China's LineShine stays on top of Top500 as new entries hit a record low
Only 27 systems joined the November list, and the entry bar rose to 2.857 PFLOPS as heavy AI spending draws hardware into data centers.
A building full of servers, storage and networking equipment that runs online services and, increasingly, AI models.
AI data centers draw so much electricity and water that power supply and local permits have become central to where and how fast they are built.
Only 27 systems joined the November list, and the entry bar rose to 2.857 PFLOPS as heavy AI spending draws hardware into data centers.
Offshore providers serving rival nations must now identify end users, and shipments to five Southeast Asian and Latin American countries need licences.
Nvidia's board is not reinstating Huang, and new licensing rules now cover advanced chip shipments to five countries.
Gas and coal supplied more than half of their power, and the study's authors say the figures may already be out of date.
Operators in Malaysia and five other countries may soon need export licences for AI chips, with existing facilities given 90 days to comply.
The letter lands ahead of a Trump–Xi meeting, as a report says the department may fast-track export licences for major US AI companies.
Megaspeed, the largest Nvidia chip buyer in Southeast Asia, has opened an internal investigation while Jensen Huang agrees to testify before House committees.
The company has since made backups default for all virtual machines and is opening a new availability zone with fewer services at first.
The suit alleges concealment and fraud over a planned data center expected to cost between $20 billion and $30 billion, and says nearby home offers have dropped or been withdrawn.
The write-off exposes how fragile gas-turbine supply planning has become for companies building AI infrastructure.
The result rules out one route quantum machines could take on some hard problems, though verification is still informal and not all routes are closed.
The expanded collaboration pairs national AI infrastructure with inference software, agent security tools and faster semiconductor simulation.
Randomising when these devices check their charge could remove the pattern, but many IoT devices are rarely patched.
Su expects data center revenue to keep climbing while overall PC sales fall by double digits in 2027.
Any move into memory would be years away and would not quickly ease the AI-driven DRAM shortage.
The round is not closed yet, and the sum collected to date falls short of the $30 billion target reported earlier.
Investors may now favour firms with stronger balance sheets as political scrutiny of AI infrastructure grows.
If NVIDIA's own and partner-reported figures hold up, agent workloads could run on fewer CPUs per unit of power, though none of the numbers have been independently benchmarked.
The company says memory supply is stable enough to avoid further increases, even as a cartridge supplier shifts capacity toward AI data-center chips.
Investors are pricing in tighter political scrutiny of AI and data centre build-outs, on top of doubts about whether heavy AI spending will pay off.