.

The AI race is increasingly becoming a competition not just over models and chips, but over access to computing power itself.

China's latest push suggests Beijing has increasingly treated computing power as a strategic national resource, similar to electricity or transportation infrastructure. The latest blueprint would push that strategy further by connecting fragmented regional data centers into a national computing network.

According to the latest Digital China Development Report issued by the China National Data Administration, the country had more than 13.7 million standard server racks in operation by the end of 2025 and had built 42 large-scale AI computing clusters. China's total intelligent computing capacity reached 1.59 million PFLOPS, ranking second globally.

A Chinese planning document from the Ministry of Industry and Information Technology targets 2028 for connecting major computing hubs into a unified national system. Much of that infrastructure is expected to be concentrated in regions such as Inner Mongolia, Ningxia, and Gansu, where abundant land and relatively inexpensive power can support energy-intensive AI workloads.

The Chinese documents also highlight the scale of China's AI ambitions. The country now has more than 6,200 AI companies and an AI industry worth more than $176.9 billion (1.2 trillion yuan), according to official data.

The timing is notable. In May, Washington cleared around 10 Chinese firms to buy Nvidia’s H200 chips, easing some restrictions aimed at slowing China's AI development.

Bloomberg reported the project could be funded primarily through sovereign debt and state-backed investment funds, underscoring China's willingness to continue spending on strategic technologies even as broader economic growth slows.

markets
Claire Yubin Oh

AST SpaceMobile rises after announcing June 17 launch date for its BlueBird 8, 9, and 10 satellites

AST SpaceMobile is up 6% in the premarket action just before the start of regular trading, after the space-based cellular broadband network operator announced that its Bluebird 8, 9, and 10 satellites will be launched on June 17 from Cape Canaveral, Florida.

Adding its BlueBird 8, 9, and 10 to its constellation, each satellite featuring the largest commercial communication array ever of ~2,400 square feet, is expected to further expand AST SpaceMobile’s direct-to-device broadband reach and nearly double the peak data speeds compared to its own initial Block 1 BlueBird satellites, per the company’s press release.

The company has seen its shares plummet in recent weeks after a Blue Origin rocket, which was to carry AST’s Block 2 BlueBird satellite, exploded while testing.

markets
Claire Yubin Oh

Applied Digital leaps on $5.2 billion deal with undisclosed US hyperscaler

Like other AI-adjacent stocks, Applied Digital has hit a bit of a speed bump of late, caught up in the malaise that sent the wider market tumbling at the end of last week. However, after unveiling a new lease agreement with an undisclosed US-based hyperscaler worth at least $5.2 billion, the stock is soaring once again today in premarket trading, up more than 11%.

The deal is with a “high investment-grade hyperscaler,” per the company’s press release, and will cover 210 megawatts of critical IT load at the company’s Delta Forge 2 AI Factory campus under a take-or-pay structure (in which the buyer is obliged to pay a minimum of $5.2 billion over 15 years) with renewal options.

If all renewal options are exercised, the deal would be worth approximately $12.7 billion over a 30-year total term. Initial operations at the Delta Forge 2 site are expected to commence in the first quarter of 2028.

Emphasizing the company’s “franchise model — a core team of design, construction, and operations professionals replicated across every campus, in every market,” CEO Wes Cummins noted that the latest lease is Applied Digital’s third long-term agreement with the same hyperscaler. The agreement also brings the company’s total base-term lease revenue to $36 billion, rising to $86 billion if all options are taken up.

Tasha Matsumoto
18h
Stocks climb higher as traders buy the dip in chip stocks

While market breadth was weak, gains in semiconductors helped propel the S&P 500 and Nasdaq 100 higher.

Group of bulls being herded
markets
Helena Cheng

Nvidia and SK Hynix strike multiyear partnership on memory chips, AI data center build-out

Nvidia shares are modestly higher after it announced a multiyear partnership with SK Hynix on memory chips and building out AI data centers.

The agreement secures a long-term pipeline of memory chips for Nvidia. At the center of the partnership is the integration of SK Hynix’s high-bandwidth memory chips into Nvidia’s newly unveiled Vera central processing units. The Vera processor is Nvidia’s first stand-alone data center microprocessor designed to compete directly against traditional enterprise server lines.

The collaboration is also structured to reshape how semiconductors are manufactured. Under the terms of the agreement, SK Hynix will implement Nvidia’s CUDA-X library and PhysicsNeMo framework directly into its memory design and manufacturing workflows.

The announcement happened during a high-profile visit to Seoul by Nvidia CEO Jensen Huang, who arrived on June 5 to align with core infrastructure partners. Over the weekend, Huang met with SK Group Chairman Chey Tae-won, SK Hynix CEO Kwak Noh-Jung, and other top South Korean technology executives during a dinner meeting, according to Nvidia’s blog posts and Reuters.

Last week, SK Hynix told investors that its proposed US listing has received strong backing, which would potentially give US investors an alternative way to play the memory chip crunch.

Latest Stories

Sherwood Media, LLC and Chartr Limited produce fresh and unique perspectives on topical financial news and are fully owned subsidiaries of Robinhood Markets, Inc., and any views expressed here do not necessarily reflect the views of any other Robinhood affiliate, including Robinhood Markets, Inc., Robinhood Financial LLC, Robinhood Securities, LLC, Robinhood Crypto, LLC, Robinhood Money, LLC, Robinhood U.K. Ltd, Robinhood Derivatives, LLC, Robinhood Gold, LLC, Robinhood Asset Management, LLC, Robinhood Credit, Inc., Robinhood Ventures DE, LLC and, where applicable, its managed investment vehicles.
©2026 Sherwood Media, LLC