Chairman of Foxconn Young Liu delivers a speech during the Hon Hai Tech Day in Taipei on Oct. 18, 2023.
I-hwa Cheng | AFP | Getty Images
The boom in corporate investment into artificial intelligence infrastructure still has some way to go as large language models are continuing to evolve, according to the CEO of Foxconn, a key supplier to Apple.
Speaking with CNBC’s Emily Tan, Chief Executive and Chairman of Foxconn Young Liu, said that the AI boom “still has some time to go” as advanced language models from the likes of OpenAI are becoming increasingly intelligent with each new iteration that comes out.
He said that the general movement in the tech industry today is trending toward a form of AI that is as intelligent — if not more intelligent — than humans. This type of AI is referred to in the industry as “AGI,” or Artificial General Intelligence.
“We … heard about AGI, and we’ll talk about different levels of intelligence. If you divide [intelligence] into four different levels, we’re at level two. There’s still level three and level four to go,” Liu told CNBC in an interview that aired Tuesday.
OpenAI is one of the leading companies pushing for AGI. Sam Altman, CEO of the Microsoft-backed startup, has previously said AGI will be developed in the “reasonably close-ish future,” however he’s also said he thinks it will “change jobs much less than we all think.”
The company, which released its upgraded GPT-4o model this summer, revealed last week that it had raised $6.6 billion at a $157 billion valuation. OpenAI is still working toward releasing its next-generation LLM, GPT-5, however it’s yet to share timing on when the new language model will launch.
Liu said that progress toward increasingly intelligent AI can only be a good thing for the AI server industry, which has been a key boon to Foxconn’s growth this year.
“I think for the AI server industry, I think we still have some time to grow,” Liu added. “With the AGI capability growing, the age [of] AI devices will be another industry we should watch carefully.”
Strong demand for Nvidia’s Blackwell chip
Foxconn, which trades as Hon Hai locally in China and Taiwan, is the world’s largest contract manufacturer for the electronics industry. The firm produces and assembles Apple products, including about two out of every three iPhones.
On Saturday, Foxconn said that it reported better-than-expected sales figures. The firm said its revenues came in at 1.85 trillion Taiwanese dollars ($57.5 billion) in the September quarter, up 20.2% year-over-year. That “exceeded the company’s original expectations of significant growth,” according to Foxconn.
The strong performance came off the back of heightened demand for AI servers, which Foxconn manufactures for several major global tech giants, including Amazon, Google, Microsoft, and Nvidia.
The company is currently on track to ship Nvidia’s next-generation Blackwell GPU (graphics processing unit), an advanced AI server chip, in the fourth quarter of 2024, Liu told CNBC. Blackwell is also known under the brand name GB200.
Asked about the order book for Blackwell, Liu said that demand for the chip is “much better than we thought,” adding that the firm is building new factories in Mexico to help service outsized demand for the product.
His comments tally with what Nvidia CEO Jensen Huang has said about demand for Blackwell previously. Last week, Huang told CNBC’s “Closing Bell Overtime” that demand for Blackwell was “insane.”
“Everybody wants to have the most and everybody wants to be first,” Huang said during the interview, which aired last Wednesday.
Blackwell, expected to cost between $30,000 and $40,000 per unit, is in hot demand from companies like OpenAI, Microsoft, Meta and other firms building AI data centers to power products like ChatGPT and Copilot.
AI devices the next growth opportunity
Beyond producing IT infrastructure for cloud-based AI applications like ChatGPT, Foxconn is also powering the new iPhone 16, which comes with the latest generative AI features Apple hopes will breathe new life into the smartphone industry.
Foxconn’s Liu said that “on-device” AI — where data is processed directly on a mobile device, rather than a cloud or server — represents the next substantial growth opportunity for the company.
“I think the genAI-related device will be the next way to grow,” Liu told CNBC. “Currently, we’re seeing this on the cloud side. You see that genAI cloud equipment was very much booming.”
“But in the next phase, what we’re seeing will be the Gen AI devices,” he added. “We think that will be the next big way to boom … We have very high hopes on those devices.”
Still, it’s worth noting though that Apple hasn’t yet released its AI system, called Apple Intelligence, on iPhone. The company is expected to release Apple Intelligence to the public in a beta version later this fall as part of a new software update.
Smartphone sales have been gaining momentum this year after several consecutive years of declines. In the second quarter of 2024, smartphone shipments climbed 6.5% year-over-year to 285.4 million units, according to preliminary data from IDC — their fourth straight quarter of growth.
Meta’s Mark Zuckerberg plans to visit South Korea, scheduling key meetings during the trip, according to a statement by Meta on Wednesday, which did not provide further details. Reportedly, Zuckerberg is anticipated to meet with Samsung Electronics chairman Jay Y. Lee later this month to discuss AI chip supply and other generative AI issues, as per the South Korean newspaper Seoul Economic Daily, citing unnamed sources familiar with the matter.
Alex Wong | Getty Images News | Getty Images
Meta extended its ban on new political ads on Facebook and Instagram past Election Day in the U.S.
The social media giant announced the political ads policy update on Monday, extending its ban on new political ads past Tuesday, the original end date for the restriction period.
Meta did not specify the day it will lift the restriction, saying only that the ad blocking will continue “until later this week.” The company did not say why it extended the political advertising restriction period.
The company announced in August that any political ads that ran at least once before Oct. 29 would still be allowed to run on Meta’s services in the final week before Election Day. Other political ads will not be allowed to run.
Organization with eligible ads will have “limited editing capabilities” while the restriction is still in place, Meta said. Those advertisers will be allowed to make scheduling, budgeting and bidding-related changes to their political ads, Meta said.
Meta enacted the same policy in 2020. The company said the policy is in place because “we recognize there may not be enough time to contest new claims made in ads.”
Google-parent Alphabet announced a similar ad policy update last month, saying it would pause ads relating to U.S. elections from running in the U.S. after the last polls close on Tuesday. Alphabet said it would notify advertisers when it lifts the pause.
Nearly $1 billion has been spent on political ads over the last week, with the bulk of the money spent on down-ballot races throughout the U.S., according to data from advertising analytics firm AdImpact.
Sam Altman, CEO of OpenAI, attends the 54th annual meeting of the World Economic Forum, in Davos, Switzerland, January 18, 2024 (L), and Amazon CEO Jeff Bezos speaks during the UN Climate Change Conference (COP26) in Glasgow, Scotland, Britain, November 2, 2021.
Reuters
Physical Intelligence, a robot startup based in San Francisco, has raised $400 million at a $2.4 billion post-money valuation, the company confirmed Monday to CNBC.
Investors included Amazon founder Jeff Bezos, OpenAI, Thrive Capital and Lux Capital, a Physical Intelligence spokesperson said. Khosla Ventures and Sequoia Capital are also listed as investors on the company’s website.
Physical Intelligence’s new valuation is about six times that of its March seed round, which reportedly came in at $70 million with a $400 million valuation. Its current roster of employees includes alumni of Tesla, Google DeepMind and X.
The startup focuses on “bringing general-purpose AI into the physical world,” per its website, and it aims to do this by developing large-scale artificial intelligence models and algorithms to power robots. The startup spent the past eight months developing a “general-purpose” AI model for robots, the company wrote in a blog post. Physical Intelligence hopes that model will be the first step toward its ultimate goal of developing artificial general intelligence. AGI is a term used to describe AI technology that equals or surpasses human intellect on a wide range of tasks.
Physical Intelligence’s vision is that one day users can “simply ask robots to perform any task they want, just like they can ask large language models (LLMs) and chatbot assistants,” the startup wrote in the blog post. In case studies, Physical Intelligence details how its tech could allow a robot to do laundry, bus tables or assemble a box.
To Barry Diller, a friend of Amazon founder Jeff Bezos, the decision for The Washington Post not to endorse a candidate in tomorrow’s presidential election was “absolutely principled” — and poorly timed, he said Monday on CNBC’s Squawk Box.
“They made a blunder — it should’ve happened months before, and it didn’t, and that’s the issue with it,” Diller said.
Diller is chairperson of both online travel company Expedia and IAC, which owns media platforms and websites like Dotdash Meredith and Care.com. He and Bezos appear to have been close friends for years, with Diller and his wife, fashion designer Diane von Furstenberg, hosting Bezos’s engagement party to fiancee Lauren Sanchez.
The decision not to endorse a presidential candidate in the 2024 race or for future presidential races came directly from Bezos, the paper’s owner, according to an article published by two of the Post’s own reporters.
The move prompted public condemnation from several staff writers, a flood of at least 250,000 digital subscription cancellations and the resignations of at least three editorial board members.
Bezos defended his position in his own op-ed late last month, calling the move a “meaningful step in the right direction” to restore low public trust in media and journalism.
“Presidential endorsements do nothing to tip the scales of an election,” Bezos wrote, emphasizing that the decision to not endorse a candidate was made “entirely internally” and without consulting either campaign. “I wish we had made the change earlier than we did, in a moment further from the election and the emotions around it.”
Diller said he spoke to Bezos following the decision.
“I think it was absolutely principled,” Diller said. “The mistake they made — and it was a mistake admitted by him — was timing.”