Connect with us

Published

on

Elon Musk claims Tesla’s new AI supercluster will grow to over 500 MW, making it one of, if not the biggest in the world. At the same time, the CEO claims Tesla is achieving some record-breaking performance with its next-en AI chip.

A few months ago, we reported that Tesla was having issues building a new expansion at Gigafactory Texas to house a new giant supercomputer to train Tesla’s AI.

At the time, we heard that Tesla was aiming for a 100 MW cluster to be ready by August. Musk canceled other projects at Tesla to focus construction resources on the expansion.

Commenting on drone videos of the expansion, Musk said that it will grow to over 500 MW over the next 18 months:

Sizing for ~130MW of power & cooling this year, but will increase to >500MW over next 18 months or so. Aiming for about half Tesla AI hardware, half Nvidia/other. Play to win or don’t play at all.

We previously noted that it was strange that Tesla was internally referring to the project as a Dojo project, which refers to Tesla’s own supercomputing hardware, but sources were also told that the cluster would use Nvidia compute power.

Now, Musk confirmed that Tesla plans to use both its own hardware and Nvidia’s, as well as other suppliers.

However, things are getting a little unclear as Musk seems to also imply that Tesla will use some of its HW4 computers for the training clusters:

HW4 generally refers to Tesla’s in-car computer with an in-house designed chip, while Dojo is used for training, like this new cluster.

It’s unclear here if Musk is talking about using inference computing for training or just talking about Tesla’s overall planned computing power.

Electrek’s Take

Elon had mentioned at Tesla’s shareholders meeting that the company now had Nvidia-level AI chips, but the stock didn’t even move from that announcement as Nvidia became the most valuable company in the world.

I think Tesla’s AI effort is still not super credible for the market. That happens when you claim that you are about to achieve self-driving by the end of the year every year for the past 5 years.

At this point, we need to see Tesla make significant improvements to FSD with each new update. It sounded like this new cluster would help achieve that but Elon also recently said that Tesla was not compute-constrained for training right now, so it’s hard to really understand what is holding up improvements at this point.

FTC: We use income earning auto affiliate links. More.

Continue Reading

Environment

Europe wants to send data centers into space — study says it’s possible

Published

on

By

Europe wants to send data centers into space — study says it's possible

A data center.

Erik Isakson | DigitalVision | Getty Images

The rise of artificial intelligence is skyrocketing demand for data centers to keep pace with the growing tech sector — and pushing Europe to explore space options for digital storage, in a bid to reduce its need for energy-hungry facilities on the ground.

Advanced Space Cloud for European Net zero emission and Data sovereignty, a 16-month long study that explored the feasibility of launching data centers into orbit, has come to a “very encouraging” conclusion, according to Damien Dumestier, manager of the project.

The 2 million-euro ($2.1 million) ASCEND study, coordinated by Thales Alenia Space on behalf of the European Commission, claims that space-based data centers are technically, economically and environmentally feasible.

“The idea [is] to take off part of the energy demand for data centers and to send them in space in order to benefit from infinite energy, which is solar energy,” Dumestier told CNBC.

‘Data tsunami’

Data centers are essential for keeping pace with digitalization, but also require significant amounts of electricity and water to power and cool their servers. The total global electricity consumption from data centers could reach more than 1,000 terrawatt-hours in 2026 —that’s roughly equivalent to the electricity consumption of Japan, according to the International Energy Agency.

The industry is about to be hit with a “wave of data tsunami,” said Merima Dzanic, head of strategy and operations at the Danish Data Center Industry Association.

“AI data centers need something like three times more energy than a traditional data center and that is a problem not just on the energy side, but also the consumption side,” she told CNBC.

A “whole different approach to how we build, design and operate data centers,” is required, Dzanic added.

Demand for data centers has really only begun, says Thor Equities CEO

The facilities that the study explored launching into space would orbit at an altitude of around 1,400 kilometers (869.9 miles) — around three times the altitude of the International Space Station. Dumestier explained that ASCEND would aim to deploy 13 space data centre building blocks with a total capacity of 10 megawatts in 2036, in order to achieve the starting point for cloud service commercialization.

Each building block — with a surface area of 6,300 square meters — includes capacity for its own data center service and is launched within one space vehicle, he said.

In order to have a significant impact on the digital sector’s energy consumption, the objective is to deploy 1,300 building blocks by 2050 to achieve 1 gigawatt, according to Dumestier.

Eco launch

ASCEND’s goal was to explore the potential and comparative environmental impact of space-based data centers to aid Europe in becoming carbon-neutral by 2050.

The study found that, in order to significantly reduce CO2 emissions, a new type of launcher that is 10 times less emissive would need to be developed. ArianeGroup, one of the 12 companies participating in the study, is working to speed up the development of such reusable and eco-friendly launchers.

The target is to have the first eco-launcher ready by 2035 and then to allow for 15 years of deployment in order to have the huge capacity required to make the project feasible, said Dumestier.

Yet Dzanic warned the somewhat “fringe” idea of space-based data centers doesn’t fully solve the issue of sustainable energy usage. “It’s just one part of the puzzle,” she said.

Michael Winterson, managing director of the European Data Centre Association, acknowledges that a space data center would benefit from increased efficiency from solar power without the interruption of weather patterns — but the center would require significant amounts of rocket fuel to keep it in orbit.

Data centers are projected to account for more than 3% of Europe’s electricity demand by 2030.

Andrey Semenov | Istock | Getty Images

Winterson estimates that even a small 1 megawatt center in low earth orbit would need around 280,000 kilograms of rocket fuel per year at a cost of around $140 million in 2030 — a calculation based on a significant decrease in launch costs, which has yet to take place.

“There will be specialist services that will be suited to this idea, but it will in no way be a market replacement,” said Winterson.

“Applications that might be well served would be very specific, such as military/surveillance, broadcasting, telecommunications and financial trading services. All other services would not competitively run from space,” he added in emailed comments. 

Dzanic also signaled some skepticism around security risks, noting, “Space is being increasingly politicised and weaponized amongst the different countries. So obviously, there is a security implications on what type of data you send out there.”

World leader

ASCEND isn’t the only study looking into the potential for orbital data centers. Microsoft, which has previously trialed the use of a subsea data center that was positioned 117 feet deep on the seafloor, is collaborating with companies such as Loft Orbital to explore the challenges in executing AI and computing in space. Its work is crucial for innovation and to “lay the groundwork for future data management solutions in space,” a Microsoft spokesperson told CNBC.

ASCEND is one way through which the EU seeks to gain a competitive advantage within the AI ecosystem, where the bloc is currently lagging behind the U.S. and China, Dzanic said.

The EU is only now “starting to wake up and smell the coffee and go in with funding these projects,” she added.

The ASCEND researchers are in talks with the International Space Agency for the next phase which includes consolidating all of the data they have gathered and work on the development of a heavy lift launcher.

“We want to ensure data sovereignty for Europe, but this kind of project can benefit other countries,” said Dumestier. “We are pushing a lot because we can tell that it is a promising project. It could be a flagship for the Europe space development.”

Continue Reading

Environment

Crypto catches M&A frenzy as bitcoin miners chase AI boom

Published

on

By

Crypto catches M&A frenzy as bitcoin miners chase AI boom

Whinstone CEO Chad Harris takes CNBC on a tour of the largest bitcoin mine in North America.

Crypto companies are suddenly at the center of deal-making. The catalyst is artificial intelligence.

Bitcoin mining companies have expansive data centers, with access to fiber lines and large amounts of power across the U.S. They’re exactly the types of facilities needed for compute-intensive AI operations, which means their sites and technology are in high demand.

Meanwhile, miners need to diversify. Following the bitcoin halving in April, an event that happens about once every four years, the business of generating new tokens has become much less profitable. JPMorgan Chase analysts wrote in a report earlier this month that “some operators are feeling the financial pinch from the recent block reward halving, which cut industry revenues in half, and are actively exploring exit strategies.”

With the burgeoning AI industry in need of capacity and bitcoin miners in search of new ways to generate returns on their hefty capital investments, mergers, financings and partnerships are rapidly coming together.

On Tuesday, U.S. bitcoin miner Core Scientific announced an expanded deal with CoreWeave, an Nvidia-backed startup that’s one of the main providers of the chipmaker’s technology for running AI models. Core Scientific will deliver 70 megawatts of computing infrastructure to support CoreWeave’s operations.

Core Scientific said the deal will generate an additional $1.2 billion in revenue over 12 years, on top of an existing arrangement that is expected to bring in $3.5 billion. In total, the company plans to provide about 270 megawatts of infrastructure to CoreWeave by the second half of 2025, with the possibility of adding an additional 230 megawatts at other Core Scientific sites.

Earlier this month, CoreWeave offered to buy Core Scientific for $1.02 billion, not long after their initial agreement. Core Scientific rejected the bid. The company, which returned to the public market in January after going through bankruptcy, is currently worth about $1.8 billion.

“The world is changing, and many data centers built in the last 20 years are not suitable to support future computing requirements,” Core Scientific CEO Adam Sullivan said in Tuesday’s press release.

Bitcoin miner Hut 8 soars more than 15% after announcing $150 million AI investment: CNBC Crypto World

A day before that announcement, bitcoin mining group Hut 8 said it raised $150 million in debt from private equity firm Coatue to help it build out its data center portfolio for AI.

Hut 8, based in Miami, is one of many crypto mining companies pivoting to AI. The company said in its first-quarter earnings report last month that it had purchased its first batch of 1,000 Nvidia graphics processing units (GPUs) and secured a customer agreement with a venture-backed AI cloud platform. Hut 8 generates 6% of sales from AI, according to CoinShares.

“The broader market is beginning to appreciate the scarcity of high-quality power assets, and Hut 8 has built a deep pipeline of highly attractive expansion assets,” Robert Yin, a partner at Coatue, said in the financing announcement.

Hut 8 CEO Asher Genoot recently told CNBC his company “finalized commercial agreements for our new AI vertical under a GPU-as-a-service model, including a customer agreement which provides for fixed infrastructure payments plus revenue sharing.”

Bit Digital dumps tokens to buy GPUs

Bit Digital, a bitcoin miner that now derives an estimated 27% of its revenue from AI, said on Monday that it had entered into an agreement with a customer to supply 2,048 Nvidia GPUs over three years, doubling the number of processors it has providing the unspecified client.

To fulfill the contract, Bit Digital ordered 256 servers from Dell Technologies, and will soon deploy them at a data center in Iceland. The company said the contract is expected to generate $92 million in annual revenue. It’s paying for the GPUs, in part, by dumping some crypto.

“The Company intends to finance the deal with a mixture of cash and digital assets on the balance sheet,” Bit Digital said.

Bit Digital also entered a so-called sale-leaseback agreement for half of the new GPUs, “which will reduce the company’s capital outlay commensurately.” With the leaseback, another company owns those GPUs, and Bit Digital leases them back, generating revenue by providing the technology to customers.

People wait in line for t-shirts at a pop-up kiosk for the online brokerage Robinhood along Wall Street after the company went public with an IPO earlier in the day on July 29, 2021 in New York City.

Spencer Platt | Getty Images

While most of the recent crypto deals involve miners, there has been at least one big notable exception.

Earlier this month, trading platform Robinhood agreed to a deal to buy Bitstamp, a Luxembourg-based crypto exchange, for around $200 million in cash.

Bitstamp holds 50 active licenses and registrations across the globe, and is popular in Europe and Asia. The purchase helps Robinhood, a retail-focused trading app, bolster its crypto operation to better take on Binance and Coinbase.

The deal, due to close next year, comes as Robinhood faces regulatory challenges in the U.S. over its crypto dealings. In May, the company said it received a Wells notice for its crypto operations. The Securities and Exchange Commission has also sued Coinbase and Binance.

Robinhood had $4.7 billion in cash and equivalents at the end of the first quarter. Its stock is up 75% this year.

Don’t miss these insights from CNBC PRO

Jack Dorsey-backed startup taps into geothermal, hydro and solar power to run bitcoin mines across Africa

Continue Reading

Environment

Don’t call it Robotaxi, Porsche recalls, what tires are best for Tesla, and BTTF

Published

on

By

Don't call it Robotaxi, Porsche recalls, what tires are best for Tesla, and BTTF

On today’s episode, Mate Rimac beats Tesla to the Robotaxi punch, Porsche recalls 31,000 electric Taycan models, Fred helps find the best tires for your Tesla Model 3, and we go Back to the Future with an electric DeLorean restomod. All this and more, on YouTube’s ONLY show about EVs: Quick Charge!

Whether or not Mate Rimac calls his new, autonomous two-seater a Robotaxi, it sure likes an awful lot like a Robotaxi to me! Let us know how it looks to you in the comments.

Prefer listening to your podcasts? Audio-only versions of Quick Charge are now available on Apple PodcastsSpotifyTuneIn, and our RSS feed for Overcast and other podcast players.

New episodes of Quick Charge are recorded Monday through Thursday (that’s the plan, anyway). We’ll be posting bonus audio content there as well, so be sure to follow and subscribe so you don’t miss a minute of Electrek’s high-voltage daily news!

Got news? Let us know!
Drop us a line at tips@electrek.co. You can also rate us on Apple Podcasts and Spotify, or recommend us in Overcast to help more people discover the show!

FTC: We use income earning auto affiliate links. More.

Continue Reading

Trending