Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/how-intel-came-crashing-back-to-earth-after-its-trump-bump.24412/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2030770
            [XFI] => 1060170
        )

    [wordpress] => /var/www/html
)

How Intel Came Crashing Back to Earth After Its Trump Bump

Daniel Nenni

Admin
Staff member
View Robbie Whelan’s  graphic link
Robbie Whelan Reporter at The Wall Street Journal based in Los Angeles

Lip-Bu Tan took over Intel last March with a mandate to cut down on what insider described privately as “drunk and disorderly” capex spending. Simply put, the company, once a Silicon Valley titan but lately fallen badly on hard times, was spending too much on its money-losing chip manufacturing business.

So Tan started slashing capex. He canceled plans for two fabs in Europe, further delayed another fab in Ohio, and made it clear internally that the “Field of Dreams” mentality of his predecessor — “If you build it, they will come…” — was over. No new plants or manufacturing lines would be built until they had solid customers locked in.

Investors responded warmly to this strategy. And they were even happier when the Trump administration in August took a 10% stake in Intel, arguing that the company was strategically important to U.S. interests and it was in danger of failing.

But one part of the cost-cutting plan nobody really noticed: Intel was also taking expensive manufacturing tools offline and writing down their value to save cash. These were the tools that were cutting-edge a year or two ago, the 3-nm and 7-nm and 10-nm process nodes that were used to make Intel’s once-popular Emerald Falls and Granite Falls data center CPUs (I know, every Intel product line sounds like one of those mysterious rooms in the Lumen headquarters from “Severance”).

Then, over the last few months, something strange happened. Demand for these somewhat-old CPUs surged. Companies like OpenAI, AWS and Google had realized that with the pivot to inference computing, agentic AI suddenly was requiring a lot more CPU capacity, rather than relying on more heavily on GPUs, the way the AI model/training phase had.

Intel got huge orders for its data center CPUs. The building boom picked up. They ran through their existing inventory, but because they had mothballed so many of he tools used to make these older chips, they couldn’t increase production. So they missed the wave.

That’s the story behind what they told investors in their earnings Thursday. And that’s why Intel has lost $46 billion in market cap on the last 36 hours.

And Intel’s Book of Job couple of years continues….

With great reporting from Sean McLain

 
This is what happens when your customers miss their forecasts. It happens to the best of us, even TSMC. There are also black swan events like the pandemic and now the AI surge that are impossible to forecast. For the foundry business customers sign wafer agreements when a design starts estimating the number of wafers that will be required over the next 3-5 years. It takes 2-3 years to tape-out a complex chip and another 1-2 years to fully ramp up production. It could take less time of course if you have flawless execution and everything goes perfectly. That however is a rare occurrence in the semiconductor industry because sometimes you do not know until you know.

TSMC has the best look at semiconductor demand. TSMC is the pulse of the semiconductor industry. Even TSMC was rattled by the pandemic when customers juggled wafer orders and now TSMC customers want more wafers than they signed up for. This is a much better position to be in than to have your fabs running at 50% of capacity, right?

Politics add even more complexity. How is Nvidia and TSMC supposed to forecast business in China?

Bottom line: This is not an execution problem. Lip-Bu Tan's plan to revive Intel is working. This bump is just how the semiconductor business works. As I have said before, never bet against Lip-Bu, absolutely.
 
Back
Top