OpenAI is planning to purchase Google's tensor processing units, according to reports from The Information. The deal, if it materializes, would mark a significant shift for OpenAI, which has long relied on Nvidia as its primary chip supplier. Google sees the potential agreement as an opportunity to expand its chip business, and its TPUs are increasingly viewed as a more affordable alternative to Nvidia's graphics processing units. The final terms of the deal have not been settled, and the financial details remain unknown.
The potential purchase deepens the relationship between OpenAI and Google, even though the two companies compete in AI products. OpenAI also unveiled a custom AI chip called Jalapeno, designed in partnership with Broadcom, as it works to diversify its hardware strategy. Meanwhile, Amazon, Microsoft, and Alphabet are collectively expected to spend up to $600 billion this year on AI infrastructure, underscoring the massive scale of investment in the sector.
In related chip news, Canon cut its full-year financial outlook due to shortages of printed circuit boards and other components driven by surging AI hardware demand. The company said the supply constraints will reduce medical sales by 90 billion yen this year. Shares in Canon fell more than 5% in Japan following the announcement, as operating expenses climbed to 1.339 trillion yen in the first nine months.
Opera has begun rolling out a local AI browser feature that runs AI models directly on users' devices. The feature supports 64-bit Windows PCs with Intel or AMD processors and Macs with Apple silicon, though Intel-based Macs, Linux, and Windows on Arm are excluded. Users must opt in to enable WebAI and download supported models, which include Google's Gemma and Apple's on-device foundation model. Website access to local models requires separate permissions from users.
Google also previewed a new crypto custody security system that uses Multi-Party Computation technology to protect digital assets. The system splits private keys across isolated hardware and cloud providers to eliminate single points of failure, specifically targeting AI-driven cyber threats. Alphabet is already using the framework to protect its own assets on Ethereum and Solana blockchains, and major financial institutions are testing it. A commercial release is planned for early 2027.
Japan's NTT Data is building a data center campus in Johor with a planned 290 MW power capacity, with the first building set to open in April 2027 using direct liquid cooling technology. Separately, a semiconductor talent program launched in October 2025 has enrolled 106 and 107 students in its first two cohorts. In other AI market data, MEXC reported that AI-related tokenized U.S. stock and ETF trading volume rose 88% in September, with semiconductors growing 100% and optical communications up 99%.
SpaceX was added to CNBC's Bullpen stocks watch list. The company, which went public in June at $135 per share, is building AI infrastructure in Memphis with its Colossus supercomputers and expects to add 2.5 to 3 gigawatts of compute by the end of 2026. Starlink, its connectivity business, has nearly 14 million consumer subscribers and is the most profitable segment. SpaceX estimates a total addressable market of $28.5 trillion across its operations.
A new AI system called LUMINA, developed by a PhD candidate, is designed to help plan lunar landing missions by evaluating potential sites, identifying hazards, and finding scientifically valuable features on the lunar surface. Built on a lunar vision-language model, it can plan multistep analyses and explain its recommendations. The team plans to test it on new lunar data, with NASA's Artemis program offering important opportunities for development.
A new system called the Decision Model Factory creates specialized decision models for different tasks and environments using a straightforward recipe to train, verify, and deploy models under operator control. Five models built through the factory improved results on Japanese decision tasks and passed a public benchmark. A new runner tool more than doubled performance on multi-question workloads, making it easier to build and deploy fast, task-specific decision models as needed.
Key Takeaways
- OpenAI plans to buy Google's tensor processing units, diversifying away from Nvidia as its main chip supplier
- Amazon, Microsoft, and Alphabet plan to spend up to $600 billion this year on AI infrastructure
- OpenAI unveiled a custom AI chip called Jalapeno designed with Broadcom
- Canon cut its full-year outlook due to AI hardware shortages, citing a 90 billion yen hit to medical sales
- Opera launched a local AI browser feature supporting Intel and AMD Windows PCs and Apple silicon Macs
- Google previewed a crypto custody system using Multi-Party Computation to combat AI-driven cyber threats, with commercial release planned for early 2027
- Japan's NTT Data is building a 290 MW data center campus in Johor, set to open in April 2027
- MEXC reported 88% growth in AI tokenized stock trading volume in September, with semiconductors up 100%
- SpaceX expects to add 2.5 to 3 gigawatts of compute by end of 2026 via its Colossus supercomputers in Memphis
- The LUMINA AI system can plan lunar landing missions and is being tested with potential NASA Artemis program applications
OpenAI plans to buy Google AI chips from TPU maker
OpenAI is planning to buy Google's AI chips called tensor processing units. This would be a big change because Nvidia has been OpenAI's main chip supplier for a long time. The deal is still in early stages and the financial details are not known. Google sees this as a chance to grow its chip business.
OpenAI plans to buy Google's AI chips, report says
OpenAI plans to buy Google's AI chips according to a report by The Information. The purchase would strengthen the relationship between the two companies even though they compete in AI products. OpenAI is one of the biggest buyers of Nvidia's graphics processing units. Google started selling its tensor processing units directly to some customers this year.
OpenAI plans to buy Google's AI chips, report says
OpenAI is planning to buy Google's AI chips in a deal that deepens ties between the two competing companies. Amazon, Microsoft and Alphabet plan to spend up to 600 billion dollars this year on AI infrastructure. OpenAI also unveiled a custom AI chip called Jalapeno designed with Broadcom. Google's tensor processing units are seen as a cheaper option to Nvidia's chips.
OpenAI plans to buy Google AI chips to expand capacity
OpenAI plans to buy AI chips from Google to expand its computing capacity and diversify its suppliers. The final terms of the deal have not been decided yet. Google's tensor processing units are emerging as an alternative to Nvidia's graphics processing units. OpenAI is also developing its own chips called Jalapeno with Broadcom.
OpenAI reportedly plans to purchase Google AI chips
OpenAI is planning to purchase AI chips from Google according to a report by The Information. The potential deal would expand the commercial relationship between the two companies. Last week OpenAI announced it would make its frontier AI models available through Google Cloud. Google's tensor processing units offer an alternative to Nvidia's graphics processing units.
OpenAI considers buying Google AI chips to challenge Nvidia
OpenAI plans to purchase Google's custom AI chips called TPUs. This move could reduce Nvidia's control over the AI hardware market. The shift may also boost Alphabet's chip business. Nvidia shares traded at USD 219.70 while Alphabet's Class A shares traded at USD 325.42 as of October 28, 2026.
Opera launches local AI browser for enterprise privacy
Opera began rolling out AI models that run directly in its browser. The feature supports 64-bit Windows PCs with Intel or AMD processors and Macs with Apple silicon. Users must opt in to enable WebAI and download models. Supported models include Google's Gemma and Apple's on-device foundation model.
Opera local AI preview limited by hardware and permissions
Opera One's local AI preview is an opt-in feature available on select Windows and Apple silicon Macs. The preview excludes Intel-based Macs, Linux, and Windows on Arm. Users enable WebAI in Settings under Opera AI, and website access to local models requires separate permissions.
Japan supports Southeast Asia AI and semiconductor development
Japan's NTT Data is building a data center campus in Johor with 290 MW planned power capacity. The first building is set to open in April 2027 using direct liquid cooling technology. A semiconductor talent program launched in October 2025 enrolled 106 and 107 students in its first two cohorts.
MEXC sees 88% growth in AI tokenized stock trading
MEXC reported that AI-related tokenized U.S. stock and ETF trading volume rose 88% in September. Four sectors grew over 90% including semiconductors at 100% and optical communications at 99%. AI-related assets made up 48% of tokenized spot trading volume. Futures volumes for Dell and AMD rose 243% and 209% respectively.
AI framework LUMINA helps plan lunar landing missions
A new AI system called LUMINA was created by a PhD candidate to help plan moon missions. It evaluates potential landing sites, identifies hazards, and finds scientifically valuable features on the lunar surface. Built on a lunar vision-language model, it can plan multistep analyses and explain its recommendations. The system is designed to support mission planners rather than work independently. The team plans to test it on new lunar data, with NASA's Artemis program offering important opportunities for development.
SpaceX added to CNBC Bullpen stocks list
CNBC's Investing Club added SpaceX to its Bullpen stocks to watch list. The company went public in June at $135 per share and operates in three segments: Space, Connectivity, and Super Intelligence. SpaceX is building AI infrastructure in Memphis with its Colossus supercomputers and expects to add 2.5 to 3 gigawatts of compute by end of 2026. Starlink, its connectivity business, has nearly 14 million consumer subscribers and is the most profitable segment. The company estimates a total addressable market of $28.5 trillion across its businesses.
Health care AI companies must contend with Epic
This article discusses why every health care AI company must face Epic in the market. It highlights the challenges AI developers face when competing against Epic's dominant position in health care technology. The piece is exclusive to STAT+ subscribers and focuses on the competitive landscape for AI innovation in health care.
Google previews crypto custody system to fight AI cyber threats
Google previewed a new crypto custody security system using Multi-Party Computation technology. It eliminates single points of failure by splitting private keys across isolated hardware and cloud providers. The system targets AI-driven cyberattacks and automated hacking techniques. Alphabet is already using the framework to protect its own digital assets on Ethereum and Solana blockchains. Major financial institutions are testing it, with a commercial release planned for early 2027.
Canon cuts outlook due to AI hardware shortages
Canon cut its full-year outlook because AI hardware demand caused shortages of printed circuit boards and components. The company said these constraints will reduce medical sales by 90 billion yen this year. Shares fell more than 5% in Japan following the announcement. Operating expenses rose to 1.339 trillion yen in the first nine months due to pay increases and currency costs. Canon said the supply hit from booming AI hardware demand is landing at a time when costs are already climbing.
New Factory Builds Fast Japanese Decision Models
The Decision Model Factory creates specialized decision models for different tasks and environments. It uses a simple recipe to train, check, and deploy these models under operator control. Five models built by this factory improved results on Japanese decision tasks and passed a public benchmark. A new runner tool speeds up serving by more than doubling performance on multi-question workloads. This system makes it easy to build and deploy fast, task-specific decision models whenever needed.
Sources
- OpenAI, a Longtime Nvidia Customer, Plans to Buy Google’s AI Chips
- OpenAI plans to buy Google’s AI chips, The Information reports
- OpenAI plans to buy Google's AI chips, The Information reports
- OpenAI plans to buy Google AI chips to expand computing capacity
- OpenAI Reportedly Plans to Purchase Google AI Chips as Demand for Computing Infrastructure Grows
- OpenAI plans to buy Google's AI chips, challenging Nvidia's dominance in AI hardware supply.
- Opera launches local browser AI, testing enterprise privacy and AI PC strategies
- Opera’s local-AI preview is opt-in, with hardware and site-permission limits
- Japan backs Southeast Asia’s AI and semiconductor growth through investment and training
- MEXC Reports 88% Growth in AI-Related Tokenized U.S. Stock and ETF Trading Volume in September, With Four Infrastructure Sectors Up Over 90%
- Could next-gen artificial intelligence help map out future space outposts?
- We're adding a new, high-profile AI play to our Bullpen stocks to watch list
- Why every health care AI company must face Epic
- Google Previews Institutional Crypto Custody Security System Amid Mounting AI Cybersecurity Threats
- Canon Cuts Outlook As AI Hardware Shortages Bite
- Decision Model Factory: Building Specialized Sovereign Decision Models
Comments
Please log in to post a comment.