To most people, Intel is a behemoth that has always been around and always will be. It’s almost inconceivable that a corporation of this size would even be considered at risk – so the vast majority never give it a second thought.
If you ever did consider it, a quick glance at the balance sheet would be enough to dispel any notion of trouble. In 2013 for example, Intel reported profits of $9.6 billion on yearly revenues of $52.7 billion.
That seems clear enough – surely the company must be in great health? Yet looking at Intel’s stock performance vs companies of similar stature, something just didn’t appear right.
Over the past two years Intel’s share price weakened considerably vs their competitors in the Standard & Poor’s 500 index – which is basically a yardstick for the performance of companies of similar types and (large) size. Googling for “Intel in trouble” will also find you plenty of recent reading material. How can this be? The reason is simply that you need to look deeper at the finances and also have a grasp of just how much Intel’s competition has changed in a relatively short space of time.
To understand Intel’s weakness we need to look at their historical strength.
If one thing separates Intel from the also-rans, it’s their manufacturing capability.
Intel spent $10.6 billion on R&D last year – a large part of which goes on manufacturing. They also spent $10.5 billion on capex which is basically for fab equipment. This consistent, huge spending has bought them an advantage of a year or two over their closest manufacturing competitors – TSMC, Samsung and GlobalFoundries.
If you read the first paragraph of the no 20nm graphics article, you will know that the reason for spending all this money is to get smaller and faster chips. This should basically mean that Intel has smaller and faster chips than their competitors – but it’s not quite that simple. Manufacturing a chip is one thing but it also has to be designed properly in the first place. It helps too when the right chips get manufactured at the right time – and this is where Intel has completely failed.
Out with the old.
Many of you reading this will have heard of AMD and might believe that they are Intel’s main chip competition – but the sad fact is that AMD long since ceased to be competition for Intel in CPUs. AMD is mostly reliant on GlobalFoundries for their chip manufacturing (GlobalFoundries was formed by spinning-off their manufacturing division in 2009) and the process gap between them and Intel is most of the reason why AMD can’t compete – Intel is on 22nm (nanometre) and moving to 14nm while AMD has only recently moved on to 28nm. That’s at least a whole node difference.
|AMD Series||Node||Year||Intel Series||Node||Year|
|K5||500, 300nm||1996||Pentium I||800 – 350nm||1993 – 1997|
|K6||350, 250nm||1997 – 1998||Pentium II||350, 250nm||1997 – 1999|
|K6 II||250, 180nm||1998 – 1999||Pentium III||250 – 130nm||1999 – 2001|
|K6 III||250, 180nm||1999 – 2000||Pentium IV||180 – 65nm||2000 – 2006|
|K7||250 – 130nm||1999 – 2005||Core||65nm||2006|
|K8||130 – 65nm||2003 – 2007||Core II||65, 45nm||2006 – 2008|
|K10 (Phenom)||65nm||2007 – 2009||Core i3, i5, i7||45nm||2009 -|
|K10.5 (Phenom II)||45nm||2009 – 2011||Westmere, Sandy Bridge||32nm||2010 – 2011|
|15h (Bulldozer)||32nm||2011 – 2013||(Ivy Bridge, Haswell)||22nm||2012 – 2013|
|Steamroller||28nm||2014 -||Broadwell||14nm||2014 – 2015|
It was close for a while, but there was never any chance of AMD keeping up with Intel – it’s all about money and when the cost of reaching each new node started to hit the $billions, Intel pulled ahead.
So Intel raked in the $billions while AMD struggled with yearly losses, big and small. Competition has been all but obliterated in servers and high-end PC’s – where the high margins and real profit is to be made. Intel’s manufacturing advantage had put them in an unassailable monopoly position. So far so good.
But then something really unexpected happened. People started to notice that their current PC was good enough, so upgraded less often. Around about the same time they realized that they preferred shiny new stuff like smartphones and tablets instead.
In with the new ARMy.
While Intel was getting fat on monopoly profits a bunch of new guys (and some old) appeared and reappeared on the scene – mostly all making chips based on British company ARM Holdings RISC architecture. Smartphone and tablet sales boomed – and Intel was nowhere near either. (It was later told by Intel’s previous CEO, Paul Otellini, that they had turned down the original iPhone contract with Apple – and that wasn’t the only mobile mistake of Otellini’s tenure either.)
After a few mobile false-starts, Intel was finally ready with Moorestown in 2010, which according to Anandtech “has the potential to be the most exciting thing to hit the smartphone market since the iPhone.” That was another false start. Medfield in 2012 also promised “dominating” performance and power consumption – yet the phone makers couldn’t get far enough away from it. It became clear that Intel was good at talking a big mobile game while bringing rusty pen-knives to gun-fights.
Intel’s excuse was that up till that point they hadn’t really been trying. For some reason they continued to optimize their manufacturing for high-performance instead of low-power and the old Atom architecture was simply no good against the Reduced Instruction Set Computing chips from ARM.
We were led to believe by Intel that Merrifield was supposed to be the “real” start instead – a brand new atom architecture built to dominate the market and this time fabbed on the most advanced node in the industry, Intel’s 22nm. The fanboys went crazy dreaming of Intel’s coming knock-out blow with their updated mobile-optimized chip and process (as you can see from the link, the press had finally learned though). How could any of the ARM guys stand in front of the now-trying Intel?
Well as it turns out, CPU performance is just one part of the overall equation. The market leader, Qualcomm, had finally started paying attention to Intel – and in a clear dig at the chip behemoth stated that CPU performance was only a small part – around 15% – of the overall phone SoC package.
They even provided a diagram.
While we waited on Merrifield, Intel released Bay Trail T (the tablet version of the Merrifield phone SoC), and it was again lauded as being much faster than anything around. It failed, utterly, forcing Intel to take massive billion dollar penalties in order to shift them all. Based on Merrifield’s specifications it’s already clear that Intel still does not have the most exciting thing since the iPhone. They’ll be lucky to capture even 1% of the market but hey – at least with that they’d make the pie-chart next time around.
How is this possible? How can Intel be failing with their new low-power phone architecture built on the industry’s most advanced process. Surely that’s a slam-dunk if ever there was one? So what is Intel missing?
Well, most of that really big part labelled “CONNECTIVITY” for a start. Over the page things take a farcical, then worrisome turn…