Inside a sprawling lab at Google headquarters in Mountain View, California, hundreds of server racks hum across several aisles, performing tasks far less ubiquitous than running the world’s dominant search engine or executing workloads for Google Cloud’s millions of customers.
Instead, they’re running tests on Google’s own microchips, called Tensor Processing Units, or TPUs.
Originally trained for internal workloads, Google’s TPUs have been available to cloud customers since 2018. In July, Apple revealed it uses TPUs to train AI models underpinning Apple Intelligence. Google also relies on TPUs to train and run its Gemini chatbot.
“The world sort of has this fundamental belief that all AI, large language models, are being trained on Nvidia, and of course Nvidia has the lion’s share of training volume. But Google took its own path here,” said Futurum Group CEO Daniel Newman. He’s been covering Google’s custom cloud chips since they launched in 2015.
Google was the first cloud provider to make custom AI chips. Three years later, Amazon Web Services announced its first cloud AI chip, Inferentia. Microsoft‘s first custom AI chip, Maia, wasn’t announced until the end of 2023.
But being first in AI chips hasn’t translated to a top spot in the overall rat race of generative AI. Google’s faced criticism for botched product releases, and Gemini came out more than a year after OpenAI’s ChatGPT.
Google Cloud, however, has gained momentum due in part to AI offerings. Google parent company Alphabet reported cloud revenue rose 29% in the most recent quarter, surpassing $10 billion in quarterly revenues for the first time.
“The AI cloud era has completely reordered the way companies are seen, and this silicon differentiation, the TPU itself, may be one of the biggest reasons that Google went from the third cloud to being seen truly on parity, and in some eyes, maybe even ahead of the other two clouds for its AI prowess,” Newman said.
‘A simple but powerful thought experiment’
In July, CNBC got the first on-camera tour of Google’s chip lab and sat down with the head of custom cloud chips, Amin Vahdat. He was already at Google when it first toyed with the idea of making chips in 2014.
Amin Vahdat, VP of Machine Learning, Systems and Cloud AI at Google, holds up TPU Version 4 at Google headquarters in Mountain View, California, on July 23, 2024.
Marc Ganley
“It all started with a simple but powerful thought experiment,” Vahdat said. “A number of leads at the company asked the question: What would happen if Google users wanted to interact with Google via voice for just 30 seconds a day? And how much compute power would we need to support our users?”
“We realized that we could build custom hardware, not general purpose hardware, but custom hardware — Tensor Processing Units in this case — to support that much, much more efficiently. In fact, a factor of 100 more efficiently than it would have been otherwise,” Vahdat said.
Google data centers still rely on general-purpose central processing units, or CPUs, and Nvidia’s graphics processing units, or GPUs. Google’s TPUs are a different type of chip called an application-specific integrated circuit, or ASIC, which are custom-built for specific purposes. The TPU is focused on AI. Google makes another ASIC focused on video called a Video Coding Unit.
The TPU, however, is what set Google apart. It was the first of its kind when it launched in 2015. Google TPUs still dominate among custom cloud AI accelerators, with 58% of the market share, according to The Futurum Group.
Google coined the term based on the algebraic term “tensor,” referring to the large-scale matrix multiplications that happen rapidly for advanced AI applications.
With the second TPU release in 2018, Google expanded the focus from inference to training and made them available for its cloud customers to run workloads, alongside market-leading chips such as Nvidia’s GPUs.
“If you’re using GPUs, they’re more programmable, they’re more flexible. But they’ve been in tight supply,” said Stacy Rasgon, senior analyst covering semiconductors at Bernstein Research.
The AI boom has sent Nvidia’s stock through the roof, catapulting the chipmaker to a $3 trillion market cap in June, surpassing Alphabet and jockeying with Apple and Microsoft for position as the world’s most valuable public company.
“Being candid, these specialty AI accelerators aren’t nearly as flexible or as powerful as Nvidia’s platform, and that is what the market is also waiting to see: Can anyone play in that space?” Newman said.
Now that we know Apple’s using Google’s TPUs to train its AI models, the real test will come as those full AI features roll out on iPhones and Macs next year.
Broadcom and TSMC
It’s no small feat to develop alternatives to Nvidia’s AI engines. Google’s sixth generation TPU, called Trillium, is set to come out later this year.
Google showed CNBC the sixth version of its TPU, Trillium, in Mountain View, California, on July 23, 2024. Trillium is set to come out later in 2024.
Marc Ganley
“It’s expensive. You need a lot of scale,” Rasgon said. “And so it’s not something that everybody can do. But these hyperscalers, they’ve got the scale and the money and the resources to go down that path.”
The process is so complex and costly that even the hyperscalers can’t do it alone. Since the first TPU, Google’s partnered with Broadcom, a chip developer that also helps Meta design its AI chips. Broadcom says it’s spent more than $3 billion to make these partnerships happen.
“AI chips — they’re very complex. There’s lots of things on there. So Google brings the compute,” Rasgon said. “Broadcom does all the peripheral stuff. They do the I/O and the SerDes, all of the different pieces that go around that compute. They also do the packaging.”
Then the final design is sent off for manufacturing at a fabrication plant, or fab — primarily those owned by the world’s largest chipmaker, Taiwan Semiconductor Manufacturing Company, which makes 92% of the world’s most advanced semiconductors.
When asked if Google has any safeguards in place should the worst happen in the geopolitical sphere between China and Taiwan, Vahdat said, “It’s certainly something that we prepare for and we think about as well, but we’re hopeful that actually it’s not something that we’re going to have to trigger.”
Protecting against those risks is the primary reason the White House is handing out $52 billion in CHIPS Act funding to companies building fabs in the U.S. — with the biggest portions going to Intel, TSMC, and Samsung to date.
Processors and power
Google showed CNBC its new Axion CPU,
Marc Ganley
“Now we’re able to bring in that last piece of the puzzle, the CPU,” Vahdat said. “And so a lot of our internal services, whether it’s BigQuery, whether it’s Spanner, YouTube advertising and more are running on Axion.”
Google is late to the CPU game. Amazon launched its Graviton processor in 2018. Alibaba launched its server chip in 2021. Microsoft announced its CPU in November.
When asked why Google didn’t make a CPU sooner, Vahdat said, “Our focus has been on where we can deliver the most value for our customers, and there it has been starting with the TPU, our video coding units, our networking. We really thought that the time was now.”
All these processors from non-chipmakers, including Google’s, are made possible by Arm chip architecture — a more customizable, power-efficient alternative that’s gaining traction over the traditional x86 model from Intel and AMD. Power efficiency is crucial because, by 2027, AI servers are projected to use up as much power every year as a country like Argentina. Google’s latest environmental report showed emissions rose nearly 50% from 2019 to 2023 partly due to data center growth for powering AI.
“Without having the efficiency of these chips, the numbers could have wound up in a very different place,” Vahdat said. “We remain committed to actually driving these numbers in terms of carbon emissions from our infrastructure, 24/7, driving it toward zero.”
It takes a massive amount of water to cool the servers that train and run AI. That’s why Google’s third-generation TPU started using direct-to-chip cooling, which uses far less water. That’s also how Nvidia’s cooling its latest Blackwell GPUs.
Despite challenges, from geopolitics to power and water, Google is committed to its generative AI tools and making its own chips.
“I’ve never seen anything like this and no sign of it slowing down quite yet,” Vahdat said. “And hardware is going to play a really important part there.”
The U.S. has placed major chip export restrictions on Huawei and Chinese firms over the past few years. This has cut off companies’ access to critical semiconductors.
Jaap Arriens | Nurphoto | Getty Images
Taiwan has added China’s Huawei and SMIC to its trade blacklist in a move that further aligns it with U.S. trade policy and comes amid growing tensions with Beijing.
Taiwan’s current regulations require licenses from regulators before domestic firms can ship products to parties named on the entity list.
In a statement on its website, Taiwan’s International Trade Administration said that Huawei and SMIC were among the 601 new foreign entities, blacklisted due to their involvement in arms proliferation activities and other national security concerns.
Huawei and SMIC are also on a U.S. trade blacklist and have been impacted by Washington’s sweeping controls on advanced chips. Companies such as contract chipmaker Taiwan Semiconductor Manufacturing Co already follow U.S. export restrictions.
However, the addition of Huawei and SMIC to the Taiwan blacklist is likely aimed at the reinforcement of this policy and a tightening of existing loopholes, Ray Wang, an independent semiconductor and tech analyst, told CNBC.
He added that the new domestic export controls could also raise the punishment for any potential breaches in the future.
TSMC had been embroiled in controversy in October last year when semiconductor research firm TechInsights found a TSMC-made chip in a Huawei AI training card.
Following the discovery, the U.S. Commerce Department ordered TSMC to halt Chinese clients’ access to chips used for AI services, according to a report from Reuters. TSMC could also reportedly face a $1 billion as penalty to settle a U.S. investigation into the matter.
Huawei has been working to create viable alternatives to Nvidia‘s general processing units used for AI. But, experts say the company’s advancement has been limited by export controls and a lack of scale and capabilities in the domestic chip ecosystem.
Still, Huawei is believed to have acquired several million GPU dies from TSMC for its AI chips by using previous loopholes before they were discovered, according to Paul Triolo, partner and senior vice president for China at advisory firm DGA-Albright Stonebridge Group.
A die refers to a small piece of silicon material that serves as the foundation for building processors and contains the intricate circuitry and components necessary to perform computations.
The Taiwanese government’s crackdown on exports to SMIC and Huawei also comes amid tense geopolitical tensions with Mainland China, which regards the democratically governed island as its own territory to be reunited by force, if necessary.
In statements reported by state media on Sunday, China’s top political adviser Wang Huning echoed Beijing’s position, calling for the promotion of national reunification with Taiwan and for resolute opposition to Taiwan independence.
An AI assistant on display at Mobile World Congress 2024 in Barcelona.
Angel Garcia | Bloomberg | Getty Images
Artificial intelligence is shaking up the advertising business and “unnerving” investors, one industry leader told CNBC.
“I think this AI disruption … unnerving investors in every industry, and it’s totally disrupting our business,” Mark Read, the outgoing CEO of British advertising group WPP, told CNBC’s Karen Tso on Tuesday.
The advertising market is under threat from emerging generative AI tools that can be used to materialize pieces of content at rapid pace. The past couple of years has seen the rise of a number of AI image generators, including OpenAI’s DALL-E, Google’s Veo and Midjourney.
In his first interview since announcing he would step down as WPP boss, Read said that AI is “going to totally revolutionize our business.”
“AI is going to make all the world’s expertise available to everybody at extremely low cost,” he said at London Tech Week. “The best lawyer, the best psychologist, the best radiologist, the best accountant, and indeed, the best advertising creatives and marketing people often will be an AI, you know, will be driven by AI.”
Read said that 50,000 WPP employees now use WPP Open, the company’s own AI-powered marketing platform.
“That, I think, is my legacy in many ways,” he added.
Structural pressure on creative parts of the ad business are driving industry consolidation, Read also noted, adding that companies would need to “embrace” the way in which AI would impact everything from creating briefs and media plans to optimizing campaigns.
A report from Forrester released in June last year showed that more than 60% of U.S. ad agencies are already making use of generative AI, with a further 31% saying they’re exploring use cases for the technology.
‘Huge transformation’
Read is not alone in this view. Advertising is undergoing a “huge transformation” due to the disruptive effects of AI, French advertising giant Publicis Groupe’s CEO Maurice Levy told CNBC at the Viva Tech conference in Paris.
He noted that AI image and video generation tools are speeding up content production drastically, while automated messaging systems can now achieve “personalization at scale like never before.”
Read more CNBC tech news
However, the Publicis chief stressed that AI should only be considered a tool that people can use to augment their lives.
“We should not believe that AI is more than a tool,” he added.
And while AI is likely to impact some jobs, Levy ultimately thinks it will create more roles than it destroys.
“Will AI replace me, and will AI kill some jobs? I think that AI, yes, will destroy some jobs,” Levy conceded. However, he added that, “more importantly, AI will transform jobs and will create more jobs. So the net balance will be probably positive.”
This, he says, would be in keeping with the labor impacts of previous technological inventions like the internet and smartphones.
“There will be more autonomous work,” Levy added.
Still, Nicole Denman Greene, analyst at Gartner, warns brands should be wary of causing a negative reaction from consumers who are skeptical of AI’s impact on human creativity.
According to a Gartner survey from September, 82% of consumers said firms using generative AI should prioritize preserving human jobs, even if it means lower profits.
“Pivot from what AI can do to what it should do in advertising,” Greene told CNBC.
“What it should do is help create groundbreaking insights, unique execution to reach diverse and niche audiences, push boundaries on what ‘marketing’ is and deliver more brand differentiated, helpful and relevant personalized experiences, including deliver on the promise of hyper-personalization.”
Jensen Huang, co-founder and chief executive officer of Nvidia Corp., left, and Emmanuel Macron, France’s president at the 2025 VivaTech conference in Paris, France, on Wednesday, June 11, 2025.
Nathan Laine | Bloomberg | Getty Images
Nvidia boss Jensen Huang has been on a tour of Europe this week, bringing excitement and intrigue to everywhere he visited.
His message was clear — Nvidia is the company that can help Europe build its artificial intelligence infrastructure so the region can take control of its own destiny with the transformative technology.
I’ve been in London and Paris this week following Huang around as he met with U.K. Prime Minister Keir Starmer, French President Emmanuel Macron, journalists, fans, analysts and gave a keynote at Nvidia’s GTC event in the capital of France.
Here’s the what I saw and the key things I learned.
At London Tech Week, the lines were long and the auditorium packed to hear him speak.
The GTC event in Paris was full too. It was like going to a music concert or sporting event. There were GTC Paris T-shirts on the back of every chair and even a merchandise store.
Nvidia GTC in Paris on 11 June 2025
Arjun Kharpal
The aura of Huang really struck me when, after a question-and-answer session with him and a room full of attendees, most people lined up to take pictures or selfies with him.
Macron and Starmer both wanted to be seen on stage with him.
Nvidia positions itself as Europe’s AI hope
Nvidia’s key product is its graphics processing units (GPU) that are used to train and execute AI applications.
But Huang has positioned Nvidia as more than a chip company. During the week, he described Nvidia as an infrastructure firm. He also said AI should be seen as infrastructure like electricity.
His pitch to all countries was that Nvidia could be the company that will help countries build out that infrastructure.
“We believe that in order to compete, in order to build a meaningful ecosystem, Europe needs to come together and build capacity that is joint,” Huang said during a speech at the Viva Tech conference in Paris on Wednesday.
Jensen Huang, CEO of Nvidia, speaks during the Viva Technology conference dedicated to innovation and startups at Porte de Versailles exhibition center in Paris, France, June 11, 2025.
Gonzalo Fuentes | Reuters
One of the most significant partnerships announced this week is between French startup Mistral and Nvidia to build a so-called AI cloud using the latter’s GPUs.
Huang spoke a lot during the week about “sovereign AI” — the concept of building data centers within a country’s borders that services its population rather than relying on servers located overseas. Among European policymakers and companies, this has been an important topic.
Huang also heaped praise on the U.K., France and Europe more broadly when it came to their potential in the AI industry.
China still behind but catching up
On Thursday, Huang decided to do a tour of Nvidia’s booth and I managed to catch him to get a few words on CNBC’s “Squawk Box Europe.”
A key topic of that discussion was China. Nvidia has not been able to sell its most advanced chips to China because of U.S. export controls and even less sophisticated semiconductors are being blocked. In its last quarterly results, Nvidia took a $4.5 billion hit on unsold inventory.
I asked Huang about how China was progressing with AI chips, in particular referencing Huawei, the Chinese tech giant that is trying to make semiconductor products to rival Nvidia.
Huang said Huawei is a generation behind Nvidia. But because there is lots of energy in China, Huawei can just use more chips to get results.
“If the United States doesn’t want to partake, participate in China, Huawei has got China covered, and Huawei has got everybody else covered,” Huang said.
In addition, Huang is concerned about the strategic importance of U.S. companies not having access to China.
“It’s even more important that the American technology stack is what AI developers around the world build on,” Huang said.
Just reading between the lines somewhat — Huang sees a world where Chinese AI tech advances. Some countries may decide to build their AI infrastructure with Chinese companies rather than American. That in turn could give Chinese companies a chance to be in the AI race.
Quantum, robotics and driverless is the future
Huang often uses public appearances to talk about the future.
I asked him about some of those areas he’s bullish on like robotics and driverless cars, technology that Nvidia’s products can power.
Huang told me this will be the “decade of” autonomous vehicles and robotics.
Nvidia boss Jensen Huang delivers a speech on stage talking about robotics.
Arjun Kharpal | CNBC
During his keynote at GTC Paris on Wednesday, he also address quantum computing, saying the technology is reaching “an inflection point.”
Quantum computers are widely believed to be able to solve complex problems that classic computers can’t. This could include things like discovering new drugs or materials.