In August 2026, the financial and technological landscape underwent fundamental changes. Following a recent initiative by NVIDIA and its CEO Jensen Huang, the company's graphics cards have officially transitioned into a new asset class. Now, computing power is viewed by investors and analysts on par with traditional financial instruments, sparking a wave of speculation and unprecedented investment in artificial intelligence infrastructure.
Technological Asset: A New Market Standard
The transformation of NVIDIA chips into an investment asset has made demand for them a determining factor for global markets. Analysts note that the company is effectively acting as the 'Federal Reserve' for the AI industry, regulating access to computing resources. However, behind this optimistic scenario lies a serious threat warned about by a key advisor to the US administration.
David Sacks' Warning: The Risk of Overproduction
David Sacks, technology advisor to Donald Trump and a member of the President's Council of Advisors on Science and Technology, voiced a worrying forecast in his recent All-In podcast episode. In his view, the main threat to the large-scale development of AI lies not in a lack of demand, but in the risk of overproduction. Sacks pointed out that tech giants, striving to capture the market, are investing in data centers with dangerous scale, which could lead to a situation where the supply of computing resources significantly exceeds actual demand.
Echoes of the Dot-Com Crash: 'Dark Fiber' 2.0
Comparing current processes with history, Sacks drew a direct parallel to the excess of 'dark fiber' (unused fiber optic cables) following the dot-com crash in the early 2000s. He cautioned that a similar surplus of unused GPUs would become a serious problem for the entire industry. Infrastructure built with the expectation of high computing costs could go unutilized, leading to asset devaluation and financial losses for investors.
Economic Model and the Cost per Watt
The overproduction crisis will be particularly painful for those who built infrastructure expecting high computing costs. Previously, Elon Musk and other industry leaders cited a benchmark of $30–50 per watt of computing power. If the market becomes saturated with supply, these prices will collapse, making many projects unprofitable. Sacks emphasizes that the current euphoria surrounding NVIDIA may mask an impending liquidity crisis in the AI sector.
