Content Insider #958 – Control
By Andy Marken – andy@markencom.com
“If the rock you stay on starts to roll, jump clean. Or you’ll go with it and be squashed. Only a fool stays aboard.” “Kyoami, Ran”, Kurosawa Productions, 1985
O.K., we get it.
We need AI to solve really complex problems like curing diseases, improving the quality of life for 20 percent of the people around the world living in extreme poverty (less than $2 a day), feeding 2.3B people facing moderate/severe food insecurity, stemming/reversing global climate change (even if you think it’s BS), figuring out how support/deal with 1.5B people 65 + by 2050 and eliminating the possibility of another pandemic.
A few folks are using it today to address these problems however, most of its use is directed at doing your work so you don’t have to and enabling your boss to reduce his/her headcount, making cute/insulting pictures and stealing stuff from other people.
Damn, in addition to being so cool, so cute and so beneficial, AI promises to have so much potential for … everyone.
The AI techies say all we have to do is keep giving it more and more data, sit back and relax and life will be good … great.
Well, they also need more than just data.
A big data center hosts 10K to 100K servers housing millions of GPUs.
More, Bigger – AI GPUs may be small, but AI data centers are big and require a lot of resources – water and power – to keep them running at peak performance. And every country/company wants to lead the parade.
Ramping Up – All of the usual cloud and data collectors – Google, Meta, Oracle, Amazon, Tencent, Baidu and Microsoft – are busy investing in bigger, better, faster AI0 data centers. Emerging hypercenter firms are adding large facilities to keep pace with AI demand.
Actually, they’re referred to as hyperscale data centers designed from the ground up to handle the growing AI workload.
No, they aren’t your hole-in-the-wall data centers, we’re talking big … we mean really big!
Expansion – Even before an AI data center is completed the company is busy planning for the next great phase of their processing capabilities. Texas may be short on a few things like water, electricity but it’s big on what hyperscale AI facilities need – land – but it’s not alone. States and countries everywhere are making room for newer, bigger, badder AI centers like Meta’s Hyperion facility.
They need AI data centers … big data centers and lots of them, everywhere.
The global AI data center market is expected to grow from $18B last year to $94B by 204003.
That kind of growth brings tears to Jensen Huang’s eyes (Nvidia CEO). He saw the value of his GPU company surpass $5T because his chips (and a few other players around the globe) are at the very heart of AI.
A big data center hosts 10K to 100K servers housing millions of GPUs.
Overwhelming – Ever since the early 1940s when the first electronic computer appeared on the screen the systems have gotten faster, more reliable in processing/crunching data. AI made it easier and faster to crunch/process more data to produce even more data.
Don’t worry, that won’t happen because the data centers are designed to quickly and accurately handle anything you can throw at them and keep consuming, processing, digesting and delivering/storing all that data, no matter how brilliant or dumb it is … they don’t judge.
Every day more than 2.5 quintillion bytes of data are created. That figure rose to more than 463 EB last year (1 quintillion bytes equals 1EB, 1000EB equals 1ZB).
That includes your streaming entertainment, emails, documents, social media (written, video) posts, gaming, web browsing, synthetic content and file sharing–anything/everything digital.
Of course, to slice/dice/move all that data around those data centers require power.
Yeah, you have to share your home/office power with those power-hungry data centers.
The small data center your organization has tucked into the corner of the building (5,000-20,000 sq ft) only requires 1-5MW of power while your home only uses about 30 kWh a day.
Cascading Effect – As AI permeates every corner of business and our lives it continues to tap, use and produce more data while constantly sucking up more resources to keep itself going and growing.
Of course, that didn’t take into consideration the hellacious amount of power AI data centers require.
AI-heavy racks require 30-50+kW per rack with today’s Nvidia GPUs consuming 1,200W and Nvidia’s next-generation AI chips will need up to 15,000W … each.
By 2030, the world’s data centers will be consuming about 220-327GW.
More, Bigger – AI GPUs may be small, but AI data centers are big and require a lot of resources – water and power – to keep them running at peak performance. And every country/company wants to lead the parade.
Yes, that’s a little more power than 225M homes require; but what’s a little brownout or blackout now and then as long as your videos and hallucinations entertain and inform/convince.
Slow Shift – While coal and natural gas are presently the leading sources of power generation, countries and companies are moving as rapidly as possible to replace the generation source to more environmentally friendly sources – solar, wind, hydropower and a newer, more safe nuclear approach.
Countries and companies around the world are doing their darndest to keep pace with and get ahead of the power demand.
Unfortunately, the electrical grid is a little fragile and under significant strain right now because it’s old and AI data centers are pushing for more and more reliable power.
In some areas, there’s a seven-year wait to get connected to the electrical grid.
But don’t worry, some government officials are aggressively pushing to harvest more natural resources to meet the demand … sending guys (they’re usually guys) back into the dirty coal mines even if the dust corrupts the lungs and spews stuff into the atmosphere.
They’re all for drilling more holes in the earth to suck up more oil to burn to make more electricity for the population and processors.
Companies and countries around the globe are making substantial investments to modernize and diversify their energy supplies and taking a longer-range approach to power generation.
Many areas are developing microgrids and installing battery storage to ensure future reliability.
Investments – Major investments are being made in wind, solar and hydro power production capabilities even as companies develop smaller, more powerful and safer nuclear energy production facilities.
There’s also a strong effort/commitment to diversify their energy supply.
Nothing happens overnight but countries and companies are focused on solving their energy requirements by solar, wind and hydro power generation. In addition, a new generation of nuclear power facilities is being developed, including Small Modular Reactors (SMRs) and Generation IV which focus on improved safety, efficiency, and cost-effectiveness.
The new facilities feature a smaller footprint with attention to climate change while quickly and economically meeting the energy demands of the new AI data centers.
While China’s energy production still relies heavily on coal, it is rapidly investing in clean energy production, particularly solar and wind power.
China is the world leader in solar power with total capacity that equals the rest of the world … combined.
Their wind power capacity (it really does work…honest) exceeds the combined capacity of the next 12 leading countries.
China continues to be the largest hydropower clean energy producer even though many scientists are concerned about the stability of their latest project – three gorges dam – the largest such construction in the world (181 M tall, 2335M across the Yangtze River).
Capable of holding 39.3B cubic meters of water, it is so large that it slightly slowed the Earth’s rotation.
While it displaced more than a million people, scientists are concerned about increased landslides, structural integrity and the impact of seismic activity.
Time will tell if the bold move was too bold but at least they are proving that they are not burdened by what worked before in meeting their power requirements for their population and AI tomorrow.
The only other thing AI data centers need is lots and lots of water.
AI data centers generate massive amounts of heat and require advanced cooling systems.
Operating 24 hours a day, the facilities need continuous chilled water loops and server-level liquid cooling and immersive cooling.
The high-density computing systems generate 60-80 KW of heat and can use millions of gallons a day to protect the high-performance processors.
A large hyperscale data center can consume 5M gallons of water per day while a 1MW facility can use up to 25.5M liters annually.
Or, to bring the matter home, a data center using 550,000 gallons per day – about as much as an Olympic-size swimming pool – is equivalent to providing water to more than 20,000+ households.
Meta’s facility outside Atlanta, GA, needs reduced water pressure in surrounding communities to a trickle of muddy, unsanitary water.
Microsoft’s central Mexico facility stretched area home water supply outages to weeks.
Seldom Discussed – When companies talk about the benefits of AI to a community and/or the world, they seldom mention the fact that the facilities require millions of gallons of water to cool the facilities and processors which must be taken from local supplies. The consumption has a major impact on surrounding communities.
Similar issues have arisen in areas across the US as well as in Brazil, Britian, India, Malasia, the Netherlands, Singapore, Spain, United Arab Emirates, India, China, Australia and …
Okay, we acknowledge that many of the facility owners/managers are working to shrink their environmental footprint.
Many are building wind farm facilities and installing their own SMR nuclear power facilities as well as developing water systems that recycle/reprocess as much water as possible to reduce the strains on an already frail system.
But as they say … it takes time.
You or AI – Many parts of the world are facing major and devastating shortages of potable water which has to be shared with 9B people and AI GPUs … tough choice. |
Despite the issues, states and countries are rushing to get their share of the AI data center boom.
States across the US are creating tax breaks, carving out large industrial zones and committing millions/billions to attract new and bigger AI data facilities.
The same is true of countries in the EU, Asia, the Middle East, Africa and well … everywhere.
Renewable energy is getting cheaper – faster.
But water is a finite thing.
It’s estimated that one in four people (2.1B) on the planet lack access to safe drinking water and in many cases that number has risen to more than 50 percent of the folks.
But at least your boss has implemented AI across the organization so the firm can do more work with fewer people.
Okay, he probably doesn’t have a clue as to what it is but gawd, it’s gonna be good.
And you can still crank out your cute YouTube, TikTok videos faster and with bigger, better – and often more embarrassing images.
Maybe it’s just best if we all agree with Kyoami who said in RAN, “In a mad world only the mad are sane.”
At least that way we can explain why AI in everything is better than thinking and sweating through a project while being able to say, “I did that.”
Naw, they’re right … AI is a better answer.
Andy Marken – andy@markencom.com – is an author of more than 900 articles on management, marketing, communications, industry trends in media & entertainment, consumer electronics, software and applications. He is an internationally recognized marketing/communications consultant with a broad range of technical and industry expertise, especially in storage, storage management and film/video production fields. He also has an extended range of relationships with business, industry trade press, online media and industry analysts/consultants.