Is this a trillion dollar AI bubble?
Tech firms face a bill likely to run into trillions of dollars on advanced microchips and data centres or risk falling behind their competitors – but the risk of getting caught in a bubble is growing
WHERE IS THE MONEY GOING?
Chips and hardware 60%
Power generation transmission, cooling 25%
Land, materials, site development 15%
BIG TECH AI INVESTMENTS
Amazon
Meta
Alphabet
Microsoft
Oracle
55GW of compute power required by 2030 at about $50bn/GW
2018
2019
2020
2021
2022
2023
2024
2025
2026
2030
Late 2022: ChatGPT released
Est. $250bn
$417bn
$490bn
$2.8tn
Global investment expected to reach $6.7 trillion
Global required investment
$6.7tn
Forecast productivity gains
$6.6tn
Including cloud computing, software, data centres, power
WARNING SIGNS
Debt-fuelled
Big Tech can no longer rely on profits to fund AI
Over-investment risk: stranded assets – rapidly depreciating hardware
Share issuance
No evidence so far of increase in IPOs
Over-valued share prices
Over-optimistic profit forecasts
Over-building data centres
Circular financing
INVESTMENT
Nvidia
$100m investment in data centres
SPENDING
OpenAI
Buys chips
$$$$$$
AI-generated image
Sources: McKinsey, Citi Research, Goldman Sachs
© GRAPHIC NEWS