Asger Risom
The latest version of Elon Musk’s Grok AI model cost nearly $500 million to train and consumed enough energy, water, and resources to match the needs of an entire small city.
The latest version of Elon Musk’s Grok AI model cost nearly $500 million to train and consumed enough energy, water, and resources to match the needs of an entire small city.
Training Grok came at a massive cost
Developed by Musk’s xAI, Grok 4 cost $490 million to train using supercomputers running nonstop for months—equivalent to 246 million computing hours.
Enough electricity to power a city
Training the model required 310 gigawatt-hours of electricity—enough to supply 4,000 people for a year. The power demand reflects the growing resource burden of cutting-edge AI.
Emissions equal to a Boeing’s three-year flight
The carbon footprint of training Grok 4 reached 154,000 tons of CO₂, comparable to what a commercial Boeing airplane emits over three years of flights.
300 Olympic pools’ worth of water
Cooling the data centers and running power systems took 754 million liters of water—enough to fill around 300 Olympic-sized swimming pools.
Colossus supercomputer powered the effort
xAI relied on its Memphis-based “Colossus” supercomputer to train Grok. The facility is equipped with Nvidia H100 GPUs and is powered by natural gas generators.
Grok 4 Fast reduces resource use
To address sustainability concerns, xAI released Grok 4 Fast—a streamlined model that uses 40% fewer thinking tokens and costs 98% less to operate, while maintaining strong performance.
Real-time web search capability added
Grok 4 Fast can now browse the web and the X social platform in real time, thanks to integrated “agentic search” features, boosting its ability to deliver up-to-date answers.
Outperforms rivals in testing
In independent benchmarks, Grok 4 Fast showed higher efficiency and intelligence density than models like OpenAI’s o3 and Google’s Gemini 2.5 Pro.
Available to all X users
xAI has rolled out Grok 4 Fast on both Android and iOS, aiming to bring high-level AI features to a broader audience while reducing operational costs.
Criticism over bias and misinformation
Grok has come under fire for spreading misinformation, promoting harmful right-wing narratives, aligning with Elon Musk’s own statements, and amplifying content that critics say threatens minority groups and public trust.
The AI sustainability and safety dilemma
Grok’s development highlights two challenges: environmental strain and ethical safety. As AI scales up, its creators face growing pressure to minimize harm—both ecological and societal.
This article is made and published by Asger Risom, which may have used AI in the preparation
