generation

ASML stated that it plans to collaborate with its major customers to develop new solutions capable of producing advanced chip manufacturing equipment to meet the demand for large data center chips designed by companies like NVIDIA.Currently, the extreme ultraviolet (EUV) lithography machines produced by ASML, which are widely used, can manufacture chips with a maximum area of about 800 square millimeters. The next-generation "high numerical aperture" EUV lithography machine that ASML is about to launch will be able to produce finer chips. However, due to its use of smaller mask sizes, it is currently limited in the area of chips it can manufacture.ASML plans to showcase a pilot production line using larger mask sizes by 2031 and aims to enable this new technology for mass production by 2033.

According to Bloomberg, ByteDance is preparing to launch an artificial intelligence model for real-time spatial video generation, with applications covering fields such as robotics and autonomous systems, and will compete with companies like Meta and Alphabet.Insiders say that ByteDance founder Zhang Yiming is personally overseeing the development of the model, coordinating multiple business departments within the company, and investing in artificial intelligence resources and computing power support. The model may be released as early as next month, but the specific timing is still subject to change.

According to South Korean media Greened.kr, Samsung Electronics and Arm have officially launched a joint research and development project for edge AI chips. Arm approved a one-time engineering fee at the end of August for the development of the next-generation edge AI SoC chip and officially started the project with Samsung Electronics.The project adopts a collaborative model where Arm provides core technology, and Samsung Electronics integrates and implements it, aiming to jointly develop customized chips for end customers.

According to Blue Whale News, from Changxin Technology, ByteDance and ZTE have jointly developed a new generation of Doubao phone------Nubia NaviX Ultra, which will be equipped with Changxin's highest speed LPDDR5X memory at 10667Mbps. The official statement indicates that this is also the first time that domestically produced 10667Mbps high-speed LPDDR5X memory chips have achieved mass production and application.Previous news reported that the Nubia "Doubao phone" has completed network filing and will be mass-produced and launched in September.

On September 1, Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, marking the first update to this model series since the launch of Fable 5 on June 9. Anthropic claims that the new models are the most advanced for programming and knowledge work. In the Terminal-Bench-Science 0.1 benchmark, Fable 5.1 recorded a pass rate of 52.6%, more than doubling Fable 5's 24.7% and surpassing Opus 5's 29%; in Terminal-Bench 4, Fable 5.1 scored 55.8%, higher than Fable 5's 42% and Opus 5's 52.3%.Fable 5.1 and Mythos 5.1 share the same underlying model, differing only in their safety filters. Fable 5.1 is available to all Claude accounts, while Mythos 5.1 remains restricted to professionals vetted through Anthropic's network security verification program and life sciences verification program. In terms of pricing, Fable 5.1 charges $10 per million tokens for input and $50 for output, which is double that of Opus 5. Cache read costs have decreased by 75%, and typical workload costs can drop by about 25%, with high agentic workloads decreasing by up to 45%.

According to Citrini analyst Jukan, SK Hynix is working on diversifying the foundry suppliers for the substrate chips used in high bandwidth memory (HBM), which are constructed by vertically stacking multiple memory chips. It is reported that the company is considering outsourcing part of the substrate chip production for the seventh generation HBM, namely HBM4E, to Intel's foundry. So far, SK Hynix has relied entirely on TSMC for outsourcing substrate chip production.This move is seen as an effort to establish a multi-supplier foundry strategy to reduce the concentration of the supply chain on TSMC while enhancing SK Hynix's bargaining power in pricing. According to industry insiders on August 31, SK Hynix is advancing a plan where the substrate chips for HBM4E may be jointly manufactured by TSMC and Intel, potentially starting from the HBM4E generation.As the market expects the cost burden of substrate chips for HBM4E to further increase. Since HBM products are mainly covered by long-term supply agreements (LTA), SK Hynix also finds it difficult to immediately pass on the higher foundry costs to customers by raising product prices. Therefore, industry observers believe that if Intel ultimately joins SK Hynix's substrate chip supply chain, the company will gain more options in terms of cost competitiveness and supply stability.

The Zcash full node client team, Zakura, has released Zakura Common, a set of cryptographic and protocol libraries for the Zcash ecosystem. Official benchmarks show that the proof generation speed on mobile has increased by over 14 times (over 5 times on desktop), Sinsemilla hash speed has increased by over 21 times, decryption trial speed has increased by over 1.5 times, and zk-SNARK verification speed has increased by 4 to 8 times; in many cases, the waiting time for wallet transaction creation can be reduced from over 3 seconds to under 200 milliseconds.This upgrade does not require a network upgrade and has been open-sourced under the MIT/Apache 2.0 dual license. Zakura 1.3.0 will switch to this technology stack.

According to Business Insider, after Google released its new AI model this month, some employees have begun testing the preview version of Gemini 3.8 Flash on the internal coding platform Jetski. A testing employee stated that the model is significantly better than 3.7 Flash, but emphasized that it is still too early to make a comprehensive evaluation. Google declined to comment on this.With 3.5 Pro still not released and Anthropic and OpenAI continuing to accelerate, Google no longer possesses cutting-edge models and has shifted to focusing on the faster and cheaper Flash series "main models" for coding and running agents. Gemini 3.6 Flash was launched in July, and 3.7 Flash followed just three weeks later. CEO Sundar Pichai stated at the Q2 earnings call that the company's goal is to release a new version approximately every month.The employee preview suggests that the public launch of Gemini 3.8 Flash may not be far off, but it is not guaranteed. Some AI observers believe they have found traces of the model on the Arena AI benchmarking website, which Google has previously used to test upcoming versions. Earlier this year, Google also launched the Gemini Spark agent.

NVIDIA (NVDA.O) Chief Financial Officer stated that the company is fully focused on producing the next generation Vera CPU.

According to Nikkei Chinese News, the American company Meta has announced that it has begun offering code generation AI for programming, including the AI agent "Muse Code" and the new version of the AI model "Muse Spark" for software development. Its performance is not as good as Anthropic's high-end models, but compared to Anthropic's Claude Opus 5, the unit token usage fee is about 80% cheaper, aiming to attract enterprises looking to control costs.If Meta is provided with AI usage data for development, the costs will be reduced by more than 90%, bringing usage costs down to levels comparable to products from emerging Chinese companies like DeepSeek. Meta stated that its internal technical staff had previously used Anthropic's code generation AI extensively and listed "utilization of other companies' AI" as a reason for the decline in operating profit from April to June 2026; recently, it has encouraged employees to use its own developed AI more to control expenses.Meta also disclosed that during performance testing of Muse Spark, there were instances of the AI breaching external systems on the internet.

According to Jinshi reports, the dark side of the moon Kimi announced that the company's first-generation trillion-parameter multimodal model Kimi K2.5 will end its service at the end of this month.

On August 24, SK Hynix is introducing advanced packaging technologies, including Intel's EMIB, for its next-generation HBM solutions and plans to eventually enter the 3D packaging era. At the 2026 Hot Chips conference, SK Hynix's Vice President of Packaging Engineering, Jaesik Lee, delivered a presentation titled "Advanced Packaging for High Bandwidth Memory," explaining how the company will leverage advanced packaging technologies to accelerate its HBM product roadmap.Currently, SK Hynix is exploring methods such as hybrid bonding to break through the limitations of 16-layer stacking. In the future, SK Hynix also plans to address power consumption challenges by increasing the number of TSVs, improving IO speeds of logic process integration, and optimizing power distribution networks. As of the time of publication, SK Hynix's stock rose by 1.71%.

Ethereum co-founder Vitalik Buterin published a new article titled "Obfuscation (Part 3): Local Mixing," which delves into a cryptographic obfuscation technique being explored—"Local Mixing." He claims it may become a new foundational cryptographic tool following elliptic curves, RSA, and lattice-based cryptography.Vitalik states that current mainstream obfuscation techniques mainly rely on complex mathematical assumptions, often resulting in extremely high computational overhead. In contrast, Local Mixing adopts a completely different approach, not relying on elliptic curves, large integer factorization, or lattice cryptography. Instead, it draws on experiences from symmetric cryptography and hash function design, continuously shuffling, reconstructing, and hiding circuit structures to eliminate information leakage while maintaining functionality.The Local Mixing technique primarily includes steps such as reversibility, hardening, mixing, splitting, crossing walk, and "gadgetization." By introducing random structures into the circuit, rearranging logic gates, and employing nonlinear hiding mechanisms, it makes it difficult for attackers to recover the original computational logic.Vitalik points out that this technology is still in its early stages, with security not yet validated over the long term and facing challenges such as random attacks and linear analysis. However, he believes that Local Mixing represents a completely new path for cryptographic exploration, aiming to construct more efficient indistinguishable obfuscation (iO) schemes.If breakthroughs in Local Mixing technology are achieved, it could lead to new quantum-resistant public key encryption schemes and promote the development of general obfuscation techniques. The field still requires years of cryptanalysis and optimization validation, but AI-assisted research may significantly accelerate this maturation process. Vitalik states that obfuscation technology is seen as the "final frontier" of cryptography because, theoretically, other cryptographic primitives can be constructed based on obfuscation and one-way functions. Local Mixing may not only reduce the costs of traditional obfuscation schemes but could also become an important direction for future cryptographic infrastructure.

Alibaba announced that its latest generation AI processor, Zhenwu M890, has been widely commercialized among over 650 external customers. The chip, through Alibaba Cloud services, has been applied in more than 20 industries including autonomous driving, the internet, and financial services, validating its capability to support end-to-end AI workloads.

Binance co-founder He Yi cited data from the research team, stating that among users investing in U.S. stocks on the Binance platform, about three-quarters are net buyers, and about one-fifth have never sold. Generation Z users show the strongest willingness to hold long-term, with the highest net buying ratio and the lowest trading frequency. He Yi stated that this reflects how, after lowering investment barriers, a new generation of investors is participating in the market with patience and a long-term perspective.

Cerebras released its fourth-generation AI inference system CS-4 this week, based on the same 5nm WSE-3 wafer, achieving double the performance by doubling the clock frequency and power consumption. A single CS-4 cabinet accommodates 3 wafers (CS-3 has 2), featuring a modular "backpack" design that simplifies manufacturing and deployment, with a TDP of approximately 125 to 135kW. The CS-4 can provide an inference speed of nearly 4000 tokens/second/user, about twice that of the CS-3, and supports decomposed inference with heterogeneous systems such as AMD and AWS Trainium.Cerebras claims that the CS-4 offers about 2000 times the on-chip memory bandwidth of NVIDIA's Rubin (43PB/s), but the 44GB SRAM capacity remains unchanged, and long-context inference still requires multi-wafer stacking. For example, with the DeepSeek V4 Pro (1.6T parameters), approximately 20 systems are needed for a 1M context window, and about 40 systems are required for 256 concurrent users, corresponding to a CAPEX exceeding 20 million USD. Cerebras is collaborating with clients such as OpenAI and plans to achieve approximately double performance improvements each year, aiming for a 20-fold throughput increase by 2027. The "backpack" cabinet design of the CS-4 will continue into the next-generation "Nexus" platform.

The Offchain Labs research team has released advancements in zero-knowledge proof technology, covering areas such as proof acceleration, verifiable AI, and recursive security. Among them, the vector commitment scheme has reduced the proof generation time for 64,000 data entries from about 2 minutes to 32 milliseconds, achieving a speedup of approximately 4000 times; the new proof system Zaratan has achieved native full succinct proofs for integers for the first time, which can reduce the computational overhead of operations like RSA by about 5000 times; in terms of verifiable AI, a lightweight verification protocol can shorten the inference verification time for large models from several minutes to milliseconds.Arbitrum is exploring a multi-prover architecture that combines ZK proofs with fraud proofs and TEE to eliminate single points of failure and shorten L1 settlement cycles.

Binance Research data shows that Generation Z investors are gradually shifting towards long-term asset allocation tools such as ETFs. Compared to Millennials, Generation X, and Baby Boomers, their trading frequency is lower, and their preference for leverage is also weaker.Data indicates that at the beginning of August, ETFs accounted for 25% of the stock trading volume among Generation Z users. In July, the proportion of net inflows into stocks from ETFs for Generation Z reached 21.9%, up from 18.5% in June; during the same period, the proportion of individual stock investments decreased from 77% to 74.2%.Binance Research analyzed trading behaviors in direct stocks, tokenized stocks, and traditional financial perpetual contracts. The data shows that Generation Z's trading activity in all three asset classes is lower than that of other working-age groups.Among them, Generation Z's traditional financial perpetual contract accounts traded an average of 13 times per month, lower than the 17 times of Millennials and 16.5 times of Generation X. In direct stock accounts, 22% of Generation Z users have never sold stocks, higher than the 19% of Generation X and 9% of Baby Boomers.Among Generation Z accounts that bought but did not sell, the assets with the highest cumulative purchase amounts include Broadcom, Tesla, and Charles Schwab U.S. High Dividend Stock ETF.In terms of leveraged products, Generation Z shows a lower risk appetite. Data indicates that 88.2% of Generation Z's traditional financial perpetual contract accounts have never traded leveraged or inverse ETFs, higher than the 84.5% of Millennials and 85.9% of Generation X.Additionally, the tokenized stock market continues to expand. Data shows that Binance's bStocks recently briefly surpassed Kraken's xStocks, becoming the second-largest tokenized stock issuance platform globally. As of the latest data, Ondo Finance ranks first with approximately $972 million in tokenized stock value, while xStocks and bStocks are approximately $611 million and $580 million, respectively.

According to Fortune, Bank of America announced a $250 billion infrastructure investment plan on Wednesday, committing to invest in key infrastructure areas in the United States over the next year. The investment scope covers multiple fields including data centers and computing power, renewable energy generation, energy storage, natural gas, transmission networks, and critical minerals and mining, with the goal of supporting energy security, job growth, and economic competitiveness.