Global GPU Computing News【20260816】

1. NVIDIA Establishes $500 Billion Computing Power Financing Platform with Six Wall Street Institutions

On August 10, NVIDIA announced the signing of a memorandum of understanding with six major financial institutions – Apollo Global Management, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR – to create an independent computing power financing platform. The long‑term goal is to mobilise over $500 billion in third‑party capital for AI infrastructure construction. Jensen Huang stated that NVIDIA’s support under this initiative would not exceed 25% of any single project, positioning the company as a complementary role. Market reaction was cautious; NVIDIA’s share price fell more than 3% intraday on the announcement.

2. CME Group to Launch World’s First Computing Power Futures Contracts in October

On August 11, CME Group and GPU market data firm Silicon Data announced plans to launch two computing power futures contracts on October 5 – the Silicon Data H100 Lease Index Futures and the B200 Lease Index Futures. Each contract unit represents 730 GPU‑hours (approximately equivalent to one calendar month of operation for a single card) and is cash‑settled. This marks the first time that AI computing power has a daily mark‑to‑market “benchmark price” that can be cleared through a central clearing system.

3. SpaceX Announces Exclusive Adoption of NVIDIA GPUs, Targeting Nearly 10 GW of Computing Capacity

In its earnings call, Elon Musk announced that SpaceX will exclusively use NVIDIA chips for all future AI infrastructure, calling the NVIDIA Vera Rubin architecture the best in the market. SpaceX expects to have over 2 GW of computing capacity by the end of 2026, and nearly 10 GW cumulative capacity by the end of 2027. The first Starmind AI satellites are scheduled for launch in 2027, with each satellite equipped with NVIDIA Rubin GPUs and Vera CPUs.

4. Tencent’s Q2 Capital Expenditure Reaches RMB 52.8 Billion, Resale Margin on Computing Power Exceeds 30%

On August 12, Tencent reported its Q2 2026 results, with capital expenditure reaching RMB 52.8 billion, far above the expected RMB 32.1 billion and nearly triple the RMB 19.1 billion from the same period last year. President Martin Lau revealed that reselling some computing power orders could generate a profit margin of over 30% above procurement costs. Combined capital spending by ByteDance, Tencent, Alibaba, and Baidu is growing at an annual rate exceeding 80%.

5. Alibaba Cloud Launches Lingjun Zhenwu M890 Supernode Instance

On August 12, Alibaba Cloud announced the official launch of the Lingjun Zhenwu M890 supernode instance, with initial sales in the Ulanqab region. This is the first domestic supernode‑based computing power capable of successfully running large models with over 2 trillion parameters. Kimi K3 and Qwen3.8 Max are already being served through this instance.

6. Domestic GPU Firm Moore Threads Reports H1 Revenue Exceeding Full‑Year 2025

Moore Threads released its first‑half 2026 report, with operating revenue of RMB 1.736 billion, up 147.42% year‑on‑year, exceeding its total revenue for all of 2025. Net loss attributable to shareholders narrowed by 95.73%, steadily approaching breakeven. The company also announced on August 9 its intention to list in Hong Kong, establishing an “A+H” dual‑listing structure.

7. AMD Launches MI455X GPU, Raises Prices Across RX 9000 Series

At its Advancing AI 2026 event, AMD officially unveiled the Instinct MI455X GPU for next‑generation AI infrastructure. Concurrently, AMD issued a Q3 price adjustment notice to partners, raising prices across the Radeon RX 9000 series, with initial increases of up to 50%. In addition, AMD completed a record $4.75 billion bond offering to boost its AI chip investments.

8. NVIDIA, Google, and Microsoft Introduce 800V High‑Voltage DC Power Architecture

NVIDIA, together with Google and Microsoft, released the industry’s first 800V high‑voltage DC power architecture through the Open Compute Project, pushing AI factories away from AC power supply. Over 80 equipment manufacturers are already following the new specification. The NVIDIA Rubin MGX platform based on this architecture is now in production and is scheduled for shipment in the second half of 2026.

9. Domestic Computing Power Industry Chain Sees Broad Earnings Surge

The AI computing power supply chain is experiencing a wave of strong earnings reports. Cambricon posted H1 revenue of RMB 5.996 billion, up 108.13% year‑on‑year, with net profit of RMB 2.311 billion, up 122.61%. Hygon Information reported H1 revenue of RMB 9.099 billion, up 66.52%. Foxconn Industrial Internet recorded H1 revenue of RMB 557.86 billion, up 54.6%, and net profit of RMB 23.74 billion, up 96%.

10. Sichuan Province Issues New Policies to Support Construction of Ten‑Thousand‑Card Intelligent Computing Clusters

On August 13, Sichuan Province released the “Measures for Promoting the Construction of Computing Network Infrastructure (Trial)” to support the deployment of intelligent computing clusters with over 10,000 cards in the Chengdu Plain core computing zone and the Panxi‑West Sichuan integrated computing‑energy development belt. The measures also encourage the construction of smart microgrids, computing power dispatch platforms, computing power supermarkets, and Token factories.