AI Performance Relying More on Power Flow: Infineon CEO

Watch AI Performance Relying More on Power Flow: Infineon CEO - Bloomberg on Dailymotion

Transcript

00:00It's all about efficiencies and we need to make sure that the power flow is not becoming the

00:06bottleneck for the race in terms of AI compute performance. And we are doing there very well

00:12after tripling our revenue last year to the previous year. Now we are targeting to double

00:20this revenue again this year and this is capacity limited, not demand limited. So a really nice

00:28growth story across many, many products, many technologies, wide band cap technologies,

00:35silicon technologies, modules, controllers, analog parts. So a great growth story for Infini.

00:44And then which partners are you focusing on in particular for this as well? Because it's

00:50obviously a build out that requires a very large ecosystem. Yeah, absolutely. It's a very large

00:55ecosystem. And you have many of the key players, of course, here in California,

01:00but also elsewhere in the world. In Taiwan, we have key players taking care of power supply units,

01:10battery backup units. But of course, it matters a lot for us to engage here with the hyperscalers and

01:18GPU makers. So you're talking about that expectation around revenue. I think you're looking at it to

01:24grow by more than maybe 100 percent this fiscal year or this financial year. But in terms of the

01:31overall revenue mix or the contribution that AI makes, how much more do you see it being a part

01:37of the business? Yeah, it grows now to 10 percent of the group's revenue. So it really becomes relevant.

01:43It becomes visible in our top line and bottom line numbers. And I think the 10 percent is not the last

01:51statement. I think for the future, we will enjoy more growth because the power semiconductor bill of

02:03material in a rack is growing strongly. If you have like in the current generation, $15,000 of power

02:13semiconductors per rack, that number will go up all the way to $100,000 per rack in 2030. So besides the

02:23number of racks, besides all the capital that is invested, we are enjoying a specific growth driver

02:30because much of the increased compute performance comes out of the fact that more power is delivered

02:37into the processors. Automotives are still a large part of the business, though. Absolutely. Yeah. And

02:43this has been a market that's had quite a few headwinds as well. What does that look like for you for

02:50this year? Yeah, it's had puts and takes. On the one hand, the e-mobility is clearly in a slow

02:58growing phase. In the US, the market is very slow. In China, also, things are tapering off in terms of

03:06volume. There's price competition. In Europe, there's some growth. But we can make use of the

03:15capacities which we have, which we used to apply for electric vehicles and dedicate them to AI.