Global GPU Computing News【20261004】

1. Nvidia's market cap returns to $5.7 trillion, DGX Spark 64GB version released

On October 3, Nvidia's stock rose nearly 3% during intraday trading, and its total market capitalization returned to $5.7 trillion, up about $78 billion from the previous trading day, just one step away from becoming the first $6 trillion company in history. The simultaneously launched DGX Spark desktop AI supercomputer 64GB unified memory version will go on sale starting October 23 through OEM partners including Acer, Asus and Dell, with a starting price of $4,999. The new version retains the GB10 Grace Blackwell superchip, with CPU and GPU sharing unified memory through NVLink-C2C. A single unit can run local inference for AI models with up to 100 billion parameters; two devices directly connected via ConnectX-7 pool memory to 128GB, raising the upper model support limit to 200 billion parameters.

2. Nvidia, OpenAI and SoftBank jointly transform a Cold War nuclear ruin into an 8 GW AI superfactory

A former Cold War uranium enrichment plant in Ohio, USA, has been jointly transformed by Nvidia, OpenAI and SoftBank into the world's largest AI factory, with plans for 8 GW of power to support 1.5 million GPUs. Nvidia CEO Jensen Huang bet $1.5 billion and provided a 20-year credit guarantee, while OpenAI signed a 20-year ultra-long lease to fully take up the compute. At full load, the factory's compute capacity is close to two-thirds of the combined total of the world's current 82 top AI data centers.

3. Tencent signs an approximately $7 billion AI chip leasing agreement with Oracle

According to the Financial Times, Tencent agreed to lease multiple Oracle data centers in Southeast Asia for a five-year term, obtaining about 100,000 high-end AI chips, with the agreement valued at about $7 billion. Tencent has already paid about 30% as a prepayment. These chips cannot be obtained within China and will be used to advance Tencent's AI model development and autonomous AI agent development.

4. HPE wins Vultr's first AMD Helios order worth $1.2 billion

HPE announced it had won the first AMD Helios rack-scale AI system order, signing a $1.2 billion agreement with private cloud vendor Vultr. Each rack integrates 72 AMD Instinct MI455X GPUs, AMD Epyc Venice CPUs, Pensando Vulcano AI NICs and ROCm software, running UALink over Ethernet. HPE also raised its fiscal 2027 networking revenue growth forecast, mainly due to Helios.

5. Nebius raises on-demand GPU rents for the second time this year, by 17% to 21%

Cloud computing service provider Nebius sent a price adjustment notice to customers, raising on-demand rents for major NVIDIA GPUs starting October 1. H100 rose from $3.85 per hour to $4.50 per hour, H200 from $4.50 to $5.40, B200 from $7.15 to $8.50, and B300 from $7.85 to $9.50, an overall increase of about 17% to 21%, the second comprehensive price increase this year. Compute leasing concept stocks subsequently surged collectively.

6. Biren Technology wins the Kunlun Lingyi Intelligent Computing Center project as sole bidder for RMB 174 million

The inquiry procurement results for the Kunlun Lingyi Intelligent Computing Center project, Phase I, were announced. Biren Technology won Package One with RMB 174 million, becoming the sole winning supplier for that package, and must complete supply within 45 calendar days after contract signing. The core products are Bili series GPUs. Package Two, with a budget of about RMB 452 million, was scrapped because fewer than three suppliers submitted response documents.

7. Moore Threads and JD Cloud cooperate to build a 100,000-card domestic intelligent computing cluster

At the 2026 JD Global Technology Explorers Conference, JD Cloud announced plans to build a 100,000-card full-function GPU cluster, using Moore Threads full-function GPUs as the compute foundation, focusing on key areas such as large model training, inference and embodied intelligence, and opening compute to the entire industry. This marks the first time domestic GPUs have entered the 100,000-card-level core intelligent computing cluster of a leading AI cloud enterprise.

8. MIIT issues the 15th Five-Year Plan for the Information and Communications Industry, with a comprehensive upgrade in compute deployment

On September 7, the Ministry of Industry and Information Technology issued the 15th Five-Year Plan for the Development of the Information and Communications Industry, proposing that by 2030 China's intelligent computing scale will grow more than fivefold compared with 2025. The plan calls for accelerating the construction of a national integrated computing network, building a multi-level computing facility system of hub-region-edge, orderly deploying 10,000-card, 100,000-card and above intelligent computing clusters, increasing efforts to adapt domestic compute chips, advancing the construction of the computing power internet, and building a unified national computing service market.

9. Inner Mongolia regional node of national computing interconnection approved for construction

MIIT officially announced the list of regional nodes for national computing interconnection, and Inner Mongolia was successfully selected, joining the important node sequence of the 1+M+N architecture. The node will coordinate the construction of six core functional systems including the national computing internet regional service center and the computing identifier management system, vertically connecting with national computing internet service nodes and horizontally linking regional computing nodes across the country, to achieve dynamic compute monitoring, resource sharing and interconnection, and cross-domain collaborative scheduling.

10. Multiple A-share companies intensively disclose large GPU compute contracts

Multiple A-share listed companies intensively disclosed progress in their compute businesses. The subsidiary of Egova plans to purchase 64 server devices, with the cumulative amount estimated at RMB 576 million; Sinnet plans to add RMB 1.265 billion in investment to the Horinger intelligent computing center project, increasing total investment to RMB 2.5 billion; the subsidiary of Aoni Electronics signed a GPU compute card procurement contract with Company B, totaling $232 million; the subsidiary of Xingyun Technology signed an intelligent computing resource leasing service contract, with a total tax-inclusive amount of RMB 872 million. Orders across the upstream and downstream of the industrial chain are being released intensively, showing that domestic intelligent computing demand is accelerating into reality.