Article by Eric Worrell
“The entire industry is chasing the AI dragon” – a single project is worth more than 27 billion US dollars. Goodbye climate action virtue signaling.
Clusters of millions of GPUs, gigawatts of power—the scale of artificial intelligence defies logic
It’s not just one hyperbolic billionaire—the entire industry is chasing the AI dragon
Tobias Mann
Thursday, December 19, 2024 // 17:30 UTCComment As the artificial intelligence boom enters its third year, next year will see some truly massive computing projects. One of the largest plans revealed to date is xAI's plan to expand its Colossus AI supercomputer from an already impressive 100,000 GPUs to 1 million.
Such numbers seem illogical. Even if you could procure enough GPUs for this new colossusthe power and cooling — not to mention the capital — required to support it would be enormous.
At $30,000 to $40,000, adding another 900,000 GPUs would set xAI back US$27 to US$36 billion. Even with big volume discounts, it's still not going to be cheap regardless of whether they're deployed within a few years. Oh, and that doesn’t even take into account the cost of the building, cooling, and power infrastructure to support all these accelerators.
Speaking of power, depending on which generation of accelerators xAI plans to deploy, the GPU nodes alone will require approximately 1.2 to 1.5 gigawatts of power. That's larger than a typical nuclear reactor, and larger than a large nuclear reactor. Again, this is just for calculations.
…
Tech giants collectively cool down artificial intelligence craze The entire nuclear industry has undergone tremendous changesCloud providers are offering cash to rehabilitate decommissioned reactors and even put their data centers behind electric meters in AWS's new Cumulus data center complex.
…
Learn more: https://www.theregister.com/2024/12/19/scale_ai_defies_logic/
Think about it: just $27 to $36 billion one These new artificial intelligence data centers. The Register lists xAI, Meta (Facebook), AWS (Amazon), Oracle, and of course we already have Microsoft – all of these companies appear to be launching Gigascale AI projects in 2025, or have already launched theirs.
Of course, these are just American tech companies. India is also racing to build its own giga-scale AI, and there are signs that China is also getting into the AI game in a big way, although specific details on China's AI efforts are sparse.
India's Reliance uses Nvidia Blackwell AI GPU to build gigawatt data center
Anton Shilov
Posted on October 26, 2024But is India’s power grid ready for this?
Frontier is the world's most powerful supercomputer, consuming between 8 megawatts and 30 megawatts of power, depending on the workload. However, the upcoming artificial intelligence data center is expected to consume more. According to Reuters, India-based Reliance will build a 1 GWh (one gigawatt hour) artificial intelligence data center that will run Nvidia's Blackwell GPU.
Nvidia CEO Jensen Huang said: “In the future, India will become an exporter of artificial intelligence.” Reuters. “You have the basics — AI, data and AI infrastructure — and you have a large number of users.”
…
Learn more: https://www.tomshardware.com/tech-industry/artificial-intelligence/indias-reliance-builds-a-gigawatt-data-center-with-nvidia-blackwell-ai-gpus
Why is artificial intelligence quickly attracting so much attention? Artificial intelligence may be complex, but the motivations of tech companies, investors and government backers are easy to understand. The country or technology company that successfully creates the first general artificial intelligence has a real chance of becoming a permanently dominant force on Earth. – At least that’s what supporters think. The prize is the ultimate power, possession of the One Ring. Being the first to control a stronger-than-human AGI is about controlling the future—always knowing in advance what your opponents are planning, and always knowing the perfect counter-move in time to advance your interests.
In homage to Tolkien's fictional ring, artificial intelligence, or general artificial intelligence, can extend the lifespan of its owner and even grant medical immortality. One of the hottest uses of artificial intelligence right now is drug discovery and reverse engineering human DNA.
What will power this AI gold rush? The Register insists it’s going to be a boom for the nuclear industry, with continued interest in nuclear data center solutions, but Facebook’s AI project powered by fossil fuels shows delays in commissioning or restoration of nuclear power plants won’t slow these guys down . It’s only a matter of time before demand outstrips available nuclear and natural gas, and tech companies and governments start building new coal-fired power plants to power their AI data centers.
Despite U.S. technology sanctions, China especially has a home field advantage when it comes to artificial intelligence. China's housing and economic slowdown could create tens of gigawatts of surplus electricity that could be used to power large artificial intelligence projects without the need to build new power plants. The existing technology sanctions in the United States have little impact on China's promotion of artificial intelligence. Artificial intelligence is an area where quantity can replace quality – the gigawatt-level artificial intelligence data center itself is proof of this. If it takes a thousand traditional Chinese chips to match the computing power of one American cutting-edge artificial intelligence chip, then China is already big in second-tier chip manufacturing. China will have no problem producing the vast quantities of second-tier AI chips it needs to maintain its competitive edge.
There is still a huge unemployment problem among graduates in China. China’s commitment to the race for new technologies has attracted the best of China’s millions of desperate, unemployed tech graduates, making China a serious contender in the race for artificial intelligence, regardless of U.S. tech sanctions and any other disadvantages.
I must admit, The immediate explosive growth of AI data centers, each consuming more power than a major city, exceeded my expectations early in the game. I think it will take at least a few years for the AI gold rush to reach this level of activity.
In the short term, this sudden burst of activity may be more about nation-states scrambling to catch up to competitors and tech companies responding to shareholder pressure to stay in the game than genuine customer demand.
But the pressure will be on to justify these crazy investments, so there will be plenty of use for these AI data centers once they are plugged in, even if that use is just for technologists exploring how to turn new enterprise AI white elephants Hyped as a profit-generating business asset from artificial intelligence – a leap into the unknown could bring vast wealth and global political influence, or even bankruptcy and destruction, to countries, politicians, tech giants and their backers.
If you’re interested in learning more about AI, here’s a closer look at why AI data centers require gigawatts of power to operate.
Relevant
Learn more from Watts Up With That?
Subscribe to have the latest posts delivered to your email.
