Mark Zuckerberg, chief executive officer of Meta Platforms Inc., during the Acquired LIVE event at the Chase Center in San Francisco, California, US, on Tuesday, Sept. 10, 2024.
David Paul Morris | Bloomberg | Getty Images
Meta announced the Quest 3S, the latest virtual reality headset to come out of the company’s Reality Labs division and a cheaper offering than its predecessor.
The device will go on sale on Oct. 15 and will retail starting at $299, down from the $499 starting price for 2023’s Quest 3. The device can be used to watch movies, as well as run VR fitness apps and gaming, Meta said Wednesday at its Connect event at its headquarters in Menlo Park, California. The company positioned the headset as a multitasking computer, putting it in competition with Apple’s $3,499 Vision Pro headset that launched in February.
Meta’s previous Quest devices are the bestselling VR headsets, with millions shipped thanks to heavy marketing and a lower price than many competitors, but those efforts have yet to spark a cultural phenomenon or a mainstream software ecosystem around VR. Including its acquisition of Oculus in 2014, Meta has poured more than $65 billion in expenses into its hardware efforts.
“I’ve been waiting for this one for a long time,” Zuckerberg said.
Meta CEO Mark Zuckerberg has defended the company’s spending as a strategic initiative to prevent Apple from controlling future hardware platforms.
Although there was hope among VR developers that Apple’s entry into the market would spur a wave of new apps and users, Apple hasn’t revealed sales for its headset and reports say that sales have been in small volumes, under 1 million units, partially due to its high price.
What it does
A Meta representative said the “S” stands for “start” — as in getting started with VR.
Many of the new Meta features that the company discussed on Tuesday for its $299 Quest 3S have counterparts on Apple’s Vision Pro, including a mode that allows for the device to be used on an airplane and another that simulates a large movie theater inside the headset.
Meta highlighted improved “passthrough,” the term used to described when a VR headset uses cameras and sensors on the outside of the device to display live real-time video inside the headset. That function is intended to make users feel like they are looking through a display and allows them to interact with the real world while keeping the headset on. For the Quest 3S, Meta added a dedicated button to turn on passthrough.
The company has emphasized the ability of the Quest 3S to multitask and run apps, positioning it as a computing device, instead of a game console.
“All the things you can do with a general purpose computer, Quest is the full package,” Zuckerberg said.
In demos provided Tuesday, Meta showcased the device running as many as four apps at one time on floating screens inside the headset, including a YouTube video, a browser, Amazon Music and Meta’s app store. Meta says the headset can handle six windows. But the demo experience was not smooth. The Amazon Music app crashed, window controls would disappear and Meta’s controllers would fall asleep after a few minutes if the user wasn’t pressing buttons.
Besides the Quest 3S, Meta also announced a price cut for last’s year Quest 3, bringing the price of the 512GB version down from $650 to $500. The Quest 3 has more advanced lenses and a superior screen with a higher resolution than the Quest 3S.
Additionally, Meta said it will discontinue the Quest Pro, its $999 headset launched in 2022 that never gained much momentum.
Eventually, glasses
Meta Orion AR glasses prototype
Meta
Zuckerberg’s justification for spending so much on VR and augmented reality is his belief that the technology will eventually end up in lightweight, transparent glasses that overlay computer graphics and information onto the real world.
Investing in VR software and hardware are early steps toward those glasses, which could take as much as a decade to develop, Zuckerberg has previously said.
Zuckerberg showed off an early concept of what those glasses could look like on Wednesday. The thick, black-framed prototype, called Orion, won’t be sold to consumers, but Meta says they will be used internally as the company continues working toward the consumer glasses it hopes to one day sell.
“This is where we are going,” Zuckerberg said.
Zuckerberg said they were Meta’s first “fully-functioning” prototype of the glasses, and would be physically tethered to a small “puck.” Zuckerberg also said that it would take advantage of a wrist-based interface, which came out of stems from the company’s 2019acquisition of CTRL-Labs.
Meta’s Orion prototype comes a week after Snapannounced its fifth-generation Spectacles AR glasses. Those thick-framed glasses will only be made available to developers, who must commit to paying $99 a month for one full year if they want to build AR apps for the device.
This isn’t the first time Meta publicly revealed a prototype of a future devices or research projects to signal to investors and employees where VR and AR technology is headed. The Orion glasses are an improvement on Project Nazare, prototype smart glasses that Zuckerberg announced in 2021, when the company changed its name from Facebook.
Ray-Ban Meta smart glasses are powered by a Qualcomm chip. Qualcomm, Samsung and Google are working on smart glasses, according to Qualcomm CEO Cristiano Amon.
Nurphoto | Nurphoto | Getty Images
Meta does sell a pair of glasses with a built-in camera in partnership with EssilorLuxottica called Ray-Ban Meta, which start at $299. While these glasses don’t have any displays, they do have tiny speakers that allow the device to play music or interact with Meta AI, the company’s voice assistant.
For example, the Ray-Ban Meta glasses can now detect when a user is looking at a sign in Spanish and, if asked, can translate in the user’s ear, a new improvement, Meta said. The camera can scan QR codes, and it can also extract information like book titles out of photos it takes. Plus, they’re stylish and look like a pair of typical sunglasses.
“One of the most important things about this is they’re just good-looking glasses,” Zuckerberg said.
Another new capability for the glasses is the ability to remember facts like where the user parked.
Li-Chen Miller, the vice president of product in charge of Ray-Ban Meta glasses, told CNBC that when she travels, she uses the glasses to take photos of her hotel room door, and later, she asks Meta AI to recall the number.
Zuckerberg is excited about the Ray-Ban Meta smart glasses, which have sold more than 730,000 units in their first three quarters, according to market researcher IDC. In July, he told investors that they were “a bigger hit sooner than expected.”
Last week, EssilorLuxottica and Meta announced that they had extended their partnership to develop more smart glasses.
AI that speaks
Zuckerberg also introduced improvements to its Meta AI chatbot that will allow people to interact with it using their voice instead of written text.
Users will now be able to have natural voice conversations with Meta AI, which is accessed through Meta apps like Messenger and Instagram. Users will be able to perform actions using their voice, such as telling Meta AI to take a photo by talking to their smartphone.
For Meta AI’s new feature, the company is using computer-generated voices from celebrities including Awkwafina, Judi Dench, John Cena, Keegan-Michael Key and Kristen Bell.
The new Siri-like Meta AI voice feature will be available over the next month for U.S., Canadian, Australian and New Zealand users of WhatsApp, Instagram, Facebook and Messenger.
The feature comes one day after rival OpenAI, the maker of ChatGPT, announced an advanced voice feature for people who pay its premium service.
The company said that the new chatbot features are based on Meta’s AI model, Llama. The company also announced a newer version of Llama, called Llama 3.2. This updated model can understand both images and text, an upgrade from its predecessors which generated responses to people’s written prompts.
Co-founder and chief executive officer of Nvidia Corp., Jensen Huang attends the 9th edition of the VivaTech trade show in Paris on June 11, 2025.
Chesnot | Getty Images Entertainment | Getty Images
Nvidia has just shelled out over $900 million to hire Enfabrica CEO Rochan Sankar and other employees at the artificial intelligence hardware startup, and to license the company’s technology, CNBC has learned.
In a deal reminiscent of recent AI talent acquisitions made by Meta and Google, Nvidia is paying cash and stock in the transaction, according to two people familiar with the arrangement. The deal closed last week, and Enfabrica CEO Rochan Sankar has joined Nvidia, said the people, who asked not to be named because the matter is private.
Nvidia has served as the backbone of the AI boom that began with the launch of OpenAI’s ChatGPT in late 2022. The company’s graphics processing units (GPUs), which are generally purchased in large clusters, power the training of large language models and allow for big cloud providers to offer AI services to clients.
Enfabrica, founded in 2019, says its technology can connect more than 100,000 GPUs together. It’s a solution that could help Nvidia offer integrated systems around its chips so clusters can effectively serve as a single computer.
A spokesperson for Nvidia declined to comment, and Enfabrica didn’t provide a comment for this story.
While Nvidia’s earlier AI chips like the A100 were single processors slotted into servers, its most recent products come in tall racks with 72 GPUs installed working together. That’s the kind of system inside the $4 billion data center in Wisconsin that Microsoft announced on Thursday.
Nvidia previously invested in Enfabrica as part of a $125 million Series B round in 2023 that was led by Atreides Management. The company didn’t disclose its valuation at the time, but said that it was a fivefold increase from its Series A funding.
Late last year, Enfabrica raised another $115 million from investors including Spark Capital, Arm, Samsung and Cisco. According to PitchBook, the post-money valuation was about $600 million.
Tech giants Meta, Google, Microsoft and Amazon have all poured money into hiring top AI talent through deals that resemble acquihires. The transactions allow the companies to bring in top engineers and researchers without worrying about the regulatory hassles that come with acquisitions.
The biggest such deal came in June, when Meta spent $14.3 billion on Scale AI founder Alexandr Wang and others and took a 49% stake in the AI startup. A month later, Google announced an agreement to bring in Varun Mohan, co-founder and CEO of artificial intelligence coding startup Windsurf, and other research and development employees in a $2.4 billion deal that also included licensing fees.
Last year, Google made a similar deal to bring in the founders of Character.AI. Microsoft did the same thing for Inflection, as did Amazon for Adept.
While Nvidia has been a big investor in AI technologies and infrastructure, it hasn’t been a significant acquirer. The company’s only billion-dollar-plus deal was for Israeli chip designer Mellanox, a $6.9 billion purchase announced in 2019. Much of Nvidia’s current Blackwell product lineup is enabled by networking technology that it acquired through that acquisition.
Nvidia tried to buy chip design company Arm, but that deal collapsed in 2022 due to regulatory pressure. In the past year, Nvidia closed a $700 million purchase of Run:ai, an Israeli company whose technology helps software makers optimize their infrastructure for AI.
On Thursday, Nvidia announced one of its most sizable investments to date. The chipmaker said it’s taken a $5 billion stake in Intel, and announced that the two companies will collaborate on AI processors. Nvidia also said this week that it invested close to $700 million in U.K. data center startup Nscale.
— Correction: A prior version of this story mistakenly included the name of a company as an investor in Enfabrica.
CrowdStrike logo is seen in this illustration taken July 29, 2024.
Dado Ruvic | Reuters
CrowdStrike shares popped about 13%, a day after the cybersecurity firm issued better-than-expected long-term guidance at its investor day.
The company on Wednesday said it expects net new annual recurring revenues to grow at least 20% in 2027, ahead of analysts’ expectations. CrowdStrike plans for ARR to hit $10 billion by 2031, and then double to $20 billion by 2036.
“CrowdStrike is by far the most advanced security platform in the industry, and the plethora of AI-based solutions announced today will further separate CrowdStrike from the competition,” wrote Wells Fargo analyst Andrew Nowinski in a note following the event.
Some Wall Street firms also boosted their price targets.
Read more CNBC tech news
Cybersecurity has taken center stage this year as businesses beef up security in the age of artificial intelligence. Many companies have harnessed AI tools to strengthen their offering as threats rise in sophistication.
Nvidia CEO Jensen Huang attends the “Winning the AI Race” Summit in Washington D.C., U.S., July 23, 2025.
Kent Nishimura | Reuters
Nvidia CEO Jensen Huang said that the company’s $5 billion investment and technology collaboration with Intel comes after the two companies held discussions for nearly a year.
Huang said that he communicated personally with Intel CEO Lip-Bu Tan about the partnership. He called Tan a “longtime friend” on a Thursday call with reporters after the companies announced that Nvidia would co-develop data center and PC chips with Intel as part of the investment deal. On the call, Tan said he and Huang have known each other for 30 years.
“We thought it was going to be such an incredible investment,” Huang said.
Nvidia said it will collaborate with the chipmaker to create artificial intelligence systems for data centers that combine Intel’s x86-based central processors with Nvidia’s graphics processors and networking.
Intel will also sell CPUs for PCs and notebooks that integrate Nvidia graphics processors, or GPUs.
The transaction itself took a few months to come together, Intel’s revenue chief Greg Ernst wrote in a LinkedIn post, adding that the agreement was reached on Saturday.
The investment highlights how the fortunes of the two companies have switched atop Silicon Valley’s pecking order as a result of the AI explosion ushered in by OpenAI’s launch of ChatGPT in late 2022.
Intel shares are down 31.78% in the last five years, while Nvidia shares are up 1,348% as of opening prices on Thursday. Nvidia is worth over $4.25 trillion, while Intel is only worth $143 billion.
How Intel and Nvidia will collaborate
For decades, the most important part in a PC or server was the central processor, and Intel dominated the market for those chips. But AI infrastructure, like the machines in the $4 billion data center Microsoft announced on Thursday, often needs two or more Nvidia GPUs for every one CPU.
Nvidia AI systems, like the NVL72 used by Microsoft, come with Arm-based CPUs, instead of Intel x86-based CPUs. On the call, Huang said Nvidia will soon support Intel’s CPUs in its NVLink racks for AI.
“We’ll buy those CPUs from from Intel, and then we’ll connect it into super chips that then becomes our compute node, that then gets integrated into a rack scale AI supercomputer,” Huang said.
Nvidia will also contribute GPU technology to Intel chips that ship in laptops and PCs, which is an underserved market, Huang said. In total, the addressable markets for the two product collaborations are worth $50 billion, Huang said.
“We’re going to become a very large customer of Intel CPUs, and we’re going to be a large supplier of GPU chiplets into Intel” chips, he said.
Huang said the deal with Intel will have “no” impact on Nvidia’s business relationship with Arm.
Thursday’s investment deal is focused on the relationship between Nvidia and Intel’s product division, not its foundry. The two companies, however, did not rule out future foundry partnerships.
“We’ve always evaluated Intel’s foundry technology, and we’re going to continue to do it, but today, this announcement, is squarely focused on these custom CPUs,” Huang said. Nvidia currently uses Taiwan Semiconductor Manufacturing Company to manufacture its chips.
The collaboration will use Intel’s packaging, which is a part chip manufacturing that occurs toward the end of the process and combines several chip components into a single part that can be installed in machines.
Intel CEO Lip-Bu Tan makes a speech on stage in Taipei, Taiwan May 19, 2025.
Ann Wang | Reuters
Tan said he was grateful for Nvidia’s vote of confidence.
“‘I’d like to thank Jensen for the confidence in me, and our team and Intel will work really hard to make sure it’s a good return for you,” Tan said.
Last year, Intel’s board removed previous CEO Pat Gelsinger because of rising costs in its manufacturing business and the company’s failure to gain a foothold in AI chips. In March, Intel named Tan, a well-connected investor who had turned around chip software firm Cadence Design Systems, its new chief executive.
Tan has focused on cutting costs and raising money in his short tenure leading Intel even as the future of the company’s manufacturing business, called Intel Foundry, remains unclear.
In addition to the $5 billion from Nvidia and $8.9 billion from the U.S. government, Intel has taken a $2 billion investment from SoftBank, sold a majority stake in its ASIC subsidiary Altera to Silver Lake for $3.3 billion and sold $1 billion in stock from Mobileye, its self-driving car subsidiary.
Intel has also cut significant staff, saying in July that it would eliminate 15% of its workforce by the end of the year.
The company develops its own chips as well as manufacturing them. It wants to manufacture chips for companies like Nvidia or Apple, but has yet to secure them as customers. Analysts say Intel needs a big foundry client to signal that its technology is stable and ready for volume production.
But cutting-edge chip manufacturing is expensive, and Intel has signaled that if it can’t get enough customers, it may not continue investing in its foundry. That could spark a reaction from Washington, whose politicians and lobbyists consider Intel to be strategically important for the nation because it is the only American company capable of manufacturing the most advanced chips.
The Trump administration took a 10% stake in Intel in August. Intel was previously in line to receive $8.9 billion in grants and loans from the CHIPS Act, but the Trump administration asked and received an equity stake in the chipmaker in exchange for the money.
Huang was with Trump this week in England to attend a State Dinner at Windsor Palace and announce new projects and investments in the U.K. But the Trump administration wasn’t involved in this deal, according to a White House official and Huang.
“Intel’s new partnership with Nvidia is a major milestone for American high-tech manufacturing,” White House spokesman Kush Desai said in a statement.