Hyperscalers Eclipse Iconic Megaprojects With AI Infrastructure Boom
Hyperscalers' AI capex has surpassed costs of historic US megaprojects, revealing massive private investment scale in the AI boom overlooked by mainstream media.
Hyperscalers have outspent most iconic US megaprojects on AI with unprecedented speed and scale.
Recent figures from Microsoft, Amazon, Google, and Meta earnings show over $200 billion directed at AI data centers and accelerators for 2024 and 2025. This exceeds inflation-adjusted costs of the Manhattan Project ($32 billion) and Apollo program ($257 billion) per data from Fin Moorhouse (https://twitter.com/finmoorhouse/status/2044933442236776794) and federal archives.
Epoch AI's tracking indicates corresponding compute growth at rates of 4x per year, consistent with scaling laws in their 2023 report. Goldman Sachs July 2024 analysis projects this infrastructure wave will drive 10-20% increases to current US electricity usage by 2030 (https://www.goldmansachs.com/insights/articles/ai-data-centers-and-power).
Such rapid private sector allocation in AI contrasts with the extended timelines of 20th century public megaprojects, demonstrating evolving patterns in technology development funding and resource commitment.
AXIOM: Private hyperscaler spending on AI is outpacing public megaprojects by orders of magnitude in velocity, pointing to an infrastructure singularity where resource demands may soon constrain further exponential growth.
Sources (3)
- [1]Hyperscalers have already outspent most famous US megaprojects(https://twitter.com/finmoorhouse/status/2044933442236776794)
- [2]AI Data Center and Power Consumption Forecast(https://www.goldmansachs.com/insights/articles/ai-data-centers-and-power)
- [3]Compute Trends Across Eras of AI(https://epochai.org/blog/compute-trends)