The tech industry’s massive investment in artificial intelligence—about $400 billion this year for specialized chips and data centers—has raised concerns about the longevity of those chips. Critics argue that the industry’s optimism about how long these components will remain useful is misplaced. Cloud‑computing giants have traditionally expected their chips and servers to last roughly six years, but rapid advances—especially from manufacturers like Nvidia, which continually releases more powerful processors—are undermining that assumption.
Mihir Kshirsagar of Princeton University’s Center for Information Technology Policy points out that wear and tear combined with swift technological obsolescence make the six‑year horizon difficult to sustain. The accelerated rollout of new chips is eroding the market value of existing ones. For example, Nvidia announced the Rubin chip, projected to deliver 7.5 times the performance of its predecessor, less than a year after launching the Blackwell chip. According to Gil Luria of D.A. Davidson, such rapid upgrades cause chips to lose 85 to 90 percent of their market value within three to four years.
In addition to depreciation, AI processors are failing more frequently because they operate at high temperatures that can cause equipment to burn out. A recent Meta study reported an annual failure rate of 9 percent for its Llama AI model. These factors suggest that the realistic lifespan of AI chips may be only two or three years, far shorter than the industry’s four‑to‑six‑year estimate.
The implications of these overly optimistic assumptions are significant. Shortening depreciation timelines would immediately affect companies’ bottom lines and reduce profits, potentially rippling through an economy increasingly dependent on AI. Large players such as Amazon, Google and Microsoft may be insulated by diversified revenue streams, but AI‑focused firms like Oracle and CoreWeave—already heavily indebted—could struggle to raise capital if they must replace equipment more often.
To mitigate the risk, some companies are exploring options such as reselling older chips or repurposing them for less demanding tasks. Nevertheless, the situation remains precarious, especially for firms that have used chips as collateral for loans. The potential fallout from the AI boom’s reliance on rapidly obsolete hardware underscores the need for realistic assessments and careful planning in the tech industry.
Comments are closed for this story.