computer chips

Researchers report new building blocks that expand what photonic chips can do, broadening the range of functions these light-based devices can perform. Photonic chips, which use light instead of electrical signals, are of growing interest for faster and more energy-efficient computing and communications, and more versatile components could speed their adoption in real-world systems.

Amazon is looking to sell roughly $8 billion worth of Nvidia chips to outside investors, according to a Financial Times report picked up by Reuters. The move suggests the company has secured more AI computing hardware than it currently needs and is seeking to share the cost and capacity burden with others.

A new video essay asks a provocative question for the AI era: what would computing look like if the industry moved away from GPUs, the chips that currently power most machine learning. The piece explores alternatives to GPU-based hardware and the economics of relying so heavily on a small number of chip suppliers. It is drawing attention on Hacker News, where readers are debating whether the GPU's dominance in AI compute is as secure as it seems.

A researcher attended the final edition of the LABScon security conference and presented work on the geopolitics of computing power. The talk, titled "Chips on the Move: Export Controls, Evasion, and the Race for Compute," examined the link between strategic policy goals and the global competition for advanced semiconductors. The event's closing edition is prompting discussion among information security researchers, with chip export restrictions and compute access remaining hot policy topics.

SpaceX would need around $40 billion to buy chips from Nvidia, according to a report by BGR. The figure highlights the enormous cost of the processors required to power large-scale computing, likely tied to SpaceX's ambitions in areas such as satellite networks or AI-related infrastructure. The scale of the number has drawn attention to how central Nvidia's chips have become to major tech and aerospace projects.

Investors are being pointed toward AI exchange-traded funds that go past the usual semiconductor makers and hyperscale cloud companies. The argument is that the next stage of AI investing will include firms across software, data infrastructure, energy and other parts of the supply chain that benefit from artificial intelligence adoption, rather than concentrating on a handful of dominant chip and computing names.

SpaceX is reportedly in discussions to raise around $40 billion in loans to fund purchases of Nvidia chips, likely to power its computing and AI ambitions. The reported loan talks, covering Nvidia's in-demand processors, drew immediate attention given the scale of the financing for a private company.

Tensor Machines has launched an open-source benchmark designed to measure the true cost of AI compute, going beyond raw performance figures to capture real-world spending on hardware, energy and infrastructure. The company says the tool will help organizations compare AI workloads more honestly across different chips and providers, as compute costs remain a central concern for the AI industry.

Broadcom is reportedly arranging a $50 billion financing deal tied to its chip supply arrangement with OpenAI. The move would fund the massive compute buildout OpenAI needs, with Broadcom supplying custom AI chips and deepening one of the most consequential partnerships in the AI hardware race. Investors are watching what the deal means for Broadcom's revenue outlook and the broader AI spending boom.

Artificial intelligence companies are taking on hundreds of billions of dollars in debt to finance the purchase of advanced computer chips, with Wall Street banks and investors fueling the borrowing spree. The surge in AI-related financing reflects the enormous capital costs of building data centers and acquiring GPUs, and is raising questions about how sustainable the debt-fueled buildout will be.

Broadcom is putting together roughly $50 billion in financing so OpenAI can buy the company's custom AI chips. The arrangement would let OpenAI acquire large volumes of Broadcom-designed processors, deepening an already significant partnership in the AI semiconductor supply chain. The deal underscores the enormous capital demands of building AI computing infrastructure, as major technology companies race to secure the specialized chips needed to train and run large AI models.

Discussion is focusing on the intensifying competition between the United States and China over artificial intelligence, covering the two countries' rivalry in developing advanced AI models, computing power and talent. Commenters and analysts are weighing how US export controls on chips, Chinese model progress and national security concerns will shape which country leads the technology.

Broadcom is reportedly looking to raise more than $50 billion in financing to support the development of custom AI chips for OpenAI, while Oracle is also said to be pursuing major chip financing of its own. The moves underscore the enormous capital demands of the AI infrastructure race, with chipmakers and cloud providers racing to secure computing capacity for artificial intelligence workloads.