Goldman Sachs committed $500 million to establish a small-business AI institute and expand its "10,000 Small Businesses" program to support 10,000 additional businesses, partnering with Anthropic, Google, Meta, and OpenAI to help U.S. small businesses integrate AI into daily operations.
Goldman Sachs Research said at its Asia Leaders Forum in Hong Kong that the performance gap among Chinese AI models is narrowing, investor attention is shifting toward cost efficiency and the Harness layer, China's semiconductor manufacturing development is exceeding expectations, and Chinese open-weight AI models are being adopted globally.
At the Milken Institute Asia Summit in Singapore on October 8, 2026, Goldman Sachs' Asia Pacific (excluding Japan) head Kevin Sneader said new hires will manage AI agents from day one instead of waiting years to become managers, and reassigning existing mid-level managers will be a generational challenge across many fields. Co-panelist Sandra Peterson, operating partner at Clayton, Dubilier & Rice, said the traditional career ladder may disappear, while MAS Managing Director Chia Der Jiun said operational staff will need to transition into managers and supervisors of AI agents, and banks will need fewer fresh graduates for analytical and preparatory roles.
On September 27, Goldman Sachs' Carbonomics team released a 50-page report raising its global data center behind-the-meter (BTM) power capacity forecast from 40GW to 67GW, expecting gas turbines, reciprocating engines, and fuel cells to supply 28% of U.S. and 25% of global data center electricity demand, with gas turbine orders booked into 2031.
Goldman Sachs strategist Guillaume Soria said in a September 15 client note that while the 3-month momentum index rose 5%, the 12-month momentum index fell 6.7%, marking the widest single-day gap in five years. The bank's AI-stock index has dropped nearly 45% from its peak, the biggest slide since ChatGPT launched, and Goldman advised hedging AI exposure with put options, noting funds are rotating from semiconductors to software.
At the Goldman Sachs Communacopia + Technology Conference, Google Cloud CEO Thomas Kurian said the average AI server payback period is under two years, custom AI accelerator payback is half that of GPU solutions, its custom AI accelerator business is more than twice the size of the second-largest hyperscaler, and it offers 2.7x, 80%, and 30% cost-performance advantages in AI training, inference, and CPU respectively.
At the Goldman Sachs Communacopia + TMT Conference in September 2026, tech executives signaled that AI investment focus is shifting from compute buildout to compute usage. Nvidia CEO Jensen Huang said 70% growth is a supply ceiling and maintained a ~70% YoY growth expectation for CY27; SpaceX CFO Bret Johnsen disclosed a 2027 plan to raise AI compute to 5-10 gigawatts; Broadcom CEO Hock Tan detailed an economic model of 1GW compute corresponding to $30B ARR; Microsoft CFO Amy Hood and others discussed agentic AI and cybersecurity.
SpaceX CFO Bret Johnsen said at the Goldman Sachs conference that the company signed an AI compute hosting deal in early September 2026, expected to generate about $1.1 billion in monthly revenue starting December 1, or roughly $13.3 billion annualized. The company is confident of reaching $100 billion in annualized revenue by the end of 2026. He also revealed plans to launch the first batch of orbital computing satellites in 2027 and expects the 14th Starship flight later this month to generate revenue for the first time.
At the Goldman Sachs Communacopia + Technology Conference in San Francisco on Sept. 10, 2026, Nvidia CEO Jensen Huang said cybersecurity will become the next major AI application, as AI-accelerated automated programming increases code vulnerabilities and attack risks. He also responded to Nvidia's approximately $13 billion acquisition of Hugging Face, denying "circular investment" and stressing that Nvidia understands customer needs and only invests in deals with assured returns.
Goldman Sachs partner John Flood released a report after a week-long survey of Asian clients, noting that Asian investors are broadly skeptical of AI trades and that current caution exceeds fundamental support. He also said American investors may underestimate, while Asian investors may overestimate, the adoption speed of China's open-source AI models.