Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Chamath said a gigawatt data center used to cost ~$5B when he started his project but now runs closer to ~$100B fully loaded. The jump is all about silicon density with $NVDA Rubin-class racks approaching 600kW and every gigawatt carrying far more GPU and HBM content.

Shay Boloor

445,512 subscribers

114,314 views • 3 months ago •via X (Twitter)

Science & Technology Finance

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Chamath: “A gigawatt now costs $100 billion.” Chamath Palihapitiya said when he started his data center project, one gigawatt was just $4 to $5 billion. Now it’s 20x higher.

Chamath: “A gigawatt now costs $100 billion.” Chamath Palihapitiya said when he started his data center project, one gigawatt was just $4 to $5 billion. Now it’s 20x higher.

Small Cap Snipa

224,379 views • 3 months ago

Eric Schmidt says the real limit on AI is cash, and he puts one gigawatt of data center at roughly $50 billion: "A gigawatt data center costs something on the order of 50 billion with a B number." "How could these people in my industry be so crazy to raise all that money and build all the data centers? And the answer is they see the demand." "So absent an AI breakthrough algorithmically, which is a change to what is called the transformer architecture which is invented at Google, thank you, in 2017, we're sort of stuck with this cost structure." "And I think the real limit to AI is not energy. It's actually cash." "Because when you add up the cost of these, if you take 50, just round numbers, 50 billion a gigawatt, so 10 gigawatts is half a trillion dollars." "How many companies, countries, and so forth can hand an industry a trillion dollars of capital? Very, very few." Even at the low end of the published estimates, around $38 billion a gigawatt, his arithmetic still puts ten gigawatts in the hundreds of billions, so the conclusion holds. Very few balance sheets can carry that, which means most companies buying AI capacity over the next few years will be renting it from the handful that can. - Eric Schmidt (Eric Schmidt), Chair of the Special Competitive Studies Project and CEO of Relativity Space, at the SCSP AI+ Expo (@SCSPAI), May 2026.

Eric Schmidt says the real limit on AI is cash, and he puts one gigawatt of data center at roughly $50 billion: "A gigawatt data center costs something on the order of 50 billion with a B number." "How could these people in my industry be so crazy to raise all that money and build all the data centers? And the answer is they see the demand." "So absent an AI breakthrough algorithmically, which is a change to what is called the transformer architecture which is invented at Google, thank you, in 2017, we're sort of stuck with this cost structure." "And I think the real limit to AI is not energy. It's actually cash." "Because when you add up the cost of these, if you take 50, just round numbers, 50 billion a gigawatt, so 10 gigawatts is half a trillion dollars." "How many companies, countries, and so forth can hand an industry a trillion dollars of capital? Very, very few." Even at the low end of the published estimates, around $38 billion a gigawatt, his arithmetic still puts ten gigawatts in the hundreds of billions, so the conclusion holds. Very few balance sheets can carry that, which means most companies buying AI capacity over the next few years will be renting it from the handful that can. - Eric Schmidt (Eric Schmidt), Chair of the Special Competitive Studies Project and CEO of Relativity Space, at the SCSP AI+ Expo (@SCSPAI), May 2026.

Karl Mehta

42,986 views • 24 days ago

AI INFRASTRUCTURE/ENERGY DEMAND STOCKS TO KEEP ON WATCH FOR 2026 Ex-CEO $GOOGL Eric Schmidt on energy demand driven by AI: "We need energy and the numbers are profound. What we need from you is energy in all forms. People are planning 10 gigawatt data centers" $IREN: owns and develops vertically integrated data centers for hyperscalers $CIFR: large power footprint, 3.4 GW pipeline for AI workloads + HPC $WULF: high capacity, low cost data center expansion across Texas and New York $NBIS: AI compute and cloud services translating energy into revenue $CLSK: America’s Bitcoin Miner transitioning to AI infrastructure with a massive power footprint across 30+ data center sites $BITF: owns and operates power sites positioned for AI workloads, yet to sign with a hyperscaler but could come this year $GLXY: multi-gigawatt data center infrastructure at scale

AI INFRASTRUCTURE/ENERGY DEMAND STOCKS TO KEEP ON WATCH FOR 2026 Ex-CEO $GOOGL Eric Schmidt on energy demand driven by AI: "We need energy and the numbers are profound. What we need from you is energy in all forms. People are planning 10 gigawatt data centers" $IREN: owns and develops vertically integrated data centers for hyperscalers $CIFR: large power footprint, 3.4 GW pipeline for AI workloads + HPC $WULF: high capacity, low cost data center expansion across Texas and New York $NBIS: AI compute and cloud services translating energy into revenue $CLSK: America’s Bitcoin Miner transitioning to AI infrastructure with a massive power footprint across 30+ data center sites $BITF: owns and operates power sites positioned for AI workloads, yet to sign with a hyperscaler but could come this year $GLXY: multi-gigawatt data center infrastructure at scale

Small Cap Snipa

116,034 views • 8 months ago

SoftBank is creating the biggest data center. On Fox Business, Arm CEO and SoftBank Group International CEO Rene Haas was asked what he thinks about Meta selling compute, which could mean a compute supply glut. Haas disagreed, pointing to SpaceX's compute deal with Anthropic as proof of real demand, not oversupply. He said Meta's situation is similar, and that only a handful of companies, SpaceX, Meta, and SoftBank, have the financial muscle to compete at gigawatt scale, which is why SoftBank is moving in.

SoftBank is creating the biggest data center. On Fox Business, Arm CEO and SoftBank Group International CEO Rene Haas was asked what he thinks about Meta selling compute, which could mean a compute supply glut. Haas disagreed, pointing to SpaceX's compute deal with Anthropic as proof of real demand, not oversupply. He said Meta's situation is similar, and that only a handful of companies, SpaceX, Meta, and SoftBank, have the financial muscle to compete at gigawatt scale, which is why SoftBank is moving in.

Lark Davis

34,017 views • 2 months ago

Elon Musk said "500 gigawatt" on stage next to Jensen Huang and the Saudi minister, then corrected himself 4 times while the room laughed. Nobody wrote down the real number, which is 50. Phase 1 of the XAI-Humain data center is 50 megawatts, the 500 is the headline, and the gigawatt was the slip that showed where his head already lives. Then Jensen took the mic and described Humain out loud as a startup with "approximately 0 billion dollars in revenues" now building a 500-megawatt site for Elon, and called it off the charts twice in 30 seconds. AWS is coming in behind them at 100 megawatts with a gigawatt ambition, Nvidia is running Omniverse digital twins for the factories, plus supercomputers to simulate quantum computers because quantum error correction burns more compute than the quantum machine ever saves. Before the announcements, the Saudi minister told a story nobody clipped: Professor Omar Yaghi, first Saudi American to win a Nobel in chemistry, metal-organic frameworks with 33-nanometer pores that pull drinking water out of dry air, and a 500-by-1,000-nanometer nanorobot running CRISPR edits on sickle cell. Both started 20 years ago and AI just collapsed the timeline. That was the pitch, and the 500 megawatts was the invoice. A company worth 0 billion is building the biggest AI site in the Gulf, Elon said gigawatt, and nobody blinked.

Elon Musk said "500 gigawatt" on stage next to Jensen Huang and the Saudi minister, then corrected himself 4 times while the room laughed. Nobody wrote down the real number, which is 50. Phase 1 of the XAI-Humain data center is 50 megawatts, the 500 is the headline, and the gigawatt was the slip that showed where his head already lives. Then Jensen took the mic and described Humain out loud as a startup with "approximately 0 billion dollars in revenues" now building a 500-megawatt site for Elon, and called it off the charts twice in 30 seconds. AWS is coming in behind them at 100 megawatts with a gigawatt ambition, Nvidia is running Omniverse digital twins for the factories, plus supercomputers to simulate quantum computers because quantum error correction burns more compute than the quantum machine ever saves. Before the announcements, the Saudi minister told a story nobody clipped: Professor Omar Yaghi, first Saudi American to win a Nobel in chemistry, metal-organic frameworks with 33-nanometer pores that pull drinking water out of dry air, and a 500-by-1,000-nanometer nanorobot running CRISPR edits on sickle cell. Both started 20 years ago and AI just collapsed the timeline. That was the pitch, and the 500 megawatts was the invoice. A company worth 0 billion is building the biggest AI site in the Gulf, Elon said gigawatt, and nobody blinked.

West Lord

101,392 views • 7 days ago

He learned what had happened to a young woman, and when he realized the people responsible were getting away with it, something inside him changed. He wanted justice for her, but when the system failed to deliver it, he decided to take matters into his own hands.He began hunting down the men responsible, one by one. What started as a quest for justice soon became a dangerous pursuit of revenge and with every person he found, he got closer to crossing a line he could never uncross. But the real question is when the law fails someone, how far is too far to go for justice?

He learned what had happened to a young woman, and when he realized the people responsible were getting away with it, something inside him changed. He wanted justice for her, but when the system failed to deliver it, he decided to take matters into his own hands.He began hunting down the men responsible, one by one. What started as a quest for justice soon became a dangerous pursuit of revenge and with every person he found, he got closer to crossing a line he could never uncross. But the real question is when the law fails someone, how far is too far to go for justice?

Ola🧘

31,614 views • 14 days ago

ELON’S “IMPOSSIBLE” 12-MONTH PLAN COULD PUT SPACEX $SPCX AHEAD OF COREWEAVE $CRWV The bet is that vertical integration lets SpaceX build AI infrastructure faster than competitors: - Elon is planning to beat CoreWeave’s five-year plan in 12 months - SpaceX builds much of its infrastructure internally, with rocket scientists working on data centres - The company is all in on NVIDIA $NVDA and has already built roughly a gigawatt of data centre capacity in about a year and three quarters

ELON’S “IMPOSSIBLE” 12-MONTH PLAN COULD PUT SPACEX $SPCX AHEAD OF COREWEAVE $CRWV The bet is that vertical integration lets SpaceX build AI infrastructure faster than competitors: - Elon is planning to beat CoreWeave’s five-year plan in 12 months - SpaceX builds much of its infrastructure internally, with rocket scientists working on data centres - The company is all in on NVIDIA $NVDA and has already built roughly a gigawatt of data centre capacity in about a year and three quarters

Ark Invest Tracker

49,373 views • 10 days ago

We asked Sholto Douglas from Anthropic about the costs of RL (Reinforcement Learning) runs. "In Dario Amodei's essay, he said that RL runs cost only $1M back in December." "RL is a more naively parallelizable and scalable than pre-training." "With pre-training, you need everything in one big data center ideally. For RL, in theory, you could scale all over the world."

We asked Sholto Douglas from Anthropic about the costs of RL (Reinforcement Learning) runs. "In Dario Amodei's essay, he said that RL runs cost only $1M back in December." "RL is a more naively parallelizable and scalable than pre-training." "With pre-training, you need everything in one big data center ideally. For RL, in theory, you could scale all over the world."

TBPN

76,696 views • 1 year ago

🔥JAMIE DIMON: "MY BIGGEST FEAR WOULD BE CYBER. AI MAKES CYBER FAR WORSE." The JPMorgan CEO says his biggest fear is cyber and he has written about the threat for years. When asked about new Anthropic technology that could enable a catastrophic cyberattack with no defense. He said it "was built to defend, but it’s a weapon that could be used to attack." "cyber is huge and it’s getting worse every single year."

🔥JAMIE DIMON: "MY BIGGEST FEAR WOULD BE CYBER. AI MAKES CYBER FAR WORSE." The JPMorgan CEO says his biggest fear is cyber and he has written about the threat for years. When asked about new Anthropic technology that could enable a catastrophic cyberattack with no defense. He said it "was built to defend, but it’s a weapon that could be used to attack." "cyber is huge and it’s getting worse every single year."

Coin Bureau

51,093 views • 2 months ago

The market is watching xAI charge $50 billion per gigawatt and the rest of the neocloud sector run up is just getting started (Save this). According to Gavin Baker of Atreides Management, this is the most important number in AI infrastructure right now, xAI is monetizing compute at $50 billion per gigawatt on the Google deal, 2 to 3 times what any neocloud competitor charges. Google is paying $920 million per month for access to roughly 110,000 Nvidia GPUs through June 2029, and Anthropic is paying $1.25 billion per month for Colossus 1's 300 megawatts. Baker's point is simple that stop tracking rocket launches, stop tracking GPU orders, model gigawatt additions. At $50 billion per gigawatt, every new gigawatt that xAI energizes over the next 12 months is a revenue event that the market has not yet priced in. But this is not just an xAI story but rather why neocloud stocks are one of the most mispriced assets in the entire AI stack. Neoclouds charge $17 to $25 billion per gigawatt in contract value, a dramatic discount to xAI's pricing, but still an extraordinary business model when the underlying infrastructure costs $9 to $12 million per megawatt to operate and customers are signing 5-year locked contracts. H100 GPU-hours from neoclouds like Nebius at $2.95 per GPU-hour are 66% cheaper than hyperscaler rates, which is the structural reason enterprise AI teams are shifting spend to neoclouds at an accelerating pace. The neocloud market is projected to grow 69% annually through 2030 to reach nearly $180 billion and right now only a handful of public companies offer direct exposure to it. Nebius is the standout among the publicly traded neoclouds. It reported Q1 2026 AI cloud revenue of $399 million, an 841% increase year over year beating estimates, with its CEO stating that demand continues to exceed available capacity and customers are actively being turned away. Nebius commands a 20 to 25% revenue premium over peers thanks to its full-stack software offering, European sovereign positioning, and data residency advantages that physically prevent hyperscalers from competing for a large portion of its customer base. It has $49 billion in contracted backlog with Meta, Microsoft, and Nvidia meaning its revenue trajectory for the next three to five years is not a forecast, it is a schedule. The competitive moat is in power, permits, and speed exactly what xAI has proven is the true bottleneck. Jensen Huang said publicly that xAI deploys data centers faster than anyone else in the ecosystem, and Baker called out that this deployment speed advantage directly translates to monetization speed, every week of earlier energization at these pricing levels is worth hundreds of millions in revenue. Neoclouds with secured power, permits, and long-term customer contracts are not in a fair race against companies still waiting on grid connections and zoning approvals. The companies with the most locked in gigawatts coming online in 2026 and 2027 are about to have very good years.

The market is watching xAI charge $50 billion per gigawatt and the rest of the neocloud sector run up is just getting started (Save this). According to Gavin Baker of Atreides Management, this is the most important number in AI infrastructure right now, xAI is monetizing compute at $50 billion per gigawatt on the Google deal, 2 to 3 times what any neocloud competitor charges. Google is paying $920 million per month for access to roughly 110,000 Nvidia GPUs through June 2029, and Anthropic is paying $1.25 billion per month for Colossus 1's 300 megawatts. Baker's point is simple that stop tracking rocket launches, stop tracking GPU orders, model gigawatt additions. At $50 billion per gigawatt, every new gigawatt that xAI energizes over the next 12 months is a revenue event that the market has not yet priced in. But this is not just an xAI story but rather why neocloud stocks are one of the most mispriced assets in the entire AI stack. Neoclouds charge $17 to $25 billion per gigawatt in contract value, a dramatic discount to xAI's pricing, but still an extraordinary business model when the underlying infrastructure costs $9 to $12 million per megawatt to operate and customers are signing 5-year locked contracts. H100 GPU-hours from neoclouds like Nebius at $2.95 per GPU-hour are 66% cheaper than hyperscaler rates, which is the structural reason enterprise AI teams are shifting spend to neoclouds at an accelerating pace. The neocloud market is projected to grow 69% annually through 2030 to reach nearly $180 billion and right now only a handful of public companies offer direct exposure to it. Nebius is the standout among the publicly traded neoclouds. It reported Q1 2026 AI cloud revenue of $399 million, an 841% increase year over year beating estimates, with its CEO stating that demand continues to exceed available capacity and customers are actively being turned away. Nebius commands a 20 to 25% revenue premium over peers thanks to its full-stack software offering, European sovereign positioning, and data residency advantages that physically prevent hyperscalers from competing for a large portion of its customer base. It has $49 billion in contracted backlog with Meta, Microsoft, and Nvidia meaning its revenue trajectory for the next three to five years is not a forecast, it is a schedule. The competitive moat is in power, permits, and speed exactly what xAI has proven is the true bottleneck. Jensen Huang said publicly that xAI deploys data centers faster than anyone else in the ecosystem, and Baker called out that this deployment speed advantage directly translates to monetization speed, every week of earlier energization at these pricing levels is worth hundreds of millions in revenue. Neoclouds with secured power, permits, and long-term customer contracts are not in a fair race against companies still waiting on grid connections and zoning approvals. The companies with the most locked in gigawatts coming online in 2026 and 2027 are about to have very good years.

Milk Road AI

74,945 views • 2 months ago

The arms race of quant funds making vids about water cooling AI data centres is unreal. Hudson River Trading uses one in Norway cooled by fjord seawater. It’s piped into a former mountain-side mineral mine now hosting GPU racks. Heated water sent to a land-based salmon farm nearby and the farm produces 15,000 tonnes of salmon a year. Salmon are very sensitive to water temperature and the data center water (that came in cold from the fjord) is now heated to a salmon-friendly 20°C. When is the RenTech data centres water optimization video dropping?

The arms race of quant funds making vids about water cooling AI data centres is unreal. Hudson River Trading uses one in Norway cooled by fjord seawater. It’s piped into a former mountain-side mineral mine now hosting GPU racks. Heated water sent to a land-based salmon farm nearby and the farm produces 15,000 tonnes of salmon a year. Salmon are very sensitive to water temperature and the data center water (that came in cold from the fjord) is now heated to a salmon-friendly 20°C. When is the RenTech data centres water optimization video dropping?

Trung Phan

107,936 views • 3 months ago

Chamath says Silicon Valley lost what made it great and only cares about money now "Silicon Valley has completely lost what made it great. It was a place for weirdos, for people at the fringes, for people that really were idealistic. And it has become something that's far from that now. It is the equivalent of what Harvard used to be. It's a credentialing place." "When we came, we used to aspire to meet the Bill Campbells of the world, Mike Moritz, John Doerr. Those folks are gone. We used to aspire to work for Steve Jobs. He's not here." "And in that breach, all we have done is focused on making money in every single way that we can, hand over fist. And there's nothing wrong with that. Except now it’s at a point where we need to do a better job of acknowledging it and trying to do a little bit better.” “Every time we don't, and we're given a chance to be better, and we're not, the hatred builds, the animus builds. And that's what we're seeing now."

Chamath says Silicon Valley lost what made it great and only cares about money now "Silicon Valley has completely lost what made it great. It was a place for weirdos, for people at the fringes, for people that really were idealistic. And it has become something that's far from that now. It is the equivalent of what Harvard used to be. It's a credentialing place." "When we came, we used to aspire to meet the Bill Campbells of the world, Mike Moritz, John Doerr. Those folks are gone. We used to aspire to work for Steve Jobs. He's not here." "And in that breach, all we have done is focused on making money in every single way that we can, hand over fist. And there's nothing wrong with that. Except now it’s at a point where we need to do a better job of acknowledging it and trying to do a little bit better.” “Every time we don't, and we're given a chance to be better, and we're not, the hatred builds, the animus builds. And that's what we're seeing now."

dnap

3,830,200 views • 24 days ago

There was a time when Mogli ji used to mock RaGa and he used to ignore.. But now Rahul Gandhi Ji is giving him all back with interest. 🔥 Today he exposed the foreign policy—Hug-plomacy of Modi which has destroyed India’s relations with every nation…

There was a time when Mogli ji used to mock RaGa and he used to ignore.. But now Rahul Gandhi Ji is giving him all back with interest. 🔥 Today he exposed the foreign policy—Hug-plomacy of Modi which has destroyed India’s relations with every nation…

Shantanu

18,563 views • 1 month ago

Callum McNeill started his pro career after Hal Sutton said he would help him if Cal quit his job and fully committed. 4 years later, and now with a family, Cal was giving Q-school one more try. And he advanced to final stage. I spoke with him about Hal and Q-school. Links ⬇️

Callum McNeill started his pro career after Hal Sutton said he would help him if Cal quit his job and fully committed. 4 years later, and now with a family, Cal was giving Q-school one more try. And he advanced to final stage. I spoke with him about Hal and Q-school. Links ⬇️

Monday Q Info

26,501 views • 9 months ago

Germany currently has about 26 gigawatt hours of battery storage. Most of it sits in home batteries with only 4.3 gigawatt hours actually serving the grid. Building that storage already cost more than 10 billion euros and at national demand levels it only covers roughly 30 minutes of summer electricity usage. The winter months bring what's known as "Dunkeflaute" - cold dark windless periods and higher energy usage. To survive a 10-day winter lull (the minimum realistic requirement), Germany would need about 12,000 gigawatt hours of batteries, 470 times today's storage. Such a system would weigh roughly 60 million tons and would be made from vast quantities of lithium, nickel, graphite, copper, aluminum and steel, all requiring intensive mining. At current battery prices, the system would cost trillions of euros. And batteries last only 10 to 15 years, meaning the entire system would need constant replacement. The conclusion is unavoidable... Wind and solar require reliable backup power, renewables need oil, coal, gas and nuclear.

Germany currently has about 26 gigawatt hours of battery storage. Most of it sits in home batteries with only 4.3 gigawatt hours actually serving the grid. Building that storage already cost more than 10 billion euros and at national demand levels it only covers roughly 30 minutes of summer electricity usage. The winter months bring what's known as "Dunkeflaute" - cold dark windless periods and higher energy usage. To survive a 10-day winter lull (the minimum realistic requirement), Germany would need about 12,000 gigawatt hours of batteries, 470 times today's storage. Such a system would weigh roughly 60 million tons and would be made from vast quantities of lithium, nickel, graphite, copper, aluminum and steel, all requiring intensive mining. At current battery prices, the system would cost trillions of euros. And batteries last only 10 to 15 years, meaning the entire system would need constant replacement. The conclusion is unavoidable... Wind and solar require reliable backup power, renewables need oil, coal, gas and nuclear.

Electroverse

61,512 views • 3 months ago

When asked about whether he would support Defense Secretary Pete Hegseth’s potential run for president in 2028, President Trump said he “thinks he’s doing a fantastic job, but it’s very early to talk about that.” “Right now, all he loves is — and all he cares about is the military and taking care of the military,” Trump said.

When asked about whether he would support Defense Secretary Pete Hegseth’s potential run for president in 2028, President Trump said he “thinks he’s doing a fantastic job, but it’s very early to talk about that.” “Right now, all he loves is — and all he cares about is the military and taking care of the military,” Trump said.

CBS News

95,479 views • 14 days ago

#WATCH | Delhi: At the #IndiaAIImpactSummit2026, Chairman & Managing Director of Reliance Industries Limited, Mukesh Ambani, says, "Today, on behalf of the Reliance Group and Jio Intelligence, I want to make three announcements. Announcement one, Jio connected India to the internet era. Jio will now connect India to the intelligence era. We will deliver intelligence to every citizen, every sector of the economy, and every facet of social development and every service of government. Jio will do so with the same reliability, quality, scale, and extreme affordability that transformed connectivity. India cannot afford to rent intelligence. Therefore, we will reduce the cost of intelligence as dramatically as we did the cost of data... Announcement 2, Jio, together with Reliance, will invest 10 lakh crores over the next seven years, starting this year... Announcement 3, Jio Intelligence will build India's sovereign compute infrastructure through three bold initiatives. One, gigawatt-scale data centres. We already started construction on multi-gigawatt AI-ready data centres at Jamnagar. Over 120 megawatts will come online in the second half of 2026 this year and a clear path to gigawatt-scale compute for training and large scale inference. Two, our green energy advantage. We have an in-house energy advantage with up to 10 gigawatts of ready green power surplus anchored by solar in both Kutch and Andhra Pradesh. Three, a nationwide edge compute, an edge compute layer deeply integrated with Jio's network will make intelligence responsive, low latency, and affordable close to where Indians live, learn, and work..." (Source: India AI)

#WATCH | Delhi: At the #IndiaAIImpactSummit2026, Chairman & Managing Director of Reliance Industries Limited, Mukesh Ambani, says, "Today, on behalf of the Reliance Group and Jio Intelligence, I want to make three announcements. Announcement one, Jio connected India to the internet era. Jio will now connect India to the intelligence era. We will deliver intelligence to every citizen, every sector of the economy, and every facet of social development and every service of government. Jio will do so with the same reliability, quality, scale, and extreme affordability that transformed connectivity. India cannot afford to rent intelligence. Therefore, we will reduce the cost of intelligence as dramatically as we did the cost of data... Announcement 2, Jio, together with Reliance, will invest 10 lakh crores over the next seven years, starting this year... Announcement 3, Jio Intelligence will build India's sovereign compute infrastructure through three bold initiatives. One, gigawatt-scale data centres. We already started construction on multi-gigawatt AI-ready data centres at Jamnagar. Over 120 megawatts will come online in the second half of 2026 this year and a clear path to gigawatt-scale compute for training and large scale inference. Two, our green energy advantage. We have an in-house energy advantage with up to 10 gigawatts of ready green power surplus anchored by solar in both Kutch and Andhra Pradesh. Three, a nationwide edge compute, an edge compute layer deeply integrated with Jio's network will make intelligence responsive, low latency, and affordable close to where Indians live, learn, and work..." (Source: India AI)

ANI

82,266 views • 6 months ago

Oakley global president spoke about Kylian Mbappe in his recent podcast He said that when he went to close the contract with Kylian, Kylian’s only concern was about the project to help children with vision problems, check their sights at the clinics etc. He also said that when they do a campaign with children, the contract says that Kylian has to stay with the children for 30 minutes, but he ends up staying with them for 3 hours and more

Oakley global president spoke about Kylian Mbappe in his recent podcast He said that when he went to close the contract with Kylian, Kylian’s only concern was about the project to help children with vision problems, check their sights at the clinics etc. He also said that when they do a campaign with children, the contract says that Kylian has to stay with the children for 30 minutes, but he ends up staying with them for 3 hours and more

〽️

47,376 views • 1 year ago

The creator of High Bandwidth Memory (HBM) put a number on the AI build that should stop every infra investor cold. A cluster of a million GPUs runs at roughly 10-20% utilization (Save this). Kim Jung-ho spent thirty years building what feeds the GPU, and his claim is that the GPU is barely working. Here is what is actually happening. Every time a model generates output, the data has to be read out of memory, computed, and written back. The read and the write swallow almost the entire cycle. While that data moves, the GPU does nothing. It sits there, fully powered, fully paid for, waiting. By Kim's estimate the memory is doing only about 30 percent of the work it needs to do. The processor idles the rest. So a million installed GPUs run at 10 to 20 percent. You are not compute constrained. You are memory constrained, and the expensive part is standing around. Adding more GPUs does not fix this. It gives you more processors starving for the same data. Here is the part that decides the next decade. Memory can grow. When a cell cannot shrink any further, you stack it into a high-rise, layer on layer. A GPU cannot be stacked. It runs too hot and needs a cooler bolted to its back, so the one move that rescues memory is closed to the processor. The thing that can keep stacking compounds. The thing that cannot plateaus. The marginal dollar in an AI build now buys more by fixing the memory path than by bolting on another idle GPU. Which is why the companies that control memory bandwidth and supply are not suppliers to the AI trade. They are the AI trade.

The creator of High Bandwidth Memory (HBM) put a number on the AI build that should stop every infra investor cold. A cluster of a million GPUs runs at roughly 10-20% utilization (Save this). Kim Jung-ho spent thirty years building what feeds the GPU, and his claim is that the GPU is barely working. Here is what is actually happening. Every time a model generates output, the data has to be read out of memory, computed, and written back. The read and the write swallow almost the entire cycle. While that data moves, the GPU does nothing. It sits there, fully powered, fully paid for, waiting. By Kim's estimate the memory is doing only about 30 percent of the work it needs to do. The processor idles the rest. So a million installed GPUs run at 10 to 20 percent. You are not compute constrained. You are memory constrained, and the expensive part is standing around. Adding more GPUs does not fix this. It gives you more processors starving for the same data. Here is the part that decides the next decade. Memory can grow. When a cell cannot shrink any further, you stack it into a high-rise, layer on layer. A GPU cannot be stacked. It runs too hot and needs a cooler bolted to its back, so the one move that rescues memory is closed to the processor. The thing that can keep stacking compounds. The thing that cannot plateaus. The marginal dollar in an AI build now buys more by fixing the memory path than by bolting on another idle GPU. Which is why the companies that control memory bandwidth and supply are not suppliers to the AI trade. They are the AI trade.

Fireside Alpha

38,370 views • 2 months ago

$People in Utah have started asking us an important question, when you build these power facilities for AI data centers, can some of that energy be sold back into the local grid? The answer is yes. That completely changes the conversation. Instead of being seen as “the evil data center guys,” we can help power Utah’s grid while also building critical AI infrastructure for the future. And despite the headlines, we’re not building the world’s largest data center overnight. We’re starting with just 1.5 gigawatts with 1 gigawatt of net IT load a small fraction of the long-term proposal and proving the model first. And it’s not a 40,000 acre data center. Our project a full build out will span 10-13k acres 7,500 of which will be preserved. We want the people of Box Elder County to come see it themselves, the EPA air quality standards, the water permits, the noise compliance, and how the project actually operates before anything scales further.$

People in Utah have started asking us an important question, when you build these power facilities for AI data centers, can some of that energy be sold back into the local grid? The answer is yes. That completely changes the conversation. Instead of being seen as “the evil data center guys,” we can help power Utah’s grid while also building critical AI infrastructure for the future. And despite the headlines, we’re not building the world’s largest data center overnight. We’re starting with just 1.5 gigawatts with 1 gigawatt of net IT load a small fraction of the long-term proposal and proving the model first. And it’s not a 40,000 acre data center. Our project a full build out will span 10-13k acres 7,500 of which will be preserved. We want the people of Box Elder County to come see it themselves, the EPA air quality standards, the water permits, the noise compliance, and how the project actually operates before anything scales further.

Kevin O'Leary aka Mr. Wonderful

71,642 views • 3 months ago