Ampere Computing CEO Renee James joins Yahoo Finance Live to discuss how her company is working to reduce the energy consumption of artificial intelligence.
video transcript
[AUDIO LOGO]
– Could the AI arms race lead to a new energy crisis? According to MIT, AI consumes more energy than traditional computing, and training one model could fill 100 US homes in a year. Consumes more electricity than eaves. Now Ampere Computing CEO Renee James is on board.
Lenny, thank you so much for being here. And folks, we’re building more energy efficient chips with complete transparency. But first of all, let’s talk about the scope of the problem. Because data centers consume a lot of energy even without this full AI capability. Is the industry doing enough to address it?
– First of all, thank you for letting me talk about something important and important to Ampere, of course, because it was founded on the principles of building sustainable computing. Aside from the demand for AI, 30% of the data centers around the world that are actually driving the growth of the cloud we are all experiencing are in the US, already using his 2% of all our energy I’m using.
So when we started the company five years ago, we said, look, I, my team and I have been developing microprocessors for a long time, but we wanted something that could deliver more power. I felt it was time to build a . — Higher performance with lower power consumption. And that’s exactly what the company is about. A highly efficient and sustainable microprocessor to drive cloud growth. Cloud growth will be further accelerated by the use of AI. That’s what we are for.
– Indeed, further down your business, if and to the extent that demand continues, what will your company do to ensure that artificial intelligence advances in a more sustainable and environmentally friendly manner? do you do? conscious way. How many of the customers you’ve spoken to are prioritizing environmental considerations and sustainable practices as they transition to AI?
– I think it’s a great question. The good news is that the cost takes sustainability into consideration, as one of the big drivers of cost is of course your electricity bill. So, of course, the more sustainable the computing solution, the lower the overall total cost of ownership due to the reduction in power and cooling.
Therefore, our customers consider this very important. This is one of the main calculations they focus on. The situation is real-time as every data center operator knows the power grid is either telling them they can’t guarantee more power or asking them to lighten the power load. As we look to add more AI, we need to find a more sustainable solution as these queries and searches only require 20+ X’s.
So in our industry there’s been a lot of talk about what it’s like for normal people who don’t make a living out of this, but what are these people saying? This is called dense computing, and it improves utilization of existing data centers built using existing capacity that you already use. In other words, the more efficient microprocessors can be used in each server rack in the data center, the higher the density and the better utilization.
And that’s exactly what Ampere is about, but now it’s impacting our clients’ ESG goals more broadly and what our users want long-term about our ability to sustain growth. I believe.
– Well, Renee, you guys are a private company. So I don’t know how much you can teach me. But what can you tell us about the demand for the chips so far?
– We are private, so we just want to say thank you for your question. Some know we are in the process of registering. We look forward to opening up the market for new growth companies at some point in the future. It was good. I think there are bright spots for that.
You see, when I went through several different phases of computing, and when the PC first made progress when I was new to the industry, what really inspired it was driving its growth. The bottom line is that there are always big new uses for computing to do. step. Our current phase is the cloud and AI on the cloud. It was the media then.
Initially, a special purpose was required. I mean, a lot of new hardware and his PC to be able to create videos, what PCs can do is play videos very efficiently. So when we think about AI, we think the same way. To be able to train absolutely accurate AI models, we need to build accelerated compute modules. That’s what Nvidia does very well.
But AI is evolving and it’s inference, if you will. Think of it as media and like playback. It’s in your browser. This is something that is built into every application to help the app. This should run most efficiently on your computer’s regular processor.
Therefore, from our point of view, building sustainable, low-power, ultra-high-performance computing is, as has been the history of our industry and microprocessors, a substitute for power for performance. Use the. Therefore, more power means more performance. We are pioneering low power, high performance, not low power, low performance, which is the history of low power consumption.
We believe this will enable data center operators and cloud (hyperscale cloud) operators to efficiently introduce artificial intelligence into everything and all services. And that’s a really exciting part of our company, and our sustainability theme has become even more important over the past five years.
– And Renee, please come soon. What should people think about what your chip will replace? In other words, Nvidia does not appear to be a direct competitor. Who are your direct competitors?
– So, what we really want is the growth of the cloud for data processing and inference processing, and think of the traditional CPU companies that have really lived in that space for the last 2-40 years. I’m here. Like Intel and AMD.
Nvidia is a great partner for us, doing fast computing and some really exciting stuff. I really appreciate those who are doing well because this business is very difficult. And yes, Nvidia has been pursuing it for 30 years, so it’s exciting for them.
– Even before the recent wave of attention around Renee, AI and chips, and what role they play in generative AI and other applications, there is still a large geopolitical landscape surrounding chips today. The impact remains. How many of them will continue to be the driver of AI’s next growth?
– Well, for the traditional computing part of the business, thankfully I don’t think it was too much of a burden for us. I believe long-term AI will be able to do the non-training part, and much of its general use will be able to run on processors that are not currently subject to export restrictions. . .
don’t know. I think it’s a big deal – there’s a lot of discussion going on. We still have a lot to discover about what we should and shouldn’t do. It’s always difficult. This happens all the time at the beginning of a new phase of computing. It’s not clear what should be held firmly or broadly deployed. And I think we are currently in discussions with the Department of Commerce about that.
– Renee, we had a really interesting conversation here today. Thank you very much for your precious time. Ampere Computing CEO Lenny James said:thanks so much
-thank you.
