BTC $83,117.91 -0.37%
ETH $2,663.85 +0.25%
BNB $757.73 -1.84%
XRP $1.48 -0.75%
SOL $117.62 -1.87%
TRX $0.3341 +0.05%
DOGE $0.0930 -1.20%
ADA $0.2432 -2.23%
BCH $305.39 -3.53%
LINK $14.71 +5.98%
HYPE $87.27 -3.06%
AAVE $148.59 -1.02%
SUI $1.11 -9.17%
XLM $0.2252 +6.66%
ZEC $1,378.80 -11.77%
AAPL $337.78 -0.73%
AMZN $246.05 -1.15%
GOOGL $341.95 -0.10%
MSFT $507.41 -1.90%
META $712.04 -2.86%
NVDA $228.38 +1.90%
TSLA $356.73 -3.70%
SNDK $1,696.12 -2.25%
INTC $113.99 -4.53%
SPCX $145.84 -2.12%
MU $1,050.83 -1.47%
AMD $605.08 -2.08%
BTC $83,117.91 -0.37%
ETH $2,663.85 +0.25%
BNB $757.73 -1.84%
XRP $1.48 -0.75%
SOL $117.62 -1.87%
TRX $0.3341 +0.05%
DOGE $0.0930 -1.20%
ADA $0.2432 -2.23%
BCH $305.39 -3.53%
LINK $14.71 +5.98%
HYPE $87.27 -3.06%
AAVE $148.59 -1.02%
SUI $1.11 -9.17%
XLM $0.2252 +6.66%
ZEC $1,378.80 -11.77%
AAPL $337.78 -0.73%
AMZN $246.05 -1.15%
GOOGL $341.95 -0.10%
MSFT $507.41 -1.90%
META $712.04 -2.86%
NVDA $228.38 +1.90%
TSLA $356.73 -3.70%
SNDK $1,696.12 -2.25%
INTC $113.99 -4.53%
SPCX $145.84 -2.12%
MU $1,050.83 -1.47%
AMD $605.08 -2.08%

cyc

All
Article
Flash

first_img Nikkei: The release cycle of AI models in China and the United States has been shortened to an average of 44 days

According to the Nikkei, the release cycle of new AI models in China and the United States has averaged 44 days since April 2026, about one-third of the previous duration. The report compiled data from five American companies, including Anthropic and OpenAI, and four Chinese companies, including Alibaba Group and Moon's Dark Side, showing that the average interval from January 2023 to March 2026 was 125 days, which shortened to 44 days from April to September 2026.Entering September, OpenAI and Anthropic successively released their latest models. Meta has been upgrading Muse Spark monthly since July, and Google launched an updated version of Gemini on September 2 after three weeks. Grok released new models for three consecutive months until August. In China, DeepSeek has been updating monthly since July, and after August, Alibaba and Z.AI, which launched the GLM series, also successively released new models.On September 17, Anthropic announced that as of August, 26% of its development work was dominated by AI, with over 90% involving AI participation, whereas the proportion of AI dominance was nearly zero in February of this year. For OpenAI, in August, AI agents operated 3.1 times longer than human researchers, with average daily usage of about $600 for ordinary researchers, and the top 10% exceeding $7,000; the amount of code written through programming by the company in August reached seven times the average level of 2025.

first_img Jonah Burian: Lessons from the Crypto Cycle Apply to AI Investment

Blockchain Capital investor Jonah Burian stated: Cryptocurrency accelerates market cycles, early companies gain liquidity through tokens, and market behavior is assumed to be public, with lessons applicable to AI investment. Huge results are contagious and trigger FOMO; after Bitcoin became a trillion-dollar asset, Ethereum and Solana proved that more significant outcomes are possible, leading to the rise of alt-L1 trading, with VCs treating new L1s as lottery funding. Similarly, OpenAI and Anthropic are moving towards the trillion-dollar scale, with each new lab being priced as a lottery.L1s were financed at billion-dollar valuations based on white papers and founding teams, while new labs are raising billions based on research papers and teams poached from OpenAI, Anthropic, or Google DeepMind. After hot money floods in, rapid capital speculators always emerge; cryptocurrency has experienced operations like tokens as products, high FDV with low circulation, and a similar dynamic is unfolding in AI, but AI prices are formed in opaque semi-liquid secondary markets, unlike publicly traded cryptocurrencies.The blockchain space has shifted from scarcity to abundance, becoming a commodity, with applications capturing most of the value. In AI, models may become commoditized, with value shifting upstream and downstream, while applications and hardware layers gain profits, and the model layer gets squeezed. During the frenzy phase, financial capital excessively funds infrastructure, and new labs bet that it is only reasonable to assume large-scale R&D returns, with the history of cryptocurrency and alt-L1s warranting skepticism unless AGI arrives.

first_img Google disassembles retired servers to recycle DDR4 in response to memory shortages

Google's Senior Director of Supply Chain Infrastructure, Nikhil Cherian, revealed that to overcome memory bottlenecks, Google is developing software and hardware solutions and dismantling retired servers to recycle DDR4 components to establish an internal recycling supply chain. Google has designed special hardware adapters to connect the previous generation DDR4 to the new generation of AI servers while importing retired servers to remove their DDR4 modules for recycling.Cherian stated that the AI industry has rapidly shifted from being compute-constrained to memory-constrained, with high-performance memory accounting for about 75% of the bill of materials cost for a given AI server. A Goldman Sachs report indicated that memory prices will continue to rise in the third quarter, with personal computer DRAM prices expected to increase by 18% to 23% and server DRAM prices expected to rise by 13% to 18%. Trendforce data shows that in August, the spot market prices for DDR4 8GB and DDR5 8GB rose to $142 and $133, respectively.The two TPU ASICs launched by Google this year have been optimized for memory design, claiming to reduce memory demand to one-sixth of the original. The TPU8i chip features a dedicated layered memory design that relies on a high-speed DDR5 memory architecture to perform host-level tasks, with each chip equipped with 288GB of HBM3e high-bandwidth memory.
app_icon
ChainCatcher Building the Web3 world with innovations.