Tencent Secures Oracle AI Chips in $7B Deal
The latest semiconductor and artificial intelligence developments highlight expanding data center demands, major hardware deals, and evolving chip design partnerships.
Radwan AbdulhadiEditor and technology writer
Tencent's Offshore AI Chip Lease With Oracle
Reports indicate that Tencent has reached an agreement to lease high-performance AI chips from Oracle in a deal valued at $7 billion over five years. While the headline figure is substantial, estimated lease prices run about 43% below standard Nvidia H100 rental rates on one-year contracts, suggesting favorable terms for Tencent. Further details are available from Tom's Hardware in the original source material.
Although specific chip models were not explicitly named, the scale and pricing metrics point toward Hopper-class hardware. Analysts are examining how such rental mechanisms might enable Chinese firms to navigate export-controlled hardware access.
OpenAI and Synopsys Partner for Autonomous Chip Design
Synopsys previously unveiled a series of strategic partnerships, most notably a collaboration with OpenAI to introduce 'GPT-Synopsys'. This specialized AI model integrates Synopsys' domain IP knowledge and electronic design automation (EDA) tools with OpenAI's large language model capabilities.
The specialized AI model is expected to assist in designing successors to OpenAI's custom hardware, such as OpenAI's Jalapeño AI ASIC. By teaming up directly, AI labs and IP holders can sidestep typical intellectual property hurdles associated with AI-generated chip designs.
Qualcomm and Huawei Patent Licensing Agreement
Qualcomm has reportedly entered a patent licensing agreement with Huawei for the first time, marking a notable shift in historical licensing arrangements where Huawei previously licensed Qualcomm technology. Initial details remain sparse and subject to regulatory approval.
Reports suggest the arrangement includes Huawei's LogicFolding chip-stacking techniques across networking, computing, and artificial intelligence products. The LogicFolding method helps target faster data transmission as an alternative to missing advanced EUV lithography equipment.
Neon Recycling and Manufacturing Pressures
With crude neon production heavily concentrated in Ukraine, ongoing regional conflicts have caused global supply disruptions for DUV lithography manufacturing. DUV machines consume over 70 percent of the world's neon supply.
To mitigate supply risks, Japanese lithography toolmaker Gigaphoton alongside rival Kanten Techno have introduced neon recycling systems. Gigaphoton recently debuted a system claiming a 50 percent recovery rate to support chipmakers reliant on older lithography processes.
Groq Lawsuit and Nvidia Partnership Context
Two former Groq employees and shareholders filed a proposed class action lawsuit against the company's former board and CEO. The plaintiffs allege that leadership breached fiduciary duties during a $20 billion deal with Nvidia in December 2025, which cost other stockholders billions.
While Nvidia is not a defendant in the lawsuit, a vast majority of Groq's engineering staff transitioned directly to Nvidia following the arrangement. Groq's language processing units have since been integrated into Nvidia's Vera Rubin hardware ecosystem.
Mistral Releases Large 4 Open-Weight AI Model
European artificial intelligence firm Mistral has released its new trillion-parameter Large 4 model, also known as ML4 or Le Chonk. Independent testing from Artificial Analysis scored the model at 38 on its overall intelligence index.
Although trailing several U.S. and Chinese open-weight alternatives in general intelligence benchmarks, ML4 demonstrates strong capabilities in cybersecurity tasks. Mistral expects to release the exact model weights before the end of October.
Sources
- Tom's HardwareThis week on Tom's Hardware Premium: October 10, 2026
Continue chronologically
Related entity coverage
Related stories
Tom's Hardware Hosts AI Chip Design Week
Tom's Hardware Premium presents AI Chip Design Week featuring an OpenAI interview, chipmaking insights, and agent safety updates.
Mistral Large 4 Trails Some Chinese Open Models in Tests
Independent benchmarking ranks Mistral's new trillion-parameter Large 4 model as a top non-US and non-China AI model, though it trails certain Chinese open alternatives.
OpenAI and Oracle Expand Partnership to Bring Apps to ChatGPT
OpenAI and Oracle expand their commercial partnership to integrate business applications and AI agents into ChatGPT and power new Oracle offerings.
Oracle turns days of work into minutes with AI
Oracle says ChatGPT Work and Codex are helping recruiters, business users and engineers turn specialist work into repeatable workflows.
