Shares of South Korean chip heavyweights Samsung Electronics and SK Hynix surged Thursday, a day after the two companies partnered with artificial-intelligence major OpenAI as part of the U.S. firm’s Stargate initiative.

Samsung shares hit their highest since January 2021, rising over 4%, while SK Hynix stock surged more than 9% — highest since 2000.

OpenAI said in a statement that this partnership will “focus on increasing the supply of advanced memory chips essential for next-generation AI and expanding data center capacity in Korea.”

The ChatGPT-maker said the two South Korean firms plan to scale up production of advanced memory chips, which are critical to power its AI models.

The announcement came as OpenAI CEO Sam Altman met with South Korean President Lee Jae Myung in Seoul, and the top leaders at Samsung and SK Hynix.

OpenAI has also signed a series of agreements to explore developing next-generation AI data centers in South Korea, including with the Korean Ministry of Science and ICT, telecommunications operator SK Telecom, as well as with Samsung subsidiaries.

Earlier this month, SK Hynix announced that it was ready to mass-produce its next-generation high-bandwidth memory chips, cementing its leading position in the AI value chain. HBM is a type of memory that is used in chipsets for artificial-intelligence computing, including in chips from global AI giant Nvidia — a major client of SK Hynix. 

HBM4 chips are expected to be the main AI memory chip needed for Nvidia’s next-generation Rubin architecture — a more powerful AI chip for global data centers.

SK Hynix has been a main chip supplier to Nvidia, while rival Samsung has reportedly been working to get its HBM4 chips certified by Nvidia. 

Samsung has traditionally been the market leader in memory, but its position has been threatened by SK Hynix that has taken a lead in the HBM space. A report from Counterpoint Research in July found that SK Hynix had caught up with Samsung’s memory revenues in the second quarter, with both now vying for the top position in the global memory market. 

Samsung’s second-quarter earnings missed expectations, as profits from its chip business declined almost 94% year on year, although its Chief Financial Officer Soon-cheol Park said that the company expects a rebound in the second half of the year.

— CNBC’s Dylan Butts contributed to this report.

Nvidia this week said it’ll invest $100 billion into OpenAI in a deal that highlights just how big the chipmaker’s investment portfolio has become since the arrival of generative AI in 2022.

That deal came just one week after Nvidia committed a $5 billion investment into one-time rival Intel, and after the company announced its intention to make $500 million investment into self-driving car startup Wayve and a £500 million ($667.7 million) investment into U.K. cloud provider Nscale.

Nvidia’s investment spending spree underscores the chipmaker’s ascendance to the top of Silicon Valley’s pecking order, providing capital and access to its highly desired artificial intelligence chips in exchange for equity and insight into where some of the hottest AI startups are headed.

If the full OpenAI investment is completed — it is expected to be carried out over an unspecified number of years — it would represent Nvidia’s largest investment ever.

Nvidia in August disclosed in a financial filing that it owned $4.33 billion in publicly traded holdings, including Applied Digital, Arm, CoreWeave, Nebius Group, Recursion Pharmaceuticals and WeRide.

At the end of July, Nvidia valued its nonmarketable equity securities at $3.8 billion, up from $1.8 billion a year ago.

The majority of Nvidia’s portfolio companies have some strategic connection to the company’s business, either developing complementary technology to its chips, selling rented access to its chips or using the chips for AI, enterprise software or robotics.

But just because Nvidia is on a company’s cap table doesn’t mean it’s one of the chipmaker’s customers.

“We do not require any of the companies we invest in to use Nvidia technology,” a company spokesperson told CNBC.

For example, Nvidia’s deal with OpenAI merely makes it the “preferred” computing power supplier to the startup, not an exclusive provider. Cohere, an enterprise AI startup that Nvidia participated in funding rounds for, announced this week that it will use AMD chips in addition to Nvidia’s.

In 2022, OpenAI launched ChatGPT and made the broader world aware of the importance of Nvidia’s graphics processing units. Since then, Nvidia’s market cap has grown from just over $420 billion to about $4.3 trillion while its annual revenue has increased by 383% from $27 billion in the company’s fiscal 2023 to $130.5 billion in the fiscal year ended in January.

The year ChatGPT first debuted, Nvidia made 16 investments into other companies, including seed rounds and stakes acquired through incubators, according to a CNBC analysis of PitchBook data. Nvidia’s investments rose to 41 in 2024, and so far in 2025, the chipmaker has made 51 such deals, not counting its commitment to OpenAI.

To be sure, Nvidia is usually just one of several investors for most of its venture capital transactions. Nvidia declined to provide CNBC dollar figures, but its sheer number of venture and public equity investments shows how central the company has become to the AI ecosystem.

Nvidia has increasing cash flow from its rising sales and the regulatory environment around acquisitions is still difficult, said Neuberger Berman analyst Jamie Zakalik. The OpenAI deal is a “win-win” because it allows Nvidia to spend some of that cash and gain influence on the implementation of artificial intelligence using its chips, she said.

“Nvidia has a lot of capital that they don’t necessarily have a ton of avenues to do stuff with,” Zakalik said.

Regulators typically don’t worry about “vertical” investments or acquisitions in which a company like Nvidia takes a stake in a supplier or customer, said Alden Abbott, a research fellow at the Mercatus Center at George Mason University. But regulators would scrutinize investments with exclusive supply contracts, which are unlikely in this case because of how fast AI is moving, he said.

These investments could also be a clue as to what companies Nvidia may be interested in buying. Earlier this year, the chipmaker acquired CentML after participating in its 2023 seed round. Nvidia also invested in Enfabrica, an infrastructure networking firm, before spending $900 million to hire its CEO and license its technology, CNBC reported last week.

The chipmaker has invested an unspecified amount in OpenAI in multiple rounds, that latest of which valued it at $500 billion. It contributed to a 1.7 billion euro round ($2 billion) into French startup Mistral AI in September, participated in a round that raised $100 million for Cohere in August and was part of a round that raised $307 million for Runway AI over multiple rounds.

Nvidia’s also reportedly participated in a April $2 billion round for Safe Superintelligence, the lab started by former OpenAI luminary Ilya Sutskever. Nvidia also participated in a $2 billion July funding round for Thinking Machine Labs, the company started by former OpenAI operating chief Mira Murati.

Some of Nvidia’s investments have already seen exits. It was an investor in Scale AI, a company that cleaned and created data for AI startups. That startup entered into a $14.3 billion hiring and licensing deal with Meta that saw its CEO and other employees join the social media company.

Nvidia also has 7% stake in CoreWeave, a cloud provider that sells access to Nvidia’s chips. CoreWeave had a successful initial public offering earlier this year.

CoreWeave is one of the neo-cloud providers that are competing and partnering with the large cloud companies such as Microsoft, Google and Amazon. CoreWeave disclosed a $6.3 billion order from Nvidia earlier this month.

Nvidia also participated in fellow neo-cloud provider Lambda Labs’ $480 million Series D round in February.

Some of Nvidia’s most interesting investments are in companies that are developing quantum computers, a still-theoretical technology that some believe could one day supercharge AI development. Quantum companies can use Nvidia’s chips to simulate their computers that don’t exist yet.

Nvidia’s stakes in those companies came after CEO Jensen Huang apologized to the industry at an Nvidia conference after one of his stray quips about a pessimistic timeline for the technology sunk their stocks.

Earlier this month, Nvidia’s venture capital arm was a participant in a $1 billion round for PsiQuantum, and the chipmaker pitched into early stage funding for Quantinuum in August.

Having Nvidia on a company’s cap table can make it more attractive to investors. In recent years, after Nvidia has revealed stakes in public companies, investors have bought up the stocks, even with tangential connections to AI. Intel’s stock rose 18% the day that Nvidia announced its investment.

WATCH: Nvidia-OpenAI deal shows capital intensity despite weak commercial monetization: GMO’s Tom Hancock

Shares of Marvell Technology plunged 18% on Friday after the artificial intelligence chipmaker’s data center revenue fell short of estimates and it gave lackluster guidance for the current quarter.

Here’s how the company did in comparison with LSEG consensus estimates:

• Earnings per share: 67 cents adjusted vs. 66 cents expected
• Revenue: $2.01 billion vs. $2.01 billion expected

Revenue jumped 58% from a year ago in the fiscal second quarter that ended Aug. 2, a record for the company that was fueled in part by “strong AI demand” for its custom silicon and electro-optics products, Marvell CEO Matt Murphy said in a statement Thursday.

The company had net income of $194.8 million, or 22 cents per share, compared with a net loss of $193.3 million, a loss of 22 cents per share, during the same period last year.

For the fiscal third quarter, the company called for revenue to be $2.06 billion, plus or minus 5%. That was slightly below the $2.11 billion forecast by analysts, according to LSEG.

Marvell is known for creating customized chips and hardware, which it offers to cloud providers such as Amazon and Microsoft.

Sales in its data center segment reached $1.49 billion during the quarter, which fell short of Wall Street’s projected $1.51 billion, according to StreetAccount.

On a conference call with investors on Thursday, Murphy said the company expects “overall data center revenue in Q3 to be flat sequentially,” which he attributed to nonlinear growth in its custom AI chips business. Fourth-quarter growth is expected to be “substantially stronger” than the third quarter, Murphy said.

He added that “lumpiness” of the guidance is normal as large hyperscalers build out infrastructure.

Still, some investors were hoping for greater clarity on the company’s pipeline of new customers.

“Without this, we find it very difficult underwriting the company’s 20% data center market share target,” Cantor analysts wrote in a Thursday note to clients. “Thus, we wait for more bottoms up granularity before potentially turning more positive.”

Analysts at Bank of America downgraded Marvell’s stock to neutral from buy on Friday and lowered their price target to $78 per share from $90, partly on concerns around the company’s AI growth prospects “in the near/medium term.”

Intel‘s stock dropped more than 8% after the chipmaker said it would slash foundry costs in its latest attempt to turn around its struggling business.

Concerns about where that leaves Intel’s chip manufacturing business overshadowed a better-than-expected earnings report late Thursday. Intel beat on revenue and issued a sales forecast for the third quarter that also topped estimates. The company reported adjusted earnings of 10 cents per share, topping the average analyst estimate of a penny, according to LSEG.

CEO Lip-Bu Tan, who was appointed to the job in March, wrote in a memo to employees that the company’s forthcoming chip manufacturing process, called 14A, will be built out based on confirmed customer commitments and that there will be “no more blank checks.” In a filing with the U.S. Securities and Exchange Commission on Thursday, Intel said it may “pause or discontinue” its foundry business entirely if it could not secure a customer on its next technology cycle.

“We have been unsuccessful to date in securing any significant external foundry customers for any of our nodes and our prospects for securing a significant external foundry customer for Intel 14A are uncertain,” the company said in the filing.

Intel’s drop on Friday wiped out most of its rally for the year. The shares lost 60% of their value in 2024, their worst year on record. The slump reflected Intel’s inability to make much headway in the artificial intelligence market, which is dominated by Nvidia, as well as skepticism surrounding its foundry bet.

The company said it is axing chip facility projects in Germany and Poland and slowing production at its Ohio plant. Intel depends on a large customer for its foundry business to succeed.

“Management wants external customer commitments to pursue the node, but in the meantime, this adds more uncertainty to product roadmaps and makes customer adoption more unlikely,” analysts at Barclays, who have the equivalent of a hold rating on the stock, wrote in a note to clients.

Tan, who replaced Pat Gelsinger as CEO, said in the memo that his first few months at the helm of the company have “not been easy.” Intel has gone through with most of its layoff plans, which will result in eliminating 15% of its workforce and finishing the year with 75,000 employees.

“Over the past several years, the company invested too much, too soon — without adequate demand,” Tan wrote. “In the process, our factory footprint became needlessly fragmented and underutilized,” he added

Intel’s net loss widened to $2.9 billion, or 67 cents per share, from $1.61 billion, or 38 cents in the year-ago period. The company recorded an $800 million impairment charge, “related to excess tools with no identified re-use.”

Analysts at JPMorgan Chase called Intel’s foundry decision a “positive step,” although ongoing market share losses remain a concern.

Shares in South Korean chipmaker SK Hynix extended gains to hit a more than 2-decade high on Tuesday, following reports over the weekend that SK Group plans to build the country’s largest AI data center.

SK Hynix shares, which have surged almost 50% so far this year on the back of an AI boom, were up nearly 3%, following gains on Monday. 

The company’s parent, SK Group, plans to build the AI data center in partnership with Amazon Web Services in Ulsan, according to domestic media. SK Telecom and SK Broadband are reportedly leading the initiative, with support from other affiliates, including SK Hynix. 

SK Hynix is a leading supplier of dynamic random access memory or DRAM — a type of semiconductor memory found in PCs, workstations and servers that is used to store data and program code.

The company’s earnings outlook has strengthened amid a recent uptick in legacy DRAM pricing, MS Hwang, research director at Counterpoint Research told CNBC.

“Notably, Morgan Stanley — historically bearish on the company — has revised its earnings estimates upward on SK Hynix,” he added.

The company’s DRAM rival, Samsung, was also trading up 4% on Tuesday. However, it’s growth has fallen behind that of SK Hynix.

On Friday, Samsung Electronics’ market cap reportedly slid to a 9-year low of 345.1 trillion won ($252 billion) as the chipmaker struggles to capitalize on AI-led demand. 

SK Hynix, on the other hand, has become a leader in high bandwidth memory — a type of DRAM used in artificial intelligence servers — supplying to clients such as AI behemoth Nvidia. 

A report from Counterpoint Research in April said that SK Hynix had captured 70% of the HBM market by revenue share in the first quarter.

This HBM strength helped it overtake Samsung in the overall DRAM market for the first time ever, with a 36% global market share as compared to Samsung’s 34%. 

“While Samsung is making technological progress, capturing [SK Hynix’s] NVIDIA market share will likely require additional time,” said Counterpoint’s Hwang. “SK Hynix’s dominance in HBM remains firm.”

Cerebras CEO Andrew Feldman said his hope is to take his company public in 2025 now that the chipmaker has obtained clearance from the U.S. government to sell shares to an entity in the United Arab Emirates.

“That’s our aspiration,” Feldman told reporters on Thursday at the company’s Supernova conference in San Francisco, after being asked if an IPO was likely this year.

Cerebras, which makes processors for artificial intelligence workloads, filed to go public in September but hasn’t provided an update on the expected size or timing of an offering. In March, the company said it had obtained clearance from a U.S. committee to sell shares to Group 42, a Microsoft-backed AI company based in the UAE.

That clearance came from the Committee on Foreign Investment in the United States, or CFIUS, and marked a key step for Cerebras in its effort to go public. Cerebras competes with Nvidia, whose graphics processing units (GPUs) are the industry’s choice for training and running AI models. More than 85% of Cerebras’ revenue in the first half of 2024 came from Group 42.

The tech IPO market broadly has been in a drought since early 2022, when rising inflation and higher interest rates pushed investors out of risky assets. Cerebras appeared poised to be the first notable pure-play AI IPO after its filing, but then the came the delay. CoreWeave, which provides AI infrastructure, debuted in March and has seen its market value jump about 65% since its IPO.

The IPO market is showing signs of life, with trading app eToro hitting the Nasdaq this week and digital health provider Hinge Health scheduled to go out next week.

The Middle East is becoming a more critical market for AI development.

Nvidia CEO Jensen Huang was in Riyadh, Saudi Arabia this week along with other tech leaders and President Donald Trump for the Saudi-U.S. Investment Forum. Nvidia said at the event that it will sell more than 18,000 of its latest AI chips to Saudi company Humain.

Group 42 is also reportedly on tap to purchase 100,000 GPUs a year as part of a bigger agreement between the U.S. and UAE.

Feldman said at the roundtable with reporters that it’s “important to be among the big dogs” and said, regarding the latest announcements, “You’ve got half the story. I can’t share the other half.”

In addition to Microsoft, Cerebras sells to Meta and IBM. Feldman said last year that the company would have another “hyperscaler” within the first half of 2025.

“We’re close with another,” he said on Thursday. “I think they haven’t been the quickest to respond.”

Earlier in the day, Cerebras announced the ability to run an open-source model from Alibaba on its chips at what it says is a lower price than what OpenAI’s GPT-4.1 model charges, and at a higher speed.

— CNBC’s Ari Levy contributed to this report.

WATCH: Cerebras postponing IPO

More than two years into the generative artificial intelligence boom, Wall Street is setting an increasingly high bar for chipmakers.

When it comes to earnings reports — most recently from Marvell Technology — good isn’t good enough. That’s because investors previously poured into the companies making the infrastructure and devices at the heart of the AI economy, bidding the stocks up to historically high levels.

They’re demanding results.

Marvell shares plummeted 20% on Thursday, their steepest slump since 2001, after guidance fell short of some elevated estimates. The company’s revenue forecast as well as its results for the latest quarter were all ahead of the average analyst estimate, according to LSEG, but Wall Street wanted more after the stock soared 83% in 2024.

“While Marvell reported a small beat and raise, the guide was definitively below buyside expectations,” analysts at Cantor wrote in a report following the results.

Nvidia suffered a similar fate in late February, with its stock dropping 8.5% the day after the leader in AI processors reported earnings and revenue that sailed past estimates. Shares of Advanced Micro Devices fell more than 6% earlier in February after beating expectations. The one troubling number for AMD was a miss in its data center business.

Optical supplier Credo Technology plummeted 14% after earnings on Wednesday and another 10% during Thursday’s session despite triple-digit revenue growth and upbeat guidance.

The VanEck Semiconductor ETF is down nearly 6% this week following last week’s 7% decline. The ETF, whose leading components are Nvidia, Taiwan Semiconductor Manufacturing Co. and Broadcom, soared 72% in 2023 and nearly 39% last year.

The challenge on Wall Street for chipmakers underscores the pressure they’re under as the AI buildout stretches into its fourth calendar year. Tariffs from the Trump administration and chip export controls have added to investor concerns.

However, not all companies in the space are getting the same treatment. Broadcom shares lost 6% during Thursday’s session in the lead-up to quarterly earnings, but the stock popped 12% after hours on better-than-expected results, including strong infrastructure and semiconductor revenue.

Qualcomm reported fourth-quarter earnings on Wednesday that beat Wall Street expectations for earnings and revenue, and the company guided to a strong December quarter.

The shares rose 10% in extended trading at one point before falling to a gain of about 4%.

Here’s how the company did versus Refinitiv consensus expectations for the quarter ending Sept. 29:

• Earnings per share: $2.69, adjusted $2.56 expected
• Revenue: $10.24 billion versus $9.90 billion expected

Qualcomm said it expects revenue in the current quarter of between $10.5 billion and $11.3 billion, with the midpoint of that range beating LSEG consensus expectations of $10.59 billion.

The company reported $2.92 billion in net income, or $2.59 per share, a sharp jump from last year’s $1.49 billion, or $1.23 per share. Qualcomm reported $33.19 billion in total revenue in its fiscal 2024, a 9% increase from 2023.

Qualcomm’s fortunes have historically been tied to the smartphone industry, where the company provides a range of chips to handset makers, including system-on-a-chip processors, modems, and antennas. The company makes the chip at the heart of most high-end Android devices, and many lower-end phones as well. Qualcomm also sells modems and related chips to Apple for its iPhones, and last year said its contract for 5G chips ran through 2026.

Qualcomm reported a 12% increase in handset chip sales to $6.1 billion, in line with FactSet estimates. Qualcomm introduced its high-end chip for 2025, called Snapdragon 8 Elite, in October.

“In handsets we delivered greater than 20% year-over-year growth in Android revenues,” said Qualcomm CFO Akash Palkhiwala on a call with analysts.

Under CEO Cristiano Amon, the company has diversified away from being a smartphone supplier and has introduced and invested heavily in producing chips for PCs, cars, and industrial machines.

“We will continue to transform Qualcomm from a wireless communications company into a connected computing company for the age of AI,” Qualcomm CEO Cristiano Amon said on the earnings call with analysts.

Qualcomm has also made efforts to brand itself as a leader in AI, having developed smartphone chips with specialized parts for machine learning since 2017. But unlike Nvidia, the company doesn’t produce the kind of graphics processors for data centers that are used for big AI programs like OpenAI’s ChatGPT.

The automotive business grew 86% on an annual basis to $899 million in sales. Qualcomm says it has billions of dollars in business with automakers currently in its development pipeline, and highlighted it was the fifth consecutive quarter of growth. Qualcomm said that it expected automotive sales in the current quarter to rise 50% on an annual basis.

The company’s “internet of things” business includes both chips for industrial purposes as well as the chips Meta uses in its Quest handsets and Ray-Ban Smart Glasses. It also includes the new business selling chips for laptops running Microsoft Windows. The division reported $1.68 billion in revenue, a 22% increase from a year earlier.

Qualcomm’s chip business, including its handset, automotive, and other chips, which together is reported as QCT, saw sales rise 18% during the quarter to $7.37 billion in total. 

The company’s profitable technology licensing business, QTL, reported $1.52 billion in revenue, a 21% increase over the same period last year.

Qualcomm said its board had approved $15 billion in additional buybacks. During the fourth quarter, it repurchased $1.3 billion worth of shares and paid out $947 million in dividends.

WATCH: CNBC’s full interview with Qualcomm CEO Cristiano Amon