Elon Musk’s xAI supercomputer has taken a major step forward with approval for 150 megawatts of power from the Tennessee Valley Authority (TVA).
This approval significantly boosts the facility’s energy supply, enabling it to run all 100,000 of its GPUs concurrently, a feat previously limited by available power.
However, this massive energy demand has raised concerns among local stakeholders regarding the impact on the region's power grid.
xAI expands power useWhen xAI first launched its supercomputer in July 2024, it required significantly more energy than was available. Initially, only 8MW of power was available at the site, which was insufficient to meet the demands of the AI data center.
Musk’s team improvised by using portable power stations to fill the gap. Over the summer, Memphis Light, Gas & Water (MLGW), a local utility company, upgraded the existing substation to provide 50MW of power, still far short of the requirements to fully operate the facility.
The xAI supercomputer, nicknamed the “Gigafactory of Compute,” is designed to support Musk’s artificial intelligence company. To run all of its 100,000 GPUs simultaneously, the data center needs an estimated 155MW of power, meaning the new approval for 150MW is just enough to get close to full capacity.
With approval for an additional 150MW, MLGW and TVA have worked to assure local residents that the increased demand from xAI will not negatively impact power reliability in the Memphis area. According to MLGW’s CEO Doug McGowen, the additional power needed for xAI’s operations is still within the utility’s peak load forecast, and measures are in place to buy more energy from TVA if necessary.
To meet these growing energy needs, many tech companies, including Amazon, Google, Microsoft, and Oracle, are investing in alternative energy sources, particularly nuclear power. However, it will take at least five years before nuclear energy solutions are ready for widespread deployment.
Until then, companies like xAI must rely on existing infrastructure to power their data centers, raising concerns about grid stability and the ability to keep up with increasing demands.
“We are alarmed that the TVA Board rubberstamped xAI’s request for power without studying the impact it will have on local communities,” says Southern Environmental Law Center senior attorney Amanda Garcia.
“Board members expressed concern about the impact large industrial energy users have on power bills across the Tennessee Valley. TVA should be prioritizing families over data centers like xAI," Garcia notes.
You might also likeWant to send a special message straight from the North Pole this year? AI video developer Synthesia has you covered with festive greetings with a dash of AI magic. You can get a digital Santa Claus speaking right to you or whoever you wish using Synthesia’s AI-powered video platform,
The personalized video messages stars a lifelike AI-generated Santa and even less tech-savvy well-wishers can use it easily. You can pick from an array of templates showing cozy living rooms adorned with Christmas trees and a comfy chair with Santa sitting and sharing your message. Synthesia’s virtual elves work their magic and your message is sent. Your heartfelt greeting is processed with Synthesia’s platform of advanced AI-powered text-to-speech and video generation technology. Santa is the latest of Synthesia's more than 230 pre-designed AI avatars, including custom creations.
Synthesia has the most comprehensive AI Santa message, but it's not alone. OpenAI debuted Santa Mode for ChatGPT last week, giving the AI chatbot a simulated version of Santa's voice for Advanced Voice Mode, which is described as "merry and bright."
Santa delivers a dose of Christmas spirit with striking realism and can speak 140 different languages. To maintain its family-friendly charm, Synthesia screens all user-submitted scripts to prevent any untoward or non-jolly messages. You can see my example below.
How to send a message from SantaIf you want to send a video from Santa, go to this website then:
1. Choose a Template: Visit Synthesia's Santa video generator page and select from festive templates.
2. Craft Your Message: Write a personalized message for your recipient. If you're unsure what to say, consider using an AI writing assistant for inspiration.
3. Submit and Generate: After finalizing your message, submit it through the platform. In just a few minutes, Synthesia's AI processes the text, generating a lifelike video featuring Santa delivering your message.
4. Share the Joy: Once the video is ready, it will be emailed directly to you. You can then share it with your loved ones, bringing a personalized touch to your holiday greetings.
You might also likeIf like me you enjoy nothing more than watching a great new movie on great home cinema kit, you've probably asked this question more than once about an exciting new release: when's it coming to one of the best streaming services – and which streamer is it coming to? And the answer is: it's complicated.
In a simpler world every movie would come to every streaming service. But of course we live in a world that's much messier, which is why we write separate guides to cover the best movies coming to Netflix, to Disney Plus, to Max, to Prime Video, to Paramount Plus and to Apple TV Plus.
So how do certain movies end up on certain streamers, and who decides who gets what and when? Let's find out.
How movies aren't sold Some streamers are also in the movie business: Netflix's Beverly Hills Cop: Axel F was a Netflix Original. (Image credit: Netflix )It's important to clarify what it means when we talk about films being released. Films aren't sold outright; in most cases, the companies that paid for them own them and like to keep it that way. What studios and their distributors do is sell licenses to show, sell or stream movies in a particular way, in a particular place, for a particular period of time.
When your local megaplex shows Wicked, it hasn't bought the movie; the theater chain has bought a licence to screen it in certain territories for a few weeks or months. Similarly, when Prime Video, Max or Paramount Plus offers a movie as part of your membership, the streamer has bought a licence that gives it the right to stream it – again, in certain territories for a certain period of time.
In addition to being licensed to different companies in different parts of the world, movie licenses are also sold in several stages. A studio movie will typically be shown in theaters first, made available digitally as pay-per-view or buy-to-own, and then made available on streaming.
Things are slightly different with TV shows, because of course there's not a theatrical release. But unless a show is made in-house by a streamer, it'll be licensed in much the same way that movies are.
Windows on the world Apple TV Plus released Wolfs in theaters, but only for one week. (Image credit: Sony Pictures / Apple)The movie business uses the term "release windows" to describe the different stages of a movie release campaign. So you have the theatrical release window, the pay-per-view window, the first streaming release window and so on.
The first release window is the original theatrical release. That used to be very long, typically three to six months or more, and then COVID came along. With theaters shut the studios either massively reduced the theatrical release window or abandoned it completely, turning to pay per view and streaming deals instead. And while COVID closures are thankfully a thing of the past, the much shorter theatrical windows remain.
Theatrical licences are traditionally exclusive: you won't usually be able to rent or buy a movie anywhere while it's still having its initial theatrical run. But the time between a movie hitting theaters and being available to rent, buy or stream is much shorter than it used to be.
Everything's accelerating Five Nights at Freddy's took an unusual approach with a simultaneous theatrical and streaming release. (Image credit: Blumhouse/Universal)According to Indiewire, the average theatrical window across all the big studio movies in 2023 was just 37 days. There were outliers such as Oppenheimer, which spent 122 days as a theater-only release. But that one was really unusual. The same studio's Super Mario Bros Movie was more typical (and more profitable), ending its theatrical run after 41 days.
Indiewire's analysis shows that Taylor Swift's Eras Tour movie ran for 60 days while Disney movies averaged 62 days, A24 movies 48 days, Paramount 42 days and Sony 35 days. The shortest theatrical window? Five Nights At Freddy's, which didn't have one: it was available on Peacock from day one.
FNAF was the third-biggest horror movie opening in cinema history but its release isn't likely to be widely copied. The consensus is that simultaneous releases are more likely to cannibalize box office sales – especially in genres such as horror, which teens like to see together in theaters – and by missing out on the word of mouth that can turn modest successes into big ones.
When the run is done Barbie followed a typical schedule: theaters first, then buy to own, then streaming (Image credit: Warner Bros. Pictures)Once the theatrical run is done, it's time for the release window that matters to streamers. That's called the Pay One Window, and it's when movies move from theaters to home viewing.
During the Pay One Window, the rights to show a movie are sold to two different markets: the PVOD/TVOD market (paid/transactional video on demand, aka pay-to-rent and buy-to-own) and the SVOD market, which is short for streaming video on demand.
SVOD means streamers such as Max, Prime, Netflix and Hulu; TVOD means rental and purchases on the likes of Apple TV (but not Apple TV+), Google Play, the Microsoft Store, Amazon (but not Included With Prime) and other digital storefronts.
It's always clearest if we use a real example, so let's take Barbie. Barbie's theatrical run started in July 2023 and stayed exclusive to theaters until September 2023, when it became available through TVOD platforms as a pay-to-rent and buy-to-own digital release. It then became available for Max subscribers to stream in December 2023.
When the Pay One Window expires, it's time to sell more licences. This next release window is known as the Pay Two Window, and it's when older movies are licensed, often to other streaming services than during the Pay One period. For example, Sony has a Pay One deal in place with Netflix but a Pay Two deal with all the Disney platforms.
Of the two windows, the Pay One Window is the more desirable: pent-up demand for big-name movies drives sales of new streaming subscriptions in a way that older movies don't. But Pay Two Window licences tend to be cheaper, and that means they enable streamers to expand their catalogs relatively cheaply.
Relationship status: it's complicated Joker Folie à Deux was released on Max two-months after its theatrical run. (Image credit: Warner Bros. Pictures)Traditionally, Pay One Window deals were for up to 18 months and were exclusive – so if a movie was going to Peacock then it wouldn't also go to Amazon Prime for at least a year and a bit. But modern movie markets are much messier, and as Variety reports, most studios now licence to multiple streamers either simultaneously or slightly staggered.
As if that wasn't complicated enough, some of the movie studios also own streamers and some of the streamers make movies. So for example Apple TV+ and Netflix make movies that hit theaters first but which were made with streaming in mind; Disney the company owns Disney the studio and Disney Plus the streamer, so Disney movies will move to Disney Plus after their theatrical run finishes. But Disney also owns 20th Century Fox and Searchlight, and those movies don't just end up with Disney+ or the Disney-owned Hulu; they are usually licensed to HBO and Max too.
Here's how Variety describes it. "Netflix gets major studio fare from Sony Pictures and additional prestige films from Sony Pictures Classics, while Prime Video offers rotating selections of Universal and Paramount titles, alongside Amazon’s MGM films after they’re done on MGM+. Disney’s streamers eventually get Sony titles, too, and Hulu is also where art-house distributor Neon’s films stream after their theatrical runs."
As we said, it's complicated – and it'll only get more so.
You might also likeMarvell Technology has unveiled a custom HBM compute architecture designed to increase the efficiency and performance of XPUs, a key component in the rapidly evolving cloud infrastructure landscape.
The new architecture, developed in collaboration with memory giants Micron, Samsung, and SK Hynix, aims to address limitations in traditional memory integration by offering tailored solutions for next-generation data center needs.
The architecture focuses on improving how XPUs - used in advanced AI and cloud computing systems - handle memory. By optimizing the interfaces between AI compute silicon dies and High Bandwidth Memory stacks, Marvell claims the technology reduces power consumption by up to 70% compared to standard HBM implementations.
Moving away from JEDECAdditionally, its redesign reportedly decreases silicon real estate requirements by as much as 25%, allowing cloud operators to expand compute capacity or include more memory. This could potentially allow XPUs to support up to 33% more HBM stacks, massively boosting memory density.
“The leading cloud data center operators have scaled with custom infrastructure. Enhancing XPUs by tailoring HBM for specific performance, power, and total cost of ownership is the latest step in a new paradigm in the way AI accelerators are designed and delivered,” Will Chu, Senior Vice President and General Manager of the Custom, Compute and Storage Group at Marvell said.
“We’re very grateful to work with leading memory designers to accelerate this revolution and, help cloud data center operators continue to scale their XPUs and infrastructure for the AI era.”
HBM plays a central role in XPUs, which use advanced packaging technology to integrate memory and processing power. Traditional architectures, however, limit scalability and energy efficiency.
Marvell’s new approach modifies the HBM stack itself and its integration, aiming to deliver better performance for less power and lower costs - key considerations for hyperscalers who are continually seeking to manage rising energy demands in data centers.
ServeTheHome’s Patrick Kennedy, who reported the news live from Marvell Analyst Day 2024, noted the cHBM (custom HBM) is not a JEDEC solution and so will not be standard off the shelf HBM.
“Moving memory away from JEDEC standards and into customization for hyperscalers is a monumental move in the industry,” he writes. “This shows Marvell has some big hyperscale XPU wins since this type of customization in the memory space does not happen for small orders.”
The collaboration with leading memory makers reflects a broader trend in the industry toward highly customized hardware.
“Increased memory capacity and bandwidth will help cloud operators efficiently scale their infrastructure for the AI era,” said Raj Narasimhan, senior vice president and general manager of Micron’s Compute and Networking Business Unit.
“Strategic collaborations focused on power efficiency, such as the one we have with Marvell, will build on Micron’s industry-leading HBM power specs, and provide hyperscalers with a robust platform to deliver the capabilities and optimal performance required to scale AI.”
More from TechRadar ProCISA has urged ‘highly targeted’ individuals in the US Government or in senior political positions to immediately review and implement strict security measures with their mobile devices.
This comes after multiple major US telecoms firms were hit by security breaches earlier in 2024, most likely originating from state-sponsored Chinese hackers.
Recent reports also confirmed that the group behind the attack, Salt Typhoon, had not yet been fully eradicated, and still lurks inside US telecoms networks.
Lockdown modeAs part of its advisory, CISA also released a best practice guide for mobile communications, which includes instructions on using end-to-end encryption, password managers, Fast Identity Online (a phishing resistance authentication), and advice on how to migrate away from SMS based multi factor authentication.
CISA recommends iPhone users turn on ‘Lockdown mode’, which strictly limits certain apps and makes many features unavailable in order to reduce the attack surface that attackers could potentially exploit.
The scale of this attack makes it a serious concern for any high-ranking officials, with major networks like Verizon, AT&T, and Lumen Technologies found to have threat actors with access inside their systems.
“Highly targeted individuals should assume that all communications between mobile devices — including government and personal devices — and internet services are at risk of interception or manipulation,” CISA said.
Communications remain an attractive target for foreign actors, as breaches can result in significant disruptions and sensitive data loss.
US communications in particular have come under attack this year, especially against government and political targets in the run up to the US election last month. A ‘hack and leak’ campaign hit president-elect Donald Trump on the campaign trail in an attempt to compromise the campaign, which resulted in two indictments for the suspected individuals.
Via The Record
You might also like