SemiAnalysis: Nvidia's system memory spending is projected to exceed 40% by 2027.
Related
AI capital expenditures are projected to reach $1.1 trillion in 2027, potentially exceeding U.S. defense spending for the first time.
According to Mars Finance, on July 5th, The Kobeissi Letter published an article stating that the AI spending boom is reshaping the US economy. AI capital expenditures by Alphabet, Amazon, Meta, Microsoft, and Oracle are projected to rise to approximately 3.2% of US GDP by 2027. If this forecast comes true, annual AI capital expenditures will surpass US defense spending for the first time, which is projected to account for approximately 2.7% of GDP next year. This year alone, the aforementioned companies' AI capital expenditures are expected to rise from 1.5% of GDP in 2025 to approximately 2.5%, approaching the approximately 2.7% share of defense spending. The AI capital expenditures of these five companies are projected to exceed $800 billion in 2026 and further rise to a record $1.1 trillion in 2027. These figures are "staggering."
Analysts are concerned that the memory chip sector may repeat Nvidia's pattern of "new highs in fundamentals, sideways stock price."
According to an article published on the X platform by Odaily researcher Jukan, market expectations for the memory chip sector are currently at a high level. Regardless of whether earnings reports exceed expectations, stock prices may face pressure. If earnings exceed expectations, the market may worry that the industry cycle has peaked; if earnings fall short of expectations, it may be interpreted as the end of the memory chip boom cycle. Jukan also expressed concern that the memory chip sector might repeat Nvidia's previous pattern of "continuously record-breaking fundamentals, but a prolonged period of sideways stock price," meaning that while corporate profits continue to break records, the stock price performance is relatively lackluster because market expectations have already been fully priced in.
SemiAnalysis: NVIDIA Delays or Adjustments to Multiple AI Rack-Scale Architectures, Limiting Rubin Ultra's Expansion Path
According to Mars Finance, on July 6th, SemiAnalysis reported that NVIDIA's latest rack-mount interconnect architecture, Kyber NVL144, has undergone a major adjustment just three months after its release, being delayed by more than 12 months to 2028. The main reason is the ongoing challenges in PCB planar design regarding manufacturing feasibility. Simultaneously, the NVL72x2 back-to-back rack architecture has been cancelled. This solution was originally intended to enhance the scalability of pure copper NVLink by deploying two Oberon racks back-to-back, but its complex structure and high operational burden on hyperscale cloud providers (CSPs) led to strong market skepticism and its eventual abandonment. Furthermore, due to the immaturity of CPO (Co-packaged Optics) technology, NVIDIA's larger-scale scaling solutions based on CPO NVSwitch (such as NVL576) may also continue to be delayed, or limited to small-batch trial production. This means that NVIDIA lacks a stable large-scale scale-up solution until CPO matures. Furthermore, the Rubin Ultra product roadmap has also changed: the originally planned "four-computing-chip" version has been cancelled, with only a "dual-computing-chip" version remaining, and the overall system-level performance is expected to drop to about half of the original solution. These adjustments mean that Nvidia's scale-up capabilities in the Rubin Ultra generation are limited. With the CPO NVSwitch unable to be deployed before the Feynman architecture, competitors such as the AMD MI500X or Google TPU v8i may have a relative window of opportunity in terms of scaling capabilities for large-scale training clusters. Meanwhile, Nvidia is expected to fill market demand during the product transition period and maintain overall supply chain pacing by shipping large quantities of Oberon Rubin racks and their "Ultra" versions.
AI capital spending by the five major tech giants is surging and is projected to reach 3.2% of US GDP in 2027, surpassing defense spending for the first time.
According to forecasts released by The Kobeissi Letter, the investment boom in AI is reshaping the US economy. It is projected that by 2027, the capital expenditures of the five major tech giants—Alphabet (Google's parent company), Amazon, Meta, Microsoft, and Oracle—on AI will climb to approximately 3.2% of US GDP. If this forecast materializes, it will be the first time in US history that annual AI capital expenditures have exceeded national defense spending (estimated at approximately 2.7% of GDP next year). Data shows that this year (2026), these five companies' AI capital expenditures are expected to exceed $800 billion, with the percentage of GDP rising sharply from 1.5% in 2025 to around 2.5%, approaching the estimated 2.7% of defense spending. By 2027, this figure is projected to surge further, reaching a record $1.1 trillion.
Apple to increase spending on its partnership with Broadcom; deal expected to exceed $30 billion.
According to Mars Finance, Apple will increase its spending on its partnership with Broadcom, an agreement expected to exceed $30 billion and produce more than 15 billion U.S.-made chips. (Cailian Press)
Omdia: China's semiconductor market will exceed $800 billion by 2026, with the memory chip market surging by 262.9%.
According to Omdia's latest report, "Semiconductor Applications Forecast Tool (AMFT) - China, Q2 2026," the Chinese semiconductor market is projected to grow at a significantly higher year-on-year rate of 92.9%, reaching $812.08 billion. This represents an overall upward revision of $265.6 billion, or 48.6%, compared to the previous report, "AMFT Shipments: China - Q4 2025 Update" (which projected a 31.26% growth rate for the Chinese semiconductor market, reaching $546.5 billion). Omdia's latest Q2 2026 data also shows that the Chinese semiconductor memory market is projected to grow at a significantly higher rate of 262.9%, reaching $449.6 billion. (Cailian Press)