A Proposal for Intel – Time to Go Private

Intel has been in the news lately, largely because of its CEO transition, which gives the impression of a company in turmoil. In truth, the root of Intel’s problems lay much deeper. The company is at the point where it has gotten so big that it has no clear path to growth, and its core business is very much under threat.  To survive, it needs to make some drastic changes. Our impression is that the management team and the Board are fully aware of the problem. Unfortunately, it seems they cannot find a way out.  The good news is that they continue to throw off immense amounts of cash, and while that will someday come under threat, it gives them options. For now.

We think the solution, painful as it may seem, is to go private and get bought out by private equity (PE). To be clear, we are not big boosters of PE. The methods these firms employ can get very extreme and often go too far. That being said, it has become very hard for companies to implement radical surgery under the scrutiny of public markets. We think the company today has the means to do this on their own, but seem to lack the will. In these situations, private equity can serve as a crucial forcing function.

First, some background. Intel has long held a central role in the electronics industry as the dominant provider of computer processors. However, the PC market peaked long ago. And while it grew in the most recent quarter, that was the first quarter the market has grown in six years. The reality is that PCs are in permanent decline. They will not go away, but demand is not coming back.

Intel has tried everything to find new growth. They have spent something like $20 billion over the last decade and a half trying to enter mobile with numerous acquisitions and immense operating losses. The end result of all that is they have exactly one customer, and no one wants to be entirely dependent on that customer. They have tried to get into automotive, with several acquisitions. Our sense is that this is not going particularly well either. They are trying to buy their way into an AI strategy, but this is also rumored to be faring poorly as well.

The company does have one key area of product strength – they have an effective monopoly in the supply of CPUs for data centers. They effectively have no or little competition in that market, with the margins to match. This market is growing very strongly as everything moves to the Cloud. And while many new compute applications like Machine Learning and Artificial Intelligence are seeing a big shift towards GPUs, all those data centers still need a lot of CPUs.

Now to be clear, the company is still massively profitable. Last year, they generated $18 billion in operating profit on $62 billion in revenue. Their server products generated almost half that profit, and the PC business is still profitable. However, it is important to understand the company’s cost structure. Intel still operates its own manufacturing facilities or fabs, and these required almost $12 billion in capex last year. Intel has a massive fixed cost base associated with this, with over $6 billion in depreciation last year.  So Intel has to focus all their energy on keeping their fabs fully utilized. Today they do this through a combination of high volume products (PCs) and high margin products (servers).

The trick with these kinds of models is that when they crack they crack faster than people expect, and they crack hard. The decline in PC volumes have left the company’s fabs less utilized than is comfortable. In an attempt to drive more volume through their fabs Intel several years ago opened up a Foundry business, supplying manufacturing capacity to other chip designers without fabs (i.e. fabless semis, which is how most chip companies operate). We think it is safe to say that this effort has failed. Not only has the company failed to attract any sizable customers, many of its non-core products are actually produced at others’ foundries.

Intel’s history is tied deeply to its manufacturing expertise. They were the leader in manufacturing processes for decades. Unfortunately for them, this long expertise has left them with a serious cultural problem. For years, the company was able to eliminate many problems by advancing its manufacturing process. After all, Moore’s Law is named for a former Intel CEO. This created the false impression internally that the next process node would solve all problems. This did not work out well in mobile, or really anywhere other than in the core CPU business. To make matters worse, the company has actually fallen behind the foundries in process technology. They have had to delay their introduction of 10nm products, stranding them at 14nm while their competitors are pushing from 10nm to 7nm.

More importantly, over the years the manufacturing team has cemented itself at the center of the company’s political orbit. We think this helps explain why the company has failed to win customers for their foundry business. Accustomed to calling the shots internally, the company’s operations team has apparently struggled to learn what it takes to provide customer service.

Which brings us to private equity. As we noted in the introduction, sometimes the only way to bring about drastic cultural change within a company is to bring in hard-eyed, sharp-elbowed new owners.

The math behind such an acquisition is fairly straightforward. Intel’s current enterprise value is $256 billion ($250 billion in market cap, plus $28 billion in debt minus $20 billion in cash). Apply a 15% premium to that, and the deal size is $300 billion. Admittedly, this would make it the largest private equity transaction ever, but as several investors have pointed out to us recently this is probably the most deal-friendly environment ever. The deal would leave the company with debt about 8.8x EBITDA, which would be a stretch, but is firmly in the realm of the possible.

The one major wrinkle would be the company’s massive capex budget, needed to support all those under-utilized fabs. The interest coverage ratio (EBITDA to Interest payments plus capex) for this deal would be about 1.1x, which is tight.

That being said. There are some pretty straightforward actions the new owners could take quickly to make this work. First, the company burned $4 billion last year on “Other”, which is probably code for mobile. Shut that down, and the debt ratio goes to 7.6x, and that interest coverage ratio goes to 1.3x which is still high but feasible. We think it is safe to say that a company of Intel’s size and lineage has considerable excess cost hidden throughout the company. True, they shut down their airline a few years ago, but there are likely still immense white elephants to be eliminated and workforces that could be more productively deployed. All of which is an incredibly diplomatic way of saying, the private equity owners would cut a lot of people.

One trick of investment bankers in this position is to work backwards. Pick a target ratio and then calculate the “Synergies” needed to achieve it. In this case, if we want to get to interest coverage of 1.5x the company would need to cut $8.8 billion in costs. We already have $4 billion from “Other”, leaving a target reduction of a further $4.8 billion. To put this in context, average comp per employee at Intel is around $200,000 per year. To cut $8.8 billion in costs would mean firing 35,000 people, a third of Intel’s 107,000 employees. To be clear, we are not advocating that scale of bloodletting. There are almost certainly many other costs to cut. (Pro Tip: Ask Broadcom’s Hok Tan for some pointers). Nonetheless, looking at these numbers it is clear why new owners would be needed, and equally clear that the process will be painful.

One final consideration. Another popular private equity practice is to acquire a company and then divest pieces of it to bring down the debt load. There is probably no easy way to do this here. Intel has shed its largest recent non-core acquisitions (e.g. McAfee, Wind River). In theory, they could sell Alterra. But the most serious step would be to split the company’s chip design and fab business. After all, the root of the whole problem is that Intel’s foundry business is going nowhere because it has not been receptive to customer service imperatives. Sell the design business and then focus the fabs on becoming a Foundry.

This level of cultural change is almost certainly burning-at-the-stake level heresy within the company. It would make for a traumatic shift. So much of the company’s history  is built on the intertwining of the two sides of the business. Splitting the two apart would be akin to surgically separating conjoined twins – there may be fatalities involved. However, there is precedent. This is exactly what Intel-rival AMD did a decade ago, selling off its foundry business into what is now known as Global Foundries. The results of that transaction are still not clear – many positives, some negatives. Nonetheless, AMD/GF proves it is possible.

We are fully aware that this is an outlandish suggestion. A drastic to solution to what may not be that drastic a problem. In many regards, Intel is in great shape. It is immensely profitable, and a paragon of technological prowess. A private equity transaction done poorly could wreck all that. That being said, there are clear signs of decay. The company has struggled for decades to break out of its rut, to no avail. It has no clear avenue for growth. Limited success at new product market entrances. And a heavy reliance on its data center business which all of its customers would love to see broken.

The company could conceivably maintain this status quo for many years. On the other hand, a few small demand disruptions could break its model faster than many can imagine. Taking drastic action now will undoubtedly be painful, but it is likely that the alternative will merely lead to a prolonged period of suffering resulting in even greater misery down the road.

6 responses to “A Proposal for Intel – Time to Go Private

  1. Pingback: Trouble for Intel in the Data Center? | DIGITS to DOLLARS·

  2. Pingback: And then there were four…or three… or maybe it’s two | DIGITS to DOLLARS·

  3. Pingback: Amazon enters the ARM Server fray | DIGITS to DOLLARS·

  4. Pingback: Intel for Mellanox? | DIGITS to DOLLARS·

  5. Pingback: Should Intel Split in Two? | DIGITS to DOLLARS·

  6. Pingback: Intel Selects Option C – All of the Above | Digits to Dollars·

Leave a Reply