Cloud Engine-Cost-effective AI GPU Rental Platform

 A100 pcie40G - $ 200/card/month;BMS:8*Ascend 910B $ 3000/month  8*4090D - $ 1000/month  8*4090 - $ 1200/month  8*A100 pcie40G - $ 1800/month  8*A100 pcie80G - $ 4000/month  8*A100 nvlink80G - $ 5400/month  8*A800 nvlink80G - $ 5400/month  8*H20 - $ 6000/month  8*L20 - $ 2000/month  8*L40 - $ 2400/month  8*L40S - $ 2500/month  8*H100 - $ 13000/month  8*H200 - $ 20000/month  8*B200 - $ 30000/month  8*B300 - $ 48000/month 

Global GPU Computing News【20260621】

  1. ByteDance in talks to purchase 50,000 AI inference chips from Tianzhi
    Industry sources indicate that ByteDance is negotiating to purchase at least 50,000 AI chips from Tianzhi (Iluvatar CoreX), primarily for large‑model inference workloads, targeting the Tianzhi Zhikai series of cloud inference GPUs. If the deal goes through, Tianzhi will become ByteDance’s third GPU supplier after Huawei and Cambricon. ByteDance has made it clear that it will completely split its training and inference chip supply chains – training uses Huawei Ascend and Cambricon high‑end training cards, while inference adopts the Tianzhi Zhikai series.


  1. ByteDance significantly raises its 2026 computing power procurement budget
    Industry forecasts suggest that ByteDance plans to invest 150 billion yuan in global computing power procurement in 2026, of which about 60–65 billion yuan will be spent domestically, with domestic chips expected to account for over 40 billion yuan. Other reports indicate that ByteDance has raised its 2026 AI infrastructure capital expenditure budget by approximately 25%, to 200 billion yuan. In the first quarter of 2026, China’s inference demand has already reached eight times training demand, making inference computing the main driver of demand.


  1. NVIDIA B200 GPU rental prices to surge 94%
    The CEO of Baseten, an AI inference infrastructure provider, revealed that its cloud service providers have notified that NVIDIA B200 GPU rental prices will rise from $2.63 per hour to $5.10 per hour upon contract renewal in October this year, an increase of about 94%. At the same time, the delivery lead time for purchasing 1,000 GPUs has extended to 12–15 months, with new orders generally scheduled for the second quarter of 2027.


  1. Domestic AI chips repriced, with quoted prices expected to rise 20–30%
    New domestic GPU products (such as Cambricon 590/690, Huawei Ascend 950, etc.) are being repriced, with quoted prices 20–30% higher than earlier expectations, and gross margins could reach 50–60%. With current high‑end domestic computing cards operating near full capacity, demand is strong, supply is tight, and customers are competing for resources, fueling continued expectations for repricing of next‑generation domestic GPUs and AI chips.


  1. Cambricon shares surge to record high, market cap exceeding 940 billion yuan
    Driven by news of ByteDance’s procurement of domestic chips, on June 18, Cambricon’s stock rose more than 14% to a record high, pushing its market capitalisation above 940 billion yuan. The overall computing chip sector strengthened, and the STAR Chip ETF gained nearly 4%. Cambricon had previously announced that it would apply for a comprehensive credit line of up to 12 billion yuan from banks.


  1. Domestic computing power sector rallies, STAR Chip Index up over 5%
    On June 18, the domestic computing power sector continued its upward trend, with the Shanghai STAR Chip Index rising 5.1%. Among its constituents, Cambricon rose 14.02%, followed by Hygon, Jinghe Integration, and Shengke Communication. The computing power leasing concept also moved higher, with Dongfang Guoxin hitting the daily limit of 20%.


  1. Major internet giants race to build computing power infrastructure
    Apart from ByteDance, Baidu is deploying 10,000‑GPU‑level AI computing clusters nationwide, and its Kunlun P800 has completed large‑scale validation. Alibaba’s single‑quarter capital expenditure in fiscal year 2026 exceeded 38 billion yuan, with an estimated total of over 380 billion yuan over the next three years to boost cloud and intelligent computing hardware. Tencent plans to introduce domestic computing power on a large scale in the second half of 2026 and offer services such as its Hunyuan large model MaaS.


  1. UBS points out that NVIDIA’s Blackwell platform will dominate 2026 shipments
    UBS analysts noted that NVIDIA’s Blackwell platform will dominate shipments in 2026, while Rubin Ultra will gradually join in the fourth quarter of 2026. In contrast, AMD’s Helios platform motherboards are expected to ship in the fourth quarter, but full rack deployments may be delayed until the end of 2026.


  1. NVIDIA and Intel cross into each other’s turf
    NVIDIA, the global GPU leader, is moving into Intel’s home ground – CPUs and PCs – launching the RTX Spark ARM‑based AI chip and the Vera‑architecture high‑end CPU. At the same time, Intel, the CPU incumbent, is also eyeing NVIDIA’s AI computing territory, as both companies compete fiercely on each other’s traditional grounds.


  1. Domestic large‑computing‑power chips showcased at China International Financial Exhibition 2026
    On June 17, at the 2026 China International Financial Exhibition, domestic CPU and GPU chip companies such as Hygon, MetaX (Muxi), Zhaoxin, and Phytium made a collective appearance. These companies presented full digital solutions tailored to the financial industry through independent exhibition areas or keynote speeches.


Sales:+8619886543278 (WhatsApp)
Email:yqtxben@163.com
Address:2nd Floor, Block A, Garden City Digital Building, Nanshan District, Shenzhen City, China.
Contact US