In his 2026 AI Trends Outlook report, Sequoia Capital partner David Cahn points out that the market is in a transitional phase where investment sentiment is cooling, but the fundamentals of AI continue to advance. Entering 2026, AI development will proceed along two paths simultaneously: on one hand, the construction of data centers and the AGI timeline will be delayed; on the other hand, the practical applications of AI will continue to expand, especially the growth momentum of AI startups, which has not stopped despite the market cooling.
Despite continued expansion of investment, the reality of AI revenue still falls short.
Cahn stated that large tech companies' capital expenditure needs for AI will remain strong well into 2026. Google and Meta continue to increase their investments, while Microsoft and Amazon have made slight adjustments, but their overall long-term strategic direction remains unchanged. However, the supply chain is beginning to show signs of fatigue, with some suppliers maintaining a cautious attitude towards the sustainability of end-user demand.
From a revenue perspective, the current scale of AI industry revenue is only in the tens of billions of dollars, a stark contrast to the trillions of dollars invested in data centers and energy infrastructure over the next five years. At the application level, products that have truly achieved scale are concentrated in programming tools and ChatGPT-like models. Most large enterprises face integration difficulties and limited success when implementing AI on their own.
With demand surging, the supply chain becomes the first stress test.
Cahn believes that the structural pressures facing the first timeline in 2026 will come from the reality that computing power demand continues to rise, but the supply chain cannot keep up with the pace of expansion.
In advanced process technologies, TSMC and ASML hold a highly concentrated, near-monopolistic position in the industry, and their capacity expansion pace is relatively cautious, which cannot be easily accelerated by external parties. Given the continued rise in demand for AI chips, this expansion pace may translate into substantial capacity constraints by 2026.
The shortage of manpower and other resources in the later stages of construction have gradually created a risk of delays.
In addition to the wafer fabrication process, Cahn also pointed out the potential risks in the later stages of data center construction. As construction enters its final phase, critical industrial equipment such as generators and cooling systems, as well as experienced technical personnel, could all become obstacles affecting the progress.
Since most AI companies actually share the same supply chain, any delay in any single link could force a postponement of the overall development timeline. Sequoia points out that 2026 will be a crucial year for the market to comprehensively examine whether these industrial supply and labor bottlenecks are truly ready.
The construction cycle has entered the showdown phase, and the AGI timeline has been postponed accordingly.
Cahn points out that an AI data center takes an average of about two years to complete. With a large number of projects starting in 2024 and construction investment beginning to be reflected in economic data in 2025, 2026 will officially enter the results verification phase.
If large cloud providers begin stockpiling AI chips but fail to deploy them in time, it will be seen by the market as a significant signal that delays are officially emerging. At the same time, another delay timeline comes from the revision of AGI development expectations. The previously widely circulated "AGI in 2027" prediction has been gradually pushed back since mid-2025, and the new mainstream view tends to believe that AGI may not be realized until the 2030s at the earliest.
AI adoption continues to advance, with startups becoming the main beneficiaries.
Compared to the uncertainties surrounding hardware development and AGI timelines, Cahn's trend toward AI adoption itself is relatively clear. Even as market enthusiasm cools, the adoption of AI in practical applications continues to accelerate, with outstanding startups rapidly growing from zero to hundreds of millions of dollars in revenue. It is anticipated that starting in 2026, startups with revenues exceeding one billion dollars will gradually emerge.
Cahn also observed that top AI startups demonstrate high operational efficiency, with some companies achieving annual revenue exceeding one million US dollars per employee. They also extensively utilize AI agents to optimize internal processes, creating a self-reinforcing operational flywheel. In contrast, large enterprises generally face organizational and integration resistance when implementing AI on their own, resulting in limited implementation effectiveness. This, in turn, creates more development space for startups that focus on a single scenario and have mature products.
This article, "Sequoia Capital's 2026 AI Trend Outlook: Although Computing Power Construction Slows Down, the Growth Momentum of AI Startups Remains Unabated," first appeared on ABMedia, a ABMedia .
