expenditure

Goldman Sachs Research released a viewpoint on September 18, stating that AI is transitioning from the experimental phase to the implementation phase, with the rise of consumer-grade AI agents marking the emergence of the platform layer. At the Communacopia + Technology Conference held in San Francisco, most companies showcased cases from experimentation to implementation. Goldman Sachs expects that by 2027, capital expenditures for U.S. mega-cap companies will reach $1.4 trillion, exceeding Wall Street consensus.Goldman Sachs analyst Eric Sheridan stated that consumer-grade AI agents are shifting from conversational relationships to action-oriented tasks. If consumers overcome trust and security issues, they could execute complex tasks such as purchasing tickets and booking hotels. The monetization of such agents in the mass market will be similar to search, achieved through advertising and subscriptions. AI is evolving from the infrastructure layer to the platform layer and application layer, with declining token unit pricing and increased utility being key drivers of mass adoption.During the conference, concerns about AI risks became a major topic, but Goldman Sachs believes this will not slow down infrastructure construction, as demand for computing power still exceeds supply and most projects have already been contracted. Supply chain constraints such as memory chips, electricity, and land may pose resistance, but the capital expenditure cycle is expected to remain high through 2027.

UBS expects global artificial intelligence capital expenditure to rise to approximately $1.4 trillion by 2027, primarily due to a significant increase in memory costs.Memory-related spending is projected to increase from $71 billion in 2025 to $923 billion in 2027, accounting for about 60% of the increase in AI capital expenditure.

According to the New York Times, OpenAI expects its computing power expenditure to reach $750 billion by 2030.

Venture capital firm Social Graph VC recently published an article on market entry stating that data center capital expenditures will exceed $1 trillion by 2026, approximately double the spending of the four major hyperscale cloud providers in 2025. Jensen Huang expects it to reach $3 to $4 trillion annually by the end of this decade, potentially hitting $10 trillion by 2031. AI investments account for about 0.9% of global GDP, rising to approximately 1.4% by 2028.The article states that Fable 5 and GPT-5.6 each have training power consumption below 2GW, with costs around $120 billion. Each generation of models can recoup costs on their own, while laboratory losses stem from the next-generation models that are expanding training scales by 10 times. Computing cannot be procured at a single price; H100 rental rates are about $1.95/hour in the market, while hyperscale cloud rates are $8-9/hour, with hourly rates fluctuating by 137% over the year. There is a financing gap of $1.5 trillion within the approximately $3 trillion global data center expenditures expected by 2028.The article also mentions that CME plans to launch cash-settled H100 and B200 monthly lease futures on NYMEX on October 5, 2026, pending regulatory review, with settlement referencing the Silicon Data Index; ICE has also announced GPU futures based on the Ornn Index.

Mizuho analyst Vijay Rakesh stated that NVIDIA's second-quarter earnings exceeding expectations has become a consensus, with the market focusing on AI spending outlook. Rakesh expects NVIDIA's adjusted earnings per share to be $2.08, with revenue at $92 billion, and the actual figures may be higher. He maintains a "Outperform" rating and a target price of $300.

The State Council Information Office held a series of themed press conferences titled "Starting Off the 14th Five-Year Plan." Officials from the Ministry of Finance introduced the role of this year's proactive fiscal policy and outlined the key focus areas for policy efforts in the second half of the year.Vice Minister of Finance Liao Min stated that since the beginning of this year, the Ministry of Finance has firmly implemented a more proactive fiscal policy:This year's fiscal expenditure budget arrangement has exceeded 30 trillion yuan for the first time.The scale of new government bonds has reached 11.89 trillion yuan, the largest in history.Central government transfers to local governments have exceeded 10 trillion yuan for the fourth consecutive year, reaching 10.42 trillion yuan.In the second half of the year, fiscal policy will continue to focus on three areas for effective results: accelerating the use of funds, increasing efforts to expand domestic demand, and strengthening fiscal reform management.Regarding incremental policies, in the second half of the year, based on the macroeconomic operation situation, the Ministry of Finance will timely plan and introduce practical and effective incremental policies to provide strong support for achieving a qualitative improvement and reasonable growth in the economy.

According to Cointelegraph, a recent report by BlocksBridge Consulting shows that the capital expenditure for AI and high-performance computing (HPC) businesses of 9 publicly listed Bitcoin mining companies will reach $5.11 billion in the first half of 2026, while the related revenue during the same period is only $340 million, resulting in an input-output ratio of about 15:1, highlighting the significant upfront investment required for transformation. Among a sample of 15 mining companies and AI data center companies, the total capital expenditure for the reporting period in 2026 reached $30.7 billion, an increase of 42.6% compared to the entire year of 2025.Despite the investment far exceeding revenue, AI and HPC revenue is still accelerating. The 9 mining companies saw a quarter-on-quarter growth of 52% to $206 million in the second quarter, led by Core Scientific, TeraWulf, and Bitdeer. BlocksBridge pointed out that although mining companies have advantages in electricity and land, converting to AI-ready capacity still requires substantial investments in substations, cooling systems, network equipment, and GPUs.

According to Cointelegraph, U.S. companies' spending in the field of artificial intelligence is rapidly increasing. Data shows that the top 1% of companies are currently spending a record $7,400 per employee per month on AI-related expenses, which is 600 times higher than the median company spending level.Related statistics reflect a significant divergence in AI investment between leading companies and average companies, with top firms significantly increasing their investment in AI applications and infrastructure at a rate far above the industry average.

According to reports from South Korean media, SK Group Chairman Chey Tae-won received a salary of 1.75 billion won in the first half of this year.According to the Financial Supervisory Service, SK Corporation disclosed this information in its semi-annual report released that day. The chairman's salary level in the first half of this year is the same as the same period last year, and he did not receive any bonuses apart from his salary. The chairman's younger brother, SK Group Vice Chairman Chey Jae-won, received a salary of 1.5 billion won during the same period. SK Corporation President Chang Yong-ho received a total of 1.8 billion won, including 1 billion won in salary and 800 million won in bonuses.In addition, SK Supex Pursuit Council Chairman Yoon Bong-yong was disclosed to have received a total of 7.715 billion won, which includes 500 million won in salary, 800 million won in bonuses, and 6.415 billion won in compensation from 12,103 performance stock units (PSUs) granted three years ago. SK Hynix's sales from Nvidia in the first half of this year exceeded 17 trillion won, accounting for about 13% of the company's total sales.The average salary of SK Hynix employees reached 144 million won (approximately $104,000) in the first half of this year, an increase of 23% year-on-year. SK Hynix significantly increased its capital expenditures and research and development (R&D) investments in the first half of this year. Cash expenditures for acquiring tangible assets exceeded 18 trillion won, an increase of over 70% year-on-year. The number of small shareholders of SK Hynix has grown more than fivefold in a year, surpassing 3.46 million.

Tencent Holdings released its Q2 2026 financial report, with revenue of 204.79 billion yuan, a year-on-year increase of 11%, slightly exceeding market expectations; Non-IFRS net profit was 68.4 billion yuan, a year-on-year increase of 9%; profit attributable to equity holders of the company was 56.022 billion yuan, a year-on-year increase of 0.7%. Capital expenditure was 52.78 billion yuan, a year-on-year increase of 176%, far exceeding the market expectation of 32.14 billion yuan, mainly used for AI infrastructure investment. The free cash flow for the quarter was negative 13.8 billion yuan; excluding prepayments for AI computing power purchases, free cash flow was 37.6 billion yuan.In terms of business, value-added services revenue was 98.4 billion yuan, a year-on-year increase of 8%, with domestic game revenue at 47.3 billion yuan, a year-on-year increase of 17%, significantly contributed by "Delta Action," "Valorant," and "Lock Kingdom: World"; international market game revenue was 18.6 billion yuan, down 0.8% year-on-year due to exchange rate effects, but up 4% year-on-year when calculated at constant exchange rates. Marketing services revenue was 43.6 billion yuan, a year-on-year increase of 22%, making it the fastest-growing segment; financial technology and enterprise services revenue was 60.3 billion yuan, a year-on-year increase of 9%.Regarding AI products, the report specifically mentioned that WorkBuddy achieved rapid user growth and a healthy user retention rate, with strong user willingness to pay. Its market-leading position is attributed to its strong harness engineering capabilities, a rich selection of models, and an industry-leading skill library. The official version of Hy3 saw a more than 68-fold increase in usage in the week following its launch compared to the previous generation, consistently ranking among the top three globally in terms of OpenRouter token consumption, with plans for a larger parameter scale Hy4 to be released soon. The AI agent Xiaowei within WeChat has initiated a small-scale gray testing. Excluding the contribution from new AI products, Non-IFRS operating profit increased by 19% year-on-year to 86.1 billion yuan. By the end of the quarter, the company's net cash decreased from 146.9 billion yuan at the end of March to 58.2 billion yuan.

According to Jinshi reports, Fed's Schmid stated that the Personal Consumption Expenditures Price Index is the best way to measure inflation.

According to Jinshi reports, SpaceX's Chief Financial Officer stated that capital expenditures in the third and fourth quarters will be similar to the levels in the second quarter.

According to Jinshi reports, SpaceX's total capital expenditure in Q2 was $18.369 billion, with a net loss of $541 million.

According to Jinshi reports, based on the latest AI server industry research by TrendForce, under the backdrop of rising demand for AI infrastructure, the total capital expenditure of the world's nine major cloud service providers (CSP) is expected to grow by approximately 90% year-on-year by 2026.Recently, ultra-large CSPs and Tier-2 data center operators have shown a significant increase in their willingness to purchase NVIDIA's integrated AI server solutions, and it is expected that the annual growth rate of AI server shipments in 2026 will be adjusted from the original 28% to nearly 31%.

According to Jinshi reports, the actual personal consumption expenditure month-on-month rate in the United States for June was 0.4%, with the previous value revised from 0.3% to 0.4%.