NVIDIA

NVIDIA's European AI programme now spans a Munich industrial AI cloud, UK capacity commitments and a USD 2bn investment in Nebius. The announcements describe computing that is promised; they do not yet describe facilities, power contracts and network paths that are operating.

Sharp is pairing Foxconn's hardware supply with local sales and support as it prepares to deliver AI servers to Japanese organisations from fiscal 2027.

Firmus is courting investors for a proposed Australian listing while five of its seven AI computing sites remain under development.

PwC expects annual data-centre investment to keep rising through 2050 as servers, GPUs and other computing equipment are replaced repeatedly.

Indonesia has power and a large digital market, but AI data centres still need dependable grid capacity and sites where that electricity can be delivered.

Nvidia is using Palantir software and its own AI models to guide supply-chain decisions, with staff retaining final control.

d-Matrix will use Nvidia's NVLink Fusion in Raptor systems due in 2027, keeping familiar rack infrastructure around a different inference chip.

Nvidia's Australian AI expansion depends on eight partners delivering the sites, power and data-centre capacity behind a 2GW target.

Huawei chairman Guo Ping says the company wants to become the "Nvidia of ICT", putting computing and connectivity at the centre of its AI strategy.

CoreWeave linked hundreds of Rubin GPUs and added storage changes aimed at reducing cross-region write delays for AI workloads.

Eight operators are testing Nokia and NVIDIA AI-RAN technology, but none has announced a commercial deployment.

Nvidia and eight Australian infrastructure partners are preparing up to 2GW of land, power and data-centre space for AI systems by 2027.

A factory designed for successive generations of computing is selling more than its first installation. The value of the next one depends on what survives—and who pays to change the rest.

OpenAI will take dedicated compute from two planned Malaysian sites as Firmus expands its regional AI infrastructure.

The Indian cloud provider has moved Nvidia Blackwell capacity deployed in May from pay-as-you-go use to a longer-term customer agreement running through June 2029.

MediaTek will help customers develop custom AI accelerators that connect to Nvidia’s interconnect and rack-scale architecture.

NVIDIA says Vera CPU systems are shipping at scale, moving its new AI processor from launch plans into customer integration.

NVIDIA has committed US$36 billion to cloud service from selected AI-cloud partners that first procure its data-centre hardware. The unusual term is not the six-year horizon. It is the switch: a partner can replace NVIDIA with a third-party customer offering a more advantageous rate. That makes the commitment a capacity backstop whose economic weight falls as outside demand—or NVIDIA’s own research—uses the machines.

NVIDIA is receiving substantial customer cash before recognizing the associated revenue, yet its quarter-end advance balance, half-year cash-to-revenue flow and long-duration contract obligations are not one metric. The distinction matters because the same quarter also carried extended payment terms in the opposite direction.

NVIDIA has more than doubled the supply and manufacturing capacity it has reserved in a single quarter. The US$279 billion figure is not current spending or guaranteed sales; it is a three-year synchronization test between memory, factories, new chip architectures and data centres that still need capital and power.

The minority investment brings Nvidia into the site and power work that takes place before AI hardware can be installed in a data centre.

The order launches L&T's AI Factory business at its Chennai campus, where the 10,000-GPU deployment must now move through integration and commissioning.

Nvidia's credit support covers the first 4.25GW tranche at PORTS-Pike, where OpenAI has a 20-year lease and phased delivery targets start in 2028.

Nvidia's guarantee covers leases for 4.25GW of IT capacity, helping SB Energy finance the Ohio campus before OpenAI starts paying rent.

Nvidia's 6 January 2025 CES announcement proposed a compact developer system with 128GB of coherent memory and a local-to-cloud software path. Its FP4 petaflop, model-size, price and availability figures described an intended product envelope, not shipment, independent testing or delivered economics.

The Miami-area demonstration used Verizon spectrum and AI software to detect and track drones, with pilot deployments planned ahead of targeted 2027 availability.

NVIDIA has recruited six major financial institutions to build independent compute-financing platforms intended to mobilise more than $500 billion over time. The scale is consequential, but the announcement contains no executed final agreement, borrower, site, drawdown or energised megawatt.

Nvidia’s proposed guarantee could support financing for OpenAI’s Ohio lease, but the terms remain under negotiation and the first phase is not yet funded.

NVIDIA Corporation is a risk and accountability case because the accountability issue is that software trust extends beyond the breached company when certificates, drivers, source code, and developer ecosystems can be reused or abused after disclosure. The public record matters for GPU users, developers, enterprises, driver distributors, endpoint-security vendors, gamers, cloud operators, and procurement teams needed evidence that software-trust repair reached certificates, binaries, and abuse monitoring.

Firmus Technologies and DayOne are developing a 360MW Nvidia DSX AI Factory campus in Batam, Indonesia, under an eight-year partnership with Nvidia, targeting AI-native multi-tenant workloads and large-scale GPU deployment.

DayOne, Firmus and Nvidia plan a 360MW AI campus in Batam, highlighting how Southeast Asia's AI infrastructure is extending beyond Singapore to markets with greater power, land and room for expansion.