1. Consumer GPU Prices Surge, Huaqiangbei Sees "Most Expensive PC-Building Season"
Driven by AI server demand and rising memory costs, consumer graphics card prices have skyrocketed. The RTX 5060 Ti 8GB rose by 700 RMB in one week to 3,400–3,500 RMB; the RTX 5070 increased by 1,000 RMB to around 6,000 RMB; and the RTX 5090 jumped nearly 4,000 RMB to 34,000 RMB. Merchants report they are "too afraid to stock up," and student foot traffic has dropped by one‑third to two‑thirds.
2. NVIDIA Raises GPU Prices Again Due to Memory Shortages
NVIDIA has issued another round of price hikes to downstream graphics card manufacturers, with increases of 20%–30%, primarily driven by the continued rise in GDDR6 and GDDR7 memory prices. This is NVIDIA's third GPU price increase this year.
3. NVIDIA Vera Rubin Enters Mass Production, Multiple Cloud Providers Deploy
NVIDIA's Vera Rubin platform has entered full‑scale mass production, with racks from several cloud service providers already operational. An AI company is also expanding its computing power by introducing thousands of Vera Rubin GPUs. The platform delivers a 10× improvement in performance per watt over the previous generation.
4. Huang Renxun Rejects "GPU as Nuclear Bomb" Analogy, Highlighting US‑China Chip Dilemma
Huang publicly dismissed comparisons of GPUs to atomic bombs as "stupid," emphasizing that GPUs are general‑purpose computing tools used by billions of people, aiming to undermine the "national security" narrative behind export controls. NVIDIA's revenue share from China has dropped sharply from its peak to about 9%.
5. 15th Five‑Year Plan Adds 4 Trillion Yuan in Computing Network Investment, Domestic Super‑Nodes Poised for Takeoff
The National Development and Reform Commission announced that the 15th Five‑Year Plan will add 4 trillion yuan in direct investment for computing network construction. Industry observers suggest that 2026 could be the first year of significant volume growth for domestic super‑nodes.
6. Google's TPU Count Expected to Reach 15 Million by 2028, Surpassing NVIDIA for the First Time
Google is projected to deploy 12 to 15 million TPUv9 units by 2028, while NVIDIA's AI GPU sales are estimated at 12.4 million units for the same year, marking the first time Google's TPUs will lead in volume. The ninth‑generation TPUv9 will use a 4‑chip package and adopt new packaging technologies.
7. Kimi K3 Exhausts Compute Cluster Within 48 Hours, Suspends New C‑User Subscriptions
Just 48 hours after launch, Kimi K3's request volume maxed out its existing cluster, prompting a suspension of new consumer‑side subscriptions. This highlights the acute shortage of computing power. The model has already been adapted to multiple domestic computing platforms, further driving demand for domestic super‑nodes.
8. MoonShot Signs Agreement with Alibaba for Approximately 20,000 NVIDIA Chips
Chinese AI startup MoonShot (developer of Kimi) has signed a contract with Alibaba for the use of about 20,000 NVIDIA chips. Although Alibaba also develops its own large language model, it continues to provide computing support to MoonShot.
9. Tech Giants Sharply Raise 2026 Capital Expenditures, AI Computing Investments Keep Expanding
Several major technology companies—including Google, Meta, Amazon, and Microsoft—have significantly increased their 2026 capital expenditure plans, sustaining the strong momentum in AI computing hardware.
10. AI Infrastructure Sector Booms, GPU Utilization Becomes a Key Focus
The explosive growth in AI inference demand has led to the commercialization of multiple open‑source inference engines, which have secured large funding rounds. Research shows that in large GPU clusters, nearly 20% of execution time and about 11% of energy are wasted on waiting. Improving GPU utilization has become a new multi‑billion‑dollar market.