BTC $79,675.53 +3.20%
ETH $2,494.96 +2.39%
BNB $709.84 +1.96%
XRP $1.51 +2.12%
SOL $101.15 +7.39%
TRX $0.3457 +0.68%
DOGE $0.0911 -0.19%
ADA $0.2237 +1.74%
BCH $275.83 +1.56%
LINK $11.77 +3.53%
HYPE $79.72 -0.88%
AAVE $133.87 -2.86%
SUI $0.8126 -1.84%
XLM $0.1963 +1.16%
ZEC $833.45 +0.06%
BTC $79,675.53 +3.20%
ETH $2,494.96 +2.39%
BNB $709.84 +1.96%
XRP $1.51 +2.12%
SOL $101.15 +7.39%
TRX $0.3457 +0.68%
DOGE $0.0911 -0.19%
ADA $0.2237 +1.74%
BCH $275.83 +1.56%
LINK $11.77 +3.53%
HYPE $79.72 -0.88%
AAVE $133.87 -2.86%
SUI $0.8126 -1.84%
XLM $0.1963 +1.16%
ZEC $833.45 +0.06%

s

All
Article
Flash

first_img Analysis: CXMT's production capacity has peaked, and DRAM prices continue to rise sharply

Analysis indicates that China's major DRAM manufacturer CXMT's monthly wafer output has reached a peak of approximately 240,000 pieces by the end of 2025, and is expected to remain flat in 2026. Due to the tightening of U.S. export controls on advanced semiconductor equipment, especially EUV lithography machines, its capacity for expansion is limited, and substantial expansion will not occur until at least 2027, depending on the progress of the domestic equipment supply chain. According to Goldman Sachs data, CXMT's coverage of domestic DRAM demand is only about 41% in 2026 and about 50% in 2028, reflecting a structural bottleneck for many years.TrendForce data shows that traditional DRAM contract prices are expected to surge by 90%-95% quarter-on-quarter in the first quarter of 2026, followed by another increase of 58%-63% in the second quarter. Jefferies predicts that prices will continue to rise by 40%-50% and 30%-40% in the third and fourth quarters, respectively. The price increase for server DRAM is even steeper, with Samsung and SK Hynix proposing price hikes of 60%-70% to clients like Microsoft and Google in the first quarter. S&P Global expects Samsung's traditional DRAM revenue per bit to rise by 116% year-on-year to $0.79 in 2026, while Micron's ASP will increase by 54% to $1.06; Bernstein predicts that SK Hynix's DRAM gross margin could reach 92.7% in the fourth quarter of 2026.Multiple forecasts suggest that effective supply relief may not occur until the end of 2027 or even 2028.

first_img NVIDIA announces full-scale production of Groq racks, which will be deployed at Nebius

On Monday local time, Nvidia Senior Director Dion Harris announced that the Groq 3 LPX rack has entered full-scale production and will be deployed in the data center of the new cloud service provider Nebius, expected to go live later this year. This move marks the commercialization of the technology obtained by Nvidia after reaching a technology licensing deal worth approximately $20 billion with Groq last December. Several core employees from Groq have joined Nvidia, with founder Jonathan Ross serving as Nvidia's Chief Software Architect.Nvidia is accelerating the production of Groq chips and providing products to customers, highlighting the importance of low-latency inference. The Groq 3 LPX, released in March this year, is an inference accelerator for the Vera Rubin platform, integrating 500 megabytes of high-speed SRAM on the chip die to reduce memory bottlenecks. Each LPX rack can integrate 256 Groq 3 chips, and Nvidia cites benchmarks indicating it can process 3,400 tokens per second; the chip is manufactured by Samsung.Harris stated that low-latency chips are not meant to replace GPUs but are designed to use the appropriate processor for different stages of workloads, allowing cloud service providers to offer higher-priced packages for latency-sensitive users. Nvidia CEO Jensen Huang plans to allocate about a quarter of the data center space for programming applications to Groq chips, with the remainder using the Vera Rubin system, and expects cumulative sales of Blackwell and Vera Rubin to reach $1 trillion by 2027.

first_img Corporate AI spending continues to increase, with the growth focus shifting from subscriptions to APIs

FundaAI released a research report on enterprise AI applications, indicating that enterprise AI budgets are still expanding, but there is a divergence in trajectories in the second half of 2026 and 2027. The AI spending guidance from large U.S. telecom operator A shows an increase from a baseline of 100 in January to about 190 in December, with an expected year-on-year increase of 40%--50% in 2027; large European automaker A has only increased by 10%--15% so far this year, with guidance for next year remaining roughly flat.Incremental spending is shifting from paid seats to API/Token consumption and production workflows. The aforementioned telecom operator's subscription and API ratio has changed from about 50%/50% to 40%/60%, and it may trend towards 35%/65%; mid-to-large biopharmaceutical company A has adjusted from 80%/20% to about 70%/30%. Open-source adoption is uneven, with active scenario usage accounting for 30%--40%, as the unit price is lower, leading to a smaller spending proportion; experts estimate that open-source inference can be about 40%--70% cheaper than closed-source cutting-edge models, with the gap narrowing to 20%--40% under full cost metrics, and model routing, caching, and context compression could further reduce API spending by about 20%--30%.On the production side, AI budgets are increasingly built from the bottom up based on workflow ROI. The typical production ROI for this telecom operator is about 1.5--2 times, with a payback period of 6--18 months, and mature use cases can reach 3--5 times. The next wave of spending is related to agents, software modernization, network operations, commoditized workflows, and longer-cycle business processes, but engineering capacity, process reengineering, governance, and data readiness are becoming tighter constraints than funding.
app_icon
ChainCatcher Building the Web3 world with innovations.