Project
Nvidia's next-generation GPU architecture and rack-scale platform succeeding Blackwell, built around HBM4 memory for AI data centers.
Microsoft's Maia 300 and Nvidia's Rubin cannot be ranked on speed, because their published performance figures use different baselines. TechRepublic's comparison found no independent, workload-matched benchmark for the two.
Reality
Nvidia contributed its Vera Rubin rack and compute tray designs to the Open Compute Project at the project's Global Summit in San Jose. Nvidia says builders get common blueprints for racks, networking, power and cooling; the benchmark figures released alongside are its own.
Perspective Coverage
4 publishersReality
NVIDIA committed $1 billion over five years and announced seven national-lab systems, led by Argonne's Solstice with around 100,000 Blackwell GPUs. Solstice targets AI work at 4 to 16 bits, so labs that need double precision are left weighing emulation against native FP64 hardware.
Reality
SpaceX is seeking $40 billion in Apollo-led debt to buy Nvidia chips, $30 billion of it investment-grade, the Financial Times reported. The deal would close in 2027, the year most of its $28 billion in AI commitments come due.
Perspective Coverage
21 publishersReality
Samsung Electronics plans up to 110 trillion won in shareholder returns this year after quarterly operating profit topped 100 trillion won for the first time. Most of that sum is still a range, and its final size depends on a memory business that earned nearly all of the quarter's profit.
Reality
CoreWeave has announced Forge, a platform linking model training, inference and evaluation so production results improve models and agents. Buyers comparing AI clouds now have a tooling claim to test next to chip capacity.
Reality
Two former Groq engineers sued the startup's board in Delaware, saying Nvidia's $20 billion license-and-hire deal was a sale that needed a stockholder vote. A ruling for them would extend that vote to other deals built the same way.
Perspective Coverage
7 publishersReality
CoreWeave is leasing 240MW from AdaniConneX in Navi Mumbai for its first India data centres, with the first phase due in mid-2028. Teams that must keep AI work in India get a 2028 supplier to plan around, though reports differ on how far the site can grow.
Perspective Coverage
3 publishersReality
South Korea's science ministry has earmarked 3.9 trillion won of a 4.7 trillion won AI budget for 10,000 Nvidia Vera Rubin GPUs, a dev.to analysis reports. Whether that buys a usable domestic model depends on who trains on the cluster and who pays for what it produces.
Reality
Samsung has more than doubled its high-bandwidth memory share in a year, eating into SK Hynix's lead, Rest of World reports. For anyone buying AI capacity, the contest could affect how fast and how cheaply new clusters arrive.
Reality
The accelerator is in full production and the headline number is a single-request generation rate at 100,000 tokens of context. That is a different purchase order than throughput.
Perspective Coverage
3 publishersReality
Record data center revenue of $75.2 billion and a 25-fold dividend increase sit next to a second-quarter guide implying 11.5 percent sequential growth, against 20 percent just delivered.
Reality
Jalapeno's lead is measured against last generation and the volumes are tiny, but a first-pass ASIC clearing Nvidia, AMD and Google parts reprices the design barrier, not the supply chain.
Perspective Coverage
7 publishersReality
Rack-scale Secure AI Factory becomes orderable in September, with Supermicro compute from October. What is being sold is validated integration, and the loss line is idle capacity.
Reality
NVIDIA's open source Topograph discovers cluster topology from five named clouds or from ibnetdiscover on-premises, then republishes it as node labels, Slurm config or Slinky ConfigMaps whenever a watched part of the cluster changes.
Reality
SpaceX committed its entire AI build to Nvidia hardware on its first earnings call as a public company, and AMD fell about 9% on a day it had a record $11.5 billion quarter on the table. Nvidia is also a SpaceX shareholder.
Reality
The Windows on Arm port and the compute-capability-107 Rubin preview are the headline items, but the two lines that touch a running cluster are the unbundled driver installer and the new CDMM default on coherent platforms.
Reality
NVIDIA Research rebuilt the serving stack for someone else's open-weight video model and got a five-second clip out in 1.653 seconds, but the default profile is lossy and the dense comparison run is the honest baseline.
Reality
The floor on Figure's compute commitment is about twice all the equity it has disclosed raising, and the vendor writing that contract now owns a piece of the customer owing it. Whether the $3.5bn is take-or-pay is undisclosed.
Reality
The contract value is public and the size of Nscale's stake in Figure is not, which leaves anyone reading a $51bn backlog ahead of a US listing unable to tell how much of that line the seller helped pay for.
Reality
Invest · September 1, 2026 · One report1 publisher