CEOInterviews.AI
Start App
Kr Sridhar
Founder, Chairman & Chief Executive Officer, Bloom Energy

Can The World's Power Grids Keep Up With AI? Bloom Energy CEO KR Sridhar Explains

📅 Jun 26, 2026 CNBC-TV18 24 MIN 608 VIEWS 87 SEGMENTS · 3 SPEAKERS
Voices From The Valley | We're manufacturing intelligence now. But can our grids keep up? Bloom Energy's KR Sridhar joins Shereen Bhan to reimagine how the world powers the AI age. #VoicesFromTheValley #BloomEnergy #ArtificialIntelligence #EnergyInnovation #CleanEnergy #GridOfTheFuture #TechAndEnergy 🔴CNBC TV18 LIVE TV: https://youtube.com/live/P857H4ej-MQ SUBSCRIBE to our Channel: https://bit.ly/3nvEcxf ---------------------------------------------------------------------------------------------------------------------

What Kr Sridhar said

Written from the verified transcript and checked against it. Every figure links to the moment it was said.

KR Sridhar, founder and CEO of Bloom Energy, discussed how AI has catalyzed a power demand revolution. He argued that electricity, not heat, is now the primary energy need, and that onsite, clean power generation is essential because grids cannot support concentrated loads. He cited India's 2,000 terawatt hours of electricity consumption versus China's 10,000, and said renewables alone cannot meet data center needs due to land and copper constraints. Sridhar highlighted Bloom's fuel cells, which he said are twice as efficient as turbines at small scale, and noted a 55 MW delivery for Oracle in 55 days versus a promised 90. He claimed carbon capture will make Bloom the best option for clean power. On AI capex, he said the trajectory is upward despite potential bumps, and called electricity 'the invisible gold.'

Key takeaways

  1. Bloom delivered 55 MW for Oracle in 55 days, beating the promised 90-day timeline.
  2. India consumes 2,000 terawatt hours of electricity versus China's 10,000, and India needs to grow fivefold to catch up.
  3. Bloom's fuel cells achieve 50% plus efficiency at small scale, double that of engines or turbines.
  4. With carbon capture and sequestration, Bloom will be 'the best, period' for converting molecules to electricity.

Numbers and commitments

FigureWhat it refers toTypeAt
55 MW Capacity delivered for Oracle in 55 days commitment 14:11
90 days Promised delivery timeline for Oracle project timeline 14:11
2,000 terawatt hours India's annual electricity consumption metric 10:53
10,000 terawatts China's annual electricity consumption metric 10:53
5 times India's electricity demand growth needed to match China metric 10:53
50 plus percent Bloom's efficiency at small scale metric 16:46
90 plus percent Efficiency when using heat for cooling metric 17:37
800 volt DC Bloom's architecture voltage other 16:46
13 kilowatts Power consumption of old CPU racks metric 5:37
3,000 homes Power equivalent for future GPU racks metric 5:37

Chapters

  1. 0:00Digital electricity thesis
  2. 4:36AI power demand explosion
  3. 7:37Onsite power necessity
  4. 10:53Renewables limitations
  5. 14:11Execution speed and Oracle
  6. 15:50Efficiency advantages
  7. 17:52Carbon capture plans
  8. 20:13AI capex and market concerns
  9. 22:31Electricity as essential

Questions asked in this interview

9
  1. 3:25But did you anticipate this kind of power demand?
  2. 4:18When we talk about the surging power demand, the explosion that we're likely to see on account of all of these build-out that's happening linked to AI. What are we talking about?
  3. 7:11But, what does it mean to be able to deliver to that consistently?
  4. 10:31Is that not the smart alternative?
  5. 13:52You know, what you were able to do in 90 days you've now been able to bring down to 55 days. How does that happen?
  6. 15:30And what's the plan in terms of enhancing efficiency?
  7. 19:27How do you see the two?
  8. 22:22Does that make you smile when you hear that?
  9. 23:24So, electricity is AI as well, right?
Narrator 0:22 ↗
For 25 years, Bloom Energy has been chasing one big idea, reimagining how the world generates power. Today, that vision has made it one of Silicon Valley's most closely watched clean tech companies with a market capitalization of over 80 billion dollars and a stock that has surged more than 1,200% over the past year. As the AI boom fuels an unprecedented demand for power, Bloom Energy finds itself at the heart of the conversation on energy security, data centers, and the future of electricity. Its fuel cell technology is already powering some of the world's largest enterprises, and now the next phase of growth could be even bigger.
Interviewer 1:12 ↗
You know, we were last year in 2016, and so much has changed for Bloom Energy since then, and so much has changed for the world that we live in.
Kr Sridhar 1:21 ↗
Absolutely. So, what AI has done is catalyzed a revolution we saw coming 25 years ago. Our thesis was the world is going from a mechanical industrial age to a digital age. Everything is transforming to dependence on digital technology. And what does digital run on? Electricity. So, for the first time in human history, unlike heat energy being the predominant need for energy to keep yourself warm and to cook your food so you don't get poisoned, electricity becomes the main staple of energy. That was our thesis, and we said digital electricity for a digital world is going to be required. And what is digital electricity? It is not a gas turbine or an engine
Interviewer 2:16 ↗
Mhm.
Kr Sridhar 2:16 ↗
that first heats the fuel, creates a fire, combustion.
Interviewer 2:21 ↗
Yeah.
Kr Sridhar 2:22 ↗
That heat energy then runs a turbine, it becomes mechanical energy. You put big copper coils around it to convert the mechanical energy to electrical energy. Then you have freeways or highways and surface streets called transmission and distribution.
Interviewer 2:37 ↗
Yeah.
Kr Sridhar 2:37 ↗
So, before you get your electricity when you flip a switch, what was happening was fuel is being consumed to make heat,
Interviewer 2:45 ↗
Yeah.
Kr Sridhar 2:45 ↗
chemical energy to heat energy, to mechanical energy, to electrical energy, to highways, to distribution streets. We said that's nonsense. You need to be able to make clean electricity from a molecule to electricity with no middleman
Interviewer 3:02 ↗
Mhm.
Kr Sridhar 3:02 ↗
who are all going to take their share and make this whole equation inefficient.
Interviewer 3:07 ↗
Mhm.
Kr Sridhar 3:08 ↗
And do that in a clean way right where you consume the electricity.
Interviewer 3:11 ↗
Absolutely.
Kr Sridhar 3:12 ↗
the Bloom thesis
Interviewer 3:13 ↗
25 years ago.
Kr Sridhar 3:14 ↗
25 years ago because internet 1.0 had just happened. And we saw where the world was going ahead with Moore's law and this continuous learning and digitalization.
Interviewer 3:25 ↗
But did you anticipate this kind of power demand?
Kr Sridhar 3:29 ↗
We anticipated an extremely large energy demand. So, if you look at our mission statement, it was affordable power for everybody on the planet. Look, I'm an abundance guy. For some people, the glass is half full. For some people, the glass is half empty. For me, it's always overflowing because if I drink the half in the glass, somebody is going to pour me another half. So, so for me, it's about abundance, right? So, I always envision the world to be this place where there's abundance of electricity. But, the butt is even I did not anticipate the extreme rate at which this growth in electricity demand is going to happen, which is predominantly being triggered by AI. And we can talk about it, right?
Interviewer 4:18 ↗
Yes, we do want to talk about it. I mean, what you know, visualization is going to probably help people understand this better. When we talk about the surging power demand, the explosion that we're likely to see on account of all of these build-out that's happening linked to AI. What are we talking about?
Kr Sridhar 4:36 ↗
So, first and foremost, right? How exciting is this? For the first time in human history, we are manufacturing intelligence as a product, as a commodity. That's amazing what it's going to do for the wellness of people all over the world and uplift lives in the long term. That's pretty amazing. That's exciting. But, manufacturing intelligence, unlike a factory that manufactures chemicals or materials, what's the input? All the data that's out there and electricity. That's it. The machines are the GPU chips. The output is intelligence. And there is no higher value product electricity would have created than intelligence. So, the value is very high. So, but then, how does it actually happen? So, these computer racks that are in big data centers, they're about the size of a big Godrej safe.
Interviewer 5:37 ↗
Yeah.
Kr Sridhar 5:37 ↗
That everybody in India understands. Yeah. That everybody in India has, right? That safe with the CPU chips, which the previous generation of great IT data centers that we're all familiar with would consume about 13 kilowatt hours per rack, 13 kilowatt per rack. 13 kilowatts will be roughly maybe 20 Indian homes. Today's GPU chips in that same Godrej safe size unit would consume more than 300 homes worth of power. And the chips that are being built 2 years and 3 years from now will probably require 2,000 to 3,000 homes worth of power in that one Godrej safe and there are thousands of these Godrej safes in a single data center. How do you bring all this power into a data center? And how do you create that much electricity? And by the way, think about this. That factory is producing all this intelligence for a reason. It's going to make every other factory making everything else a lot more productive, a lot more efficient. It's going to allow us to do things in hospitals, in other places, in research centers that we could never do. They all need one or two of these Godrej safes sitting everywhere. They're sitting in every laboratory, in every building needing 3,000 homes worth of power coming into that. So, it's an explosion of demand and that's what's creating this need.
Interviewer 7:11 ↗
You know, thank you for giving us that visualization of the kind of demand explosion that we're likely to see and that is why people are excited about what Bloom Energy is doing. Now, talk to us about what it means to actually be able to cater to that kind of demand. It's been reflected in your numbers. I mean, we've seen record numbers as far as your financial performance is concerned. But, what does it mean to be able to deliver to that consistently?
Kr Sridhar 7:37 ↗
Great question. So, when you have that large a demand of power needed in one single place. The only way to produce that power is right where you need that power, not producing it somewhere else and bringing it over. Onsite power. That's very clear and it should be clear even from the industrial age. Very large petrochemical plants, very large cement factories, they produce their own power. They didn't depend on the grid because the grid only brings power to spread it out to lots of people, not for a single concentrated use. So, onsite power. Once you get to onsite power, countries like India should learn from what we are already facing in the US. The initial thought was we could use the grid. When you use the grid, very quickly you find out that the grid is not meant to support those kind of very large loads that can go up and down and create instability in the grid, so reliability goes down. If 95% of all the freeways and surface streets like we talked about transmission distribution is being used by a single customer, then everybody else should not pay the toll and if they did, the rates for all of these people are going up and that's going to create a backlash. And you're seeing that happen here in the US, right? So, we decided we need to do onsite power, but the knee-jerk reaction was let's bring the power plant that was hundreds of miles away and put that in our backyard. And it was okay as long as it was somebody else's backyard. If it is your backyard, you don't want the power plant in your backyard. It's noisy, it's heat, it's pollution, it's socks, knocks, particulates. And it is unsightly to see, which means your property values are going to go down and nobody wants it in their backyard. So, you need not just power on site, but you need that power to be clean. No air pollution. You want that power plant not to consume water and take away the water that you need for your community. You want it to be quiet. Bloom thought about these things 25 years ago and said that's what digital electricity needs to look like. So, we don't combust the gas. So, we provide no air pollution. We produce no air pollution. We don't consume any water in it. And because we are modular and fault tolerant, we don't require backup generators that are running on diesel. You're very familiar in Delhi about having backup generators and what that does to your air. We use none of that to provide reliable power. So, we were custom designed deliberately for applications like this for the modern age and finally now people are realizing that hence the demand.
Interviewer 10:31 ↗
So, let me push back on this because I think the general narrative is that yes, we get the fact that we don't have power data centers using thermal, but you know, renewable energy and why can't we do renewable energy? Is that not the smart alternative? And that is the way that India is looking at it. You're saying that's not sufficient and that in fact will not be enough at all.
Kr Sridhar 10:53 ↗
You know, I think we need all of the above. We are all of the above believers in the total electricity system as you design it. Should there be nuclear? The answer is yes. Should there be solar wind hydro? The answer is yes. However, let's again go back to some numbers. India today consumes roughly 2,000 terawatt hours of electricity. So, just remember the number 2,000 because we don't care about the terawatt hours. It's hard for people to fathom it. 2,000. The US with 1/4 the population consumes 4,000 terawatts of electricity. China, about the same size as India in terms of per capita, consumes 10,000 terawatts also. So, China consumes 10 times more power and is building at a rate lot faster than India. India's at 2,000. So, imagine just getting to even today's China's numbers 3 years 5 years from now. That's five times the amount of electricity that India produces today. That has to happen in a very short amount of time. No one technology, especially solar and wind, because of the amount of land use that you have, and the amount of land India has, and there are other needs for the land. There are other needs for the water. And remember, solar and wind are like rainwater. Nature gives it to you when it wants to, not all the time, which is not conducive for data. And even if you build those renewables, bringing that reliably to a data center in those large numbers we talked about earlier, there's not enough copper that India can import to be able to build those lines. So, it's an add solution. So, we should be adding renewables to power the general public. You need dedicated data centers running today on natural gas and in the future on green molecules such as hydrogen. But, that is farther away. Reliably today on natural gas. What is great about this? Friendly countries like Australia, like Canada, like the US, like many countries in the Middle East. So, there are multiple sources of natural gas that India can rely on to create energy security and build a data center economy because the speed at which AI is moving.
Interviewer 13:27 ↗
Mhm.
Kr Sridhar 13:28 ↗
If India misses this by a few years, the economic impact of keeping up is going to be pretty severe.
Interviewer 13:52 ↗
Well, you know, speed is of the essence and time is critical as you point out and that's the other thing that I want to talk to you about, which is very specific to the Bloom success story and that is your execution timeline. You know, what you were able to do in 90 days you've now been able to bring down to 55 days. How does that happen?
Kr Sridhar 14:11 ↗
So, I think you're referring to the example where we did 55 MW for Oracle. We had promised that we would deliver from the day we signed a contract in 90 days and we delivered in 55 days.
Interviewer 14:24 ↗
Mhm.
Kr Sridhar 14:24 ↗
And the reason is again, our digital electricity platform was built with a hope that one day it'll be like an appliance. You basically bring that home, plug it into the gas pipeline, plug it into your system and it's ready to run. That's when I would say we are really at our peak of speed, but we're doing very well today on that journey. So, it used to take us months.
Interviewer 14:52 ↗
Mhm.
Kr Sridhar 14:52 ↗
Now we have shortened that to days. Our goal is to shorten it to a day where it's so easy to just apply it and it's a LEGO block-like model where not only can you do it at speed,
Interviewer 15:03 ↗
Mhm.
Kr Sridhar 15:04 ↗
but you only pay as you grow. Typically, what will happen in a power plant is you'll build a very large power plant expecting a 5-year demand because it takes 5 years to build it.
Interviewer 15:12 ↗
Yeah.
Kr Sridhar 15:13 ↗
So, you would project out what you need and you'd build it and it'll sit idle that excess capacity will sit idle until such time that you need to use it. Why would you do that?
Interviewer 15:23 ↗
Mhm.
Kr Sridhar 15:23 ↗
With us, just buy the LEGO blocks you need today and we'll keep adding the LEGO blocks as you grow.

43 more exchanges in this transcript

Sign in free to read the rest of this interview. No card required.

Sign in to read the full transcript

Cite this transcript

APA, MLA, BibTeX
APA

Sridhar, K. (2026, June 26). Can The World's Power Grids Keep Up With AI? Bloom Energy CEO KR Sridhar Explains [Interview transcript]. CNBC-TV18. CEOInterviews.AI. https://ceointerviews.ai/interview/1148956/

MLA

Kr Sridhar. "Can The World's Power Grids Keep Up With AI? Bloom Energy CEO KR Sridhar Explains." CNBC-TV18, 26 Jun. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/1148956/.

BibTeX
@misc{sridhar2026_1148956,
  author       = {Kr Sridhar},
  title        = {Can The World's Power Grids Keep Up With AI? Bloom Energy CEO KR Sridhar Explains},
  howpublished = {Interview transcript, CNBC-TV18. CEOInterviews.AI},
  year         = {2026},
  month        = {jun},
  url          = {https://ceointerviews.ai/interview/1148956/},
  note         = {Speaker-attributed transcript with timestamps}
}