Daily: Google’s Waymo develops custom chip

Share Chip Briefing

1.1

Reuters (08/20): Nvidia denies report it is rolling out China AI chip by year-end

Nvidia on Thursday denied a report by The Information that it plans to begin shipping a language processing unit (LPU) ​tailored for Chinese customers by year-end, saying it has no China-specific ‌version of the product on its roadmap.

Earlier on Thursday, The Information reported, citing two Nvidia employees, that ​the company planned to begin small-batch shipments of an artificial intelligence chip designed for Chinese customers by the end of the year and that several customers had already placed orders.

According ​to the report, the chip was a version of Nvidia’s LPU, ​a processor developed using technology licensed from startup Groq that works alongside graphics processing ‌units (GPUs) ⁠to speed up AI chatbot responses.

2.1

Reuters (08/20): Samsung Electroncis to announce more than $72 billion shareholder return programme, media reports

Samsung Electronics is set to announce a new shareholder return policy ​later this month worth more than ‌100 trillion won ($71.75 billion), media reported on Thursday, as the company plans to share ​record profits with shareholders amid an ​AI-driven chip supercycle.

Samsung will spend 50% of ​its free cash flow on the new programme, the ‌report ⁠said.

Local rival SK Hynix on Wednesday unveiled a 40 trillion won share buyback and cancellation plan, the largest shareholder return ​programme announced ​by a ⁠publicly listed South Korean company.

2.2

Reuters (08/20): SK Hynix workers to get 60% of this year’s bonuses in stock rather than all cash

Chipmaker SK Hynix said on Thursday it has reached a tentative wage deal with its South Korean workers ​that will see them receive at least 60% of this year’s bonuses in ‌company stock rather than the all-cash payout they received last year.

The AI boom has sent SK Hynix’s earnings soaring, leading to eye-popping bonuses for its employees. Workers are set to be paid an average 779 million ​won ($547,000) for 2026, according to Reuters calculations.

3.1

Bloomberg (08/21): Alphabet’s Waymo Has Built a Custom Chip for Its Robotaxis

Alphabet Inc.’s Waymo has built a custom chip that the company said will improve the performance of its robotaxis, while also diversifying its chip supply beyond third-party companies such as Nvidia Corp.

The ASIC chip — short for application-specific integrated circuit — is in production in Waymo’s latest robotaxi generation, the company said Thursday. Waymo said using its own chip, alongside others and with sensors the company designed itself, will improve its robotaxis’ reflexes and navigational skills in complex urban environments.

3.2

Bloomberg (08/20): AI Startup Callosum Raises $100 Million to Match Tasks to Chips

Callosum, a startup that makes software to match specific artificial intelligence tasks with different models and chips, has raised $100 million in early financing from backers including the UK’s public AI fund.

Venture capital firm Atomico led the seed financing, which includes participation from investors Plural and DCVC, as well as a “significant” investment from the UK’s £500 million ($677 million) Sovereign AI fund, the London-based startup said on Thursday. Callosum didn’t share its valuation.

3.3

Reuters (08/20): Micron unveils $10 billion AI memory research lab in Boise

Micron said on Thursday it plans to spend $10 billion on its newly established research ​lab at Boise, Idaho, over the next decade ‌to advance memory technologies, develop compute systems and support future chip manufacturing.

Large memory chipmakers such as Micron, Samsung Electronics and SK Hynix have been benefiting from a ​surge in demand from the rapid expansion ⁠of AI infrastructure.

3.4

Chosun Ilbo (08/20): SK Hynix Solves AI Bottlenecks with Light-Based Communication

SK Hynix has proposed a roadmap for next-generation optical communication technology to resolve data bottlenecks in AI servers using “light.” The plan involves replacing existing copper wiring connecting chips and servers with optical communication to enhance the performance and power efficiency of AI data centers where thousands of graphics processing units (GPUs) and high-bandwidth memory (HBM) are interconnected.

SK Hynix announced on the 20th that it published a paper in the international academic journal *Nature Electronics* outlining a technology roadmap for co-packaged optics (CPO) for AI and high-performance computing (HPC), co-authored by global researchers. Hong Seung-hoon, head of SK Hynix’s AI Infrastructure Team, and Professor Lee Kyu-sang of the University of Virginia served as corresponding authors, with researchers from the University of Illinois Urbana-Champaign (UIUC), Nanyang Technological University, the Massachusetts Institute of Technology (MIT), and Yonsei University participating in the joint study.