Intel: Power games between bosses and investors

Contents

"It's been a tough few weeks," admits Intel's CEO Pat Gelsinger at a Deutsche Bank investor conference. He requires an explanation because the direction he is aiming for is currently swallowing up many billions of US dollars and Intel is in the red. Even 15,000 job cuts are not enough for investors as a cost-cutting measure – the share price fell further after the announcement.

Intel is apparently prepared to make concessions, allegedly including delays to planned semiconductor plants, the so-called fabs. This could also affect the plant in Magdeburg, which costs a good 30 billion euros, writes the news agency Reuters.

Gelsinger wants to make Intel the world's second-largest chip contract manufacturer by 2030, after the global market leader TSMC. A major undertaking, as Intel has so far produced virtually no chips for external customers. To change this, the Intel Foundry production division is becoming independent.

Underestimated problems

So far, however, the internal processes have not been designed for external customers. The company will probably only be able to attract more external customers with the Intel 18A production generation. The design tools required for chip production (Electronic Design Automation, EDA) and licensable chip components (IP blocks) from industry giants such as Synopsis and Cadence will be available.

"We underestimated, I underestimated, that in addition to making good wafers, the EDA and the IP ecosystem that needs to be put in place to get the designs to the foundry is a lot of work", Gelsinger explains (audio recording at Intel, transcript at Seeking Alpha).

The biggest problem until then: Intel cannot utilize its own (and costly) fabs to capacity because more and more chips for Intel processors are also coming off the production line at TSMC. Added to this are the billions in investments for the construction of new plants that Intel considers necessary by 2030.

Deeper insights into Intel 18A

Gelsinger is trying to reassure investors. Once again, he emphasizes that there are around a dozen customers for 18A. Why have no names been mentioned so far? The companies do not want to jeopardize their good relations with TSMC currently, it is said.

The insights into the current status are unusual. Normally, chip manufacturers keep the yield rates of their production processes safeguarded. Now Gelsinger says for reassurance: 18A has a defect rate of less than 0.4 defects per square centimeter – a very good value, according to his own statements, for a process that will only go into series production a good year later.

Intel could therefore already produce a hypothetical 50 mm² device today, with a yield of a good 82 percent of fully functional chips. The compute chiplets of today's multi-chip processors are in this order of magnitude. Intel itself calls such chiplets tiles – that of the mobile Core Ultra 100 (Meteor Lake) measures around 40 mm².

According to Intel, it is currently working on eight so-called product tape-ins. This means that customers have sent their designs to the Intel Foundry, which is now producing suitable exposure masks and then the first test chips. Corresponding products are expected from 2026. From 2027, Intel Foundry should be profitable, i.e. generate net profits.

The company's own processors Panther Lake (desktop, mobile) and Clearwater Forest (server) are already one step ahead. The first prototypes are running in the labs.

"Unacceptable margins"

Meanwhile, Gelsinger is making concessions to investors: "[...] The margins are unacceptable. And internally, we are very focused on the fact that we have to get back into the 4s and then back into the 5s very quickly. And, of course, some of our product lines need to get into the 6s."

This refers to the gross margin, which currently stands at 35.4 percent and should rise to more than 50 percent in the long term. Intel's high profitability once made it particularly attractive to investors.

Intel board wants to discuss options

Reuters and BNN Bloomberg both report that the Intel Board of Directors will meet later this month to discuss various options. Further spin-offs, for example of the FPGA manufacturer Altera, would be an obvious option. Altera has recently been operating in the red, so a sale is unlikely to bring in much money. Marvell is a possible buyer.

Other partnerships at the semiconductor plants are also conceivable. There are already two precedents for this: the Canadian asset manager Brookfield Asset Management is backing almost half of the 30 billion US dollar Arizona plants. In return, a revenue share beckons. Meanwhile, asset manager Apollo has announced that it is taking a 49 percent stake in the USD 11 billion expansion of the Irish site.

Ultimately, a spin-off of Intel Foundry is not completely unthinkable. However, it would be a loss of face for Gelsinger, who is betting everything on the project, and would be tantamount to a vote of no confidence. Intel is currently protecting itself from aggressive investors.

Meanwhile, more and more parallels are emerging with AMD in the late 2000s: in 2007 and 2008, the company made losses of more than three billion US dollars (4.5-5 billion adjusted for inflation). This was followed in 2009 by the spin-off of chip production as Globalfoundries.

(mma)