SpaceX Eyes Title of World's Biggest Datacenter Builder
SpaceX has raised its projection to roughly 13GW of capacity in 2027. That would be the largest amount of datacenter capacity ever built by a single company. Source: newsletter.semianalysis.com
AI chips, GPUs, data centers, cloud compute, and the hardware powering the AI revolution.
870 articles
SpaceX has raised its projection to roughly 13GW of capacity in 2027. That would be the largest amount of datacenter capacity ever built by a single company. Source: newsletter.semianalysis.com
Buildout plans for 2027 are largely locked in, but 2028 is far less certain. A prolonged permit freeze in Texas is now seen as the single biggest risk to the US buildout, and both parties are in play. Source: newsletter.semianalysis.com
UC San Diego has launched a research initiative, backed by Qualcomm Technologies, that uses AI to tackle the challenges of chip design. As AI reshapes industries worldwide, researchers are now turning the technology on itself, applying it to the very hardware that powers it. Source: today.ucsd.edu
A Monday filing shows Anthropic expects to owe nearly $70 billion in contractual obligations by September 2027. Total commitments come to roughly $447.5 billion, and the compute bill could climb toward $1 trillion. Source: theregister.com
Wall Street lenders are joining Blackstone in an unprecedented $63bn debt package. The financing will pay for Broadcom AI hardware to expand Anthropic's computing capacity. Source: businesschief.com
Marvell Technology is acquiring optical components maker Lumentum in an all-stock deal valued at $115 billion, including debt. It ranks among the largest deals in the tech industry. Source: reuters.com
South Korea will double its spending on AI and semiconductors in 2027 to a record high, the president said on Monday. He pledged to use the technology to transform the country. Source: finance.yahoo.com
OpenAI plans to purchase AI chips from Google as the ChatGPT developer expands its computing capacity. The move reduces its reliance on its established hardware suppliers. Source: techradar.com
Google says it has successfully run an AI model on a data-center-class processor on board a satellite. The test is described as an early milestone toward its ambitions for orbital data centers. Source: spacenews.com
Bank of America, Citi, Morgan Stanley and Blackstone are funding a $63 billion special purpose vehicle that buys Broadcom AI chips and leases them to Anthropic. The structure is backed by support from Broadcom. Source: globaldatacenterhub.com
Researchers at the Allen Telescope Array have adopted NVIDIA's Holoscan platform in their search for signals of possible extraterrestrial origin. The catch: the signals they detect might just as well be coming from the AI systems they built themselves. Source: blogs.nvidia.com
Gemma 4 ran on a data center TPU orbiting over Svalbard, while Alphabet's capex reached $132bn over nine months. Heat and radiation tests are next, raising the question of whether orbit can host AI compute at scale. Source: ppc.land
Google marked the arrival of a poem sent from Suncatcher, its first experimental AI datacenter satellite in orbit. The Friday announcement celebrated the milestone, though the verse proved an extremely expensive way to say hello from space. Source: theregister.com
With production hardware beginning to ship, NVIDIA has adjusted several specifications of its Rubin GPU platforms. Source: servethehome.com
U.S. and Chinese tech giants are investing billions in massive data centers in Malaysia. The buildout is straining local resources and testing the limits of Washington's influence. Source: nytimes.com
For decades, semiconductor companies competed on the strength of the individual chip. One of the biggest AI semiconductor trends now under way is a move toward competing at the system level. Source: siliconangle.com
NVIDIA's TensorRT Edge-LLM aims to simplify getting started with AI development for edge applications. Developers can launch and run models with only a handful of commands. Source: developer.nvidia.com
AI is changing how organizations build applications, handle information and automate decisions. Making it safe and secure depends on the underlying infrastructure, spanning compute, networking, storage and firmware. Source: blogs.nvidia.com
AM Intelligence, an Indian AI infrastructure company, is looking to raise as much as $3 billion in financing. The money would pay for a previously announced order of Nvidia hardware. Source: bloomberg.com
Nvidia is negotiating billions of dollars in credit guarantees for debt held by Nscale, a young AI cloud provider. The financing would support the development of a 460-megawatt data center tied to Anthropic. Source: theinformation.com