三季度科技圈重要事件一览：科技巨头财报密集发布，存储大会与Hot Chips 2026接力登场
Related
Serenity: XFAB upgraded to "Buy"; AI and photonic chips may unlock long-term growth potential.
BlockBeats reported on July 6th that Serenity reposted Kelper's latest research report, upgrading semiconductor foundry XFAB's rating to "Buy." The report argues that Melexis demand is continuously improving, and the demand for silicon carbide (SiC), gallium nitride (GaN), and photonic chips driven by AI is expected to offset cyclical fluctuations in the automotive industry. Furthermore, the report notes that specialized foundries, scarce in the West, are receiving higher valuation premiums. Data shows that XFAB's current 2027 EV/EBITDA is only 4.8x, significantly lower than the industry median of 14.5x. Serenity stated that this indicates XFAB's core fundamentals are improving, but they are more optimistic about the growth potential of photonic chips, which the market has not yet fully priced in. They pointed out that XFAB leads the EU's PhotonixFAB project, which aims to build a European silicon photonics industry chain and promote the large-scale manufacturing of photonic chips. In the future, XFAB is expected to benefit from the growth in data center optical communication demand driven by AI giants such as Nvidia. However, since related commercial orders have not yet been implemented on a large scale, traditional valuation models still cannot reflect this long-term growth logic.
Samsung begins mass production of next-generation enterprise-grade SSDs for Nvidia Vera Rubin.
According to BlockBeats, on July 8th, Samsung Electronics began mass production of its most advanced data center storage hard drive, which will be used in NVIDIA's upcoming Vera Rubin platform. This enterprise-class solid-state drive, named PM1763, was first announced earlier this year at NVIDIA's GTC conference. Samsung also showcased its next-generation high-bandwidth memory (HBM4) and low-power SOCAMM2 modules at the time, as part of a comprehensive solution for AI data centers. Samsung stated on Wednesday that the PM1763 features the company's latest V-NAND flash memory chips and a newly developed 4nm controller, offering more than double the read and write speeds of its predecessor. The PM1763 is designed to reduce data latency for advanced processors and AI accelerators. To maintain high-speed performance during AI training and inference, the drive employs a liquid-cooling design.
The recovery in semiconductor demand is spreading from AI to a wider range of industrial end markets, and supply chain tensions are shifting from digital chips to analog chips.
According to BlockBeats, on July 4th, analog chip giant Analog Devices (ADI) sent a formal letter to its customers, stating that due to a significant increase in overall semiconductor demand and continued supply shortages, lead times for some product portfolios have been extended to up to six months. ADI recommends that customers place orders at least six months in advance through channel partners and extend their order backlog accordingly. Critini Research analyst Jukan stated that this is an important signal of a recovery in semiconductor demand. Previously, market focus was on the shortage of AI training chips and HBM memory. The extended lead times for analog chips indicate that the demand recovery is spreading from AI to broader end markets such as industry and automotive, and the supply chain tensions are shifting from digital chips to analog chips.
SemiAnalysis: SPHBM4 shifts the complex engineering burden of AI chips to the substrate layer
According to a Odaily by SemiAnalysis on the X platform, SPHBM4 will shift the complex engineering burden of AI chips. Chip manufacturers will no longer purchase the extremely expensive proprietary "silicon interposer + ABF substrate" combination, but will instead turn to purchasing ultra-large, high-layer-count ABF substrates, or adopt glass substrates ahead of time when performance requirements are directly pushed to the substrate layer. The substrate boom has just begun.
South Korea plans to invest 5 trillion won in excess tax revenue from semiconductors to develop "sovereign AI" and purchase 10,000 GPUs to support a single team.
According to Beating's monitoring, the Korea Economic Daily revealed that South Korea's Ministry of Science and ICT is in discussions with the Presidential Office and the Ministry of Finance, planning to utilize approximately 5 trillion won in excess tax revenue from the semiconductor industry to purchase about 10,000 of NVIDIA's most advanced Vera Rubin superchip GPU modules this year. This investment will be used to support elite teams developing world-class "sovereign AI" models. This decision stems directly from the sense of crisis caused by US export controls on cutting-edge AI models—Anthropic's Fable 5 and OpenAI's GPT-5.6 have been restricted from external access, forcing South Korean companies to urgently modify systems built on these models. The concept of "national strategic assetization" of AI models has thus become a reality. Currently, although four teams in South Korea—LG AI Research Institute, SK Telecom, Upstage, and Motif Technologies—are advancing independent basic model projects, each team only receives support from 700 to 800 GPUs. This dispersed investment makes it difficult to compete with globally leading models that utilize tens of thousands of GPUs. The government has therefore decided to shift to a "selection and concentration" strategy, allocating 10,000 GPUs to the best teams at once, and plans to include special funds in the supplementary budget for attracting top overseas AI talent. South Korean President Lee Jae-myung publicly stated last month that the GPU procurement pace was too slow and demanded that related funds be ensured when compiling the supplementary budget. The current global competitive window is extremely short, with companies like Anthropic and OpenAI iterating new models weekly. France, with its single-focus strategy, has surpassed Meta to become the seventh largest AI provider globally, thanks to Mistral AI. If South Korea cannot catch up in the short term, it risks becoming an "AI vassal state." The South Korean semiconductor academic community has given this positive feedback, believing that investing the funds gained from hardware prosperity into building software competitiveness is essentially an investment in the country's future. --------------------------------- Click the original link below to join the Beating · Lark AI news channel for 24/7 monitoring of global AI hotspots and news.
Perplexity announces it will use Nvidia's new CPU.
Mars Finance reports that AI startup Perplexity announced on Tuesday its plans to use Nvidia's new CPU chips. Nvidia has stated that it expects sales of its general-purpose Vera CPUs to reach $20 billion by the end of this fiscal year. Perplexity declined to disclose how many Nvidia CPUs it plans to purchase. However, Nvidia has previously revealed that OpenAI, Anthropic, and Oracle also plan to use its CPUs. (Cailian Press)