Saturday, September 05, 2020

AMD has risen from the ashes to win Round one, but Intel has not been knocked out yet

I have not covered the semiconductor industry for a long time, but analysts’ reactions to Intel’s recent earnings call are too loud to ignore. The stir has been caused by the delayed rollout of Intel’s 7nm chips by six months, which are now planned to be launched by late 2022. AMD is now expected to gain substantial lead time over Intel, as its latest chips are already built on a 7nm process and will gain further lead by moving to the 5nm process by late 2021.

A Bernstein analyst went on to say that the earnings call was the ‘worst we have seen’ in the company’s history. Indeed, Intel’s share price tanked by 20% after the call, and has been languishing there for more than a month now.

That said, despite the pessimism exuding from analysts, revised price targets are not that bleak, down mostly in the 6%-14% range. The analyst reaction prompted Bill Maurer to comment in Seeking Alpha that ‘it seems everyone thinks Intel is dead’. He goes on to say that while the bleak commentary would have you believe that projections moving forward would be quite bearish, it is not really the case, and Maurer therefore recommends to ‘skip the funeral’. Of course, this is partly because Intel is no longer just a chip company – about half of its operating profit is now contributed by the data center group.

This is not the first time when AMD has left Intel behind on the development front. In the past twenty years or so, AMD’s share price has done significantly better than Intel in two instances. One is now, and the other one was when AMD held sway with its Athlon processor.

A brief history of AMD will reveal that fortunes in the processor market can be fleeting, and tables can turn overnight. Perhaps this prompted Andy Grove to say “only the paranoid survive”. The point that I wish to put across is that AMD has been here before, but it needs tremendous effort to stay at the top when you have a rival like Intel. While AMD’s recent feats are by no means trivial, Intel has the capability to bounce back if it gets really serious about the CPU market.

The beginnings of AMD as an underdog

Let us start from the start then. Both companies trace their roots back to Fairchild Semiconductor. Intel was formed when two employees of Fairchild, Robert Noyce and Gordon Moore, left the company in 1968 to form N M Electronics. Less than a year later, eight people left Fairchild and set up AMD. Historically, AMD was considered the cheap, copycat option. Before 1999, what Intel designed, AMD simply tried to make better. But even though AMD used to ‘reverse engineer’ Intel’s processors, they trumped Intel on performance – for instance, the Am486 processor launched in 1993 offered roughly 20% more performance than Intel’s i486. AMD’s success with the 386 and 486 helped double its revenue to USD 2bn during the 1990-94 period.

Days of Glory

In 1996, AMD acquired NexGen, a chip design company whose Nx686 processor demonstrated a 180 MHz core speed when AMD’s K5 and K6 (under development) were facing scaling issues. After the buyout, the Nx686 became AMD’s K6 and thus began an era of AMD’s dominance over Intel. Interestingly, Vinod Dham, who is called the Father of Pentium, had left Intel in 1995 to join NextGen. A classic case study that underlines the importance of keeping track of key management movement. The K6 processors, launched in 1997, gave stiff competition to Intel’s Pentium series and costed half as much as equivalent Pentium chips.

 

AMD retained its research edge, for instance, the K6-2 was known for offering an easier route to accessing the CPU’s floating-point capabilities. Finally, in 1999, AMD launched Athlon, which truly rocked the boat. It stood apart on several counts, from overclockability, to DDR RAM support and full speed L2 cache. In 2003, Athlon was upgraded to Athlon 64, because it added a 64-bit extension to the x86 instruction set. Till early 2006, AMD had the upper hand in the server market as well with its Opteron chips. This was the period when AMD’s share price did far better than Intel, whose Netburst architecture was looking like a failure. The company’s rating was upgraded from B2 to Ba3 in January 2006 by Moody’s. And then, AMD’s juggernaut hit an iceberg.

The ATI acquisition blunder

AMD decided to acquire ATI Technologies, the graphics card manufacturer, in July 2006. The deal size was USD 5.4bn, comprising USD 4.3bn in cash and loan, and the rest raised through equity. This was about half the market capitalisation of AMD at that time. ATI had no manufacturing sites and negligible revenue, and the value could almost entirely be attributed to its intellectual property. Over the next few years, AMD had to write off USD 3.2bn related to ATI's acquisition, revealing the extent to which the company had overpaid for ATI. In what could be termed as an embarrassing blunder, AMD also sold off the handheld graphics division of ATI for just USD 65mn. That division is now known as Adreno.

On the development front, AMD’s planned to challenge Intel in the server market through its quad core K10 Barcelona processor (Intel’s Xeon chips were just dual core). However, a bug was discovered in the K10, and by the time a BIOS patch was rolled out, AMD’s reputation had been damaged.

Intel makes a comeback

Meanwhile, Intel dumped Netburst and switched to their older Pentium M architecture. The Pentium name, now much maligned, was relegated to low-end budget models and was replaced with “Core”. In end-July 2006, the Core 2 Duo was launched and by the end of the year, Intel had reconquered the summit. AMD tried to retaliate by pushing the quad-core K10 design to the desktop market in late 2007. While the K10 was superior in some aspects, it was slightly slower and more expensive. It seemed that AMD’s research pipeline had run dry.

AMD perhaps then thought that the Athlon brand had suffered some damage, and launched Phenom to challenge Intel’s popular Core 2 Quad Q6600. But by this time Intel was well-entrenched on the software side of business, and unlike AMD, advertised more. Intel had its own foundries tailored to its own products and a large marketing budget. Intel did not stop here however. They used their large cash reserves to keep AMD CPUs out of new computers. Intel paid Dell a whopping USD 723m to remain the sole provider of its processors. AMD stood little chance to compete. The company was downgraded back to B2 by 2008.

AMD rises from the ashes

In 2010, AMD pulled up its socks by cutting down costs, disposing assets (chipmaking foundries) and focusing solely on processor design. The initial products were disappointing – the FX series could not effectively compete with Intel’s Core i7 for instance on power consumption and size. Support came from unexpected quarters though. The purchase of ATI helped AMD to come up with a combined CPU and GPU package, which was later used in the Jaguar architecture and selected for the Xbox One and PlayStation 4 in 2013. However, this helped to build reputation instead of profitability, as margins tend to be weak in the console business. AMD generated operating losses in each year from 2012 to 2016 (except 2013), and revenue was on a downtrend. But then in 2012, AMD hired former lead architect Jim Keller, who returned after a 13 year absence, and Lisa Su from Freescale Semiconductors.

Meanwhile, cracks were starting to appear in Intel’s citadel. In 2012, Intel expected to release CPUs on the 10nm node within 3 years. The company however released the first 14nm CPU in 2015. Yield issues kept cropping with 10nm.

The Ryzen effect

In March 2017, AMD launched Ryzen on its entirely new architecture (Zen). Ryzen was remarkably different from the processors that had come before it, and as Ryzen 3 and 5 models were released, they gave stiff competition to Intel on price. The developments kept on coming – in April 2018, AMD came up with Zen+ and in mid-2019, Zen 2 was launched. Helped by the success of Ryzen, AMD’s cash flows and profits witnessed a welcome turnaround. The CFO/Debt improved from just 6% in FY16 to 96% in LTM June 2020.

 

How closely Ryzen’s launch and success coincides with a turnaround in AMD’s fortunes can be seen from its rating history. The company witnessed a string of upgrades since 2017, at very short intervals. From being at a precarious Caa1 in September 2016, AMD was upgraded to B3 in February 2017, and is now rated Ba2, almost close to investment grade.

At the same time, Radeon (AMD’s graphics division), which had its roots in ATI, started to do well when given full independence. The new architecture is being deployed by Microsoft and Sony for their new consoles (Xbox and PlayStation 5). That said, Radeon is still far behind Nvidia and is seen as a budget option.

AMD has won the battle but not the war

As of now, with USD 7.6bn of revenue in LTM June 2020, AMD is even smaller than Nvidia (USD 13.1bn in LTM Jul 2020), forget Intel. In fact, Nvidia is so strong that its market capitalisation has doubled YTD at roughly USD 340bn, and is worth more than Intel and AMD combined. New titles like the Flight Simulator 2020 that need tremendous graphics power will keep the growth going for Nvidia. While one can argue that Nvidia is focused only on graphics cards and lacks diversification, AMD too is heavily dependent on the Ryzen series. AMD also has limited presence in the Enterprise segment, where Intel rules the roost (Intel derived USD 27.7bn from its data center group alone). AMD will also face problems because of its weak marketing. The ‘Intel inside’ campaign is still yielding dividends for its rival.

This time AMD is not competing with Intel alone. With Amazon’s Graviton2 and Ampere (that has a whopping 128 CPU cores) also in the race, the competition is heating up. Also, Radeon is still not out of the woods. Only recently gamers faced serious driver related issues. In a poll of 49,000 AMD GPU users, about half stated that they experienced driver problems.

AMD however has a few aces up its sleeve. Nvidia chose AMD over Intel for their new AI compute clusters, because AMD offers more cores and faster PCI Express lanes. AMD’s share for laptop processors reached a new all-time high of 19.9%, and new Ryzen 4000 mobile processors are expected to go into more than 100 laptop designs this year. Moreover, AMD’s latest chips are already built on a 7nm process and the 5nm process is expected by late 2021, whereas Intel’s 7nm chips will arrive by late 2022.

The real risk to Intel might be on the server (data-center) front

The bigger threat to Intel is from future in-house processors by cloud players (like Graviton2 from Amazon). These can hurt Intel where it now rules supreme – the server market. Intel holds a ~90% market share in the lucrative data-center/server market (operating margins of ~44%), although AMD is gradually gaining share just by offering lower prices. Competition can also come from Ampere, and from unexpected fronts like SiPearl, a Parisian company which is being backed by the European Commission as a part of European Processor Initiative Project. If price-led competition gets tough, Intel might be forced to sell Xeon chips at less than USD 1,000. The question to ask would be “Can Intel sustain if Xeon margins collapse?”.

Intel has historically shied away from low-margin businesses. It ignored Steve Jobs’ suggestion that they fabricate an ARM chip for the iPhone because of the thin margins in phone CPUs. Intel’s strategy has been that of a high-volume, high-margin company, which cannot afford to offer low-margin versions of its products for fear of cannibalisation. If the server market mirrors that of mobile processors, Intel’s days of glory might be well behind it. But that story is yet to play out.

The Empire strikes back

It appears that AMD has the upper hand as of now, but Intel has been quick to make changes. The Chief Engineering Officer Murthy Renduchintala left the company in August and Intel is considering to hire a third-party foundry to build its processors. A few days ago, Intel launched its new 11th Gen Tiger Lake laptop chips. When put to test against an AMD Ryzen 7 4800U device and a machine running an Nvidia MX350, the an 11th Gen Iris Xe-powered laptop (Tiger Lake machine) clearly comes out top, with significantly better frame rates.

Intel has previously tackled the AMD challenge either by its sheer balance sheet and marketing power, or on the development front (Core 2 Duo against the Athlon). The company is not going to go down easily and I am tempted to use one of my favourite witticisms from Mark Twain “The reports of my death have been greatly exaggerated”.

 

Sources:

1.      Google Finance

2.      Yahoo Finance

3.      CapitalIQ

4.      AMD

5.      Techspot

6.      AnandTech (https://www.anandtech.com/show/161/4)

7.      Techcrunch

8.      PC Gamer (https://www.pcgamer.com/intel-tiger-lake-11th-gen-launch/)

9.      All About Circuits (https://www.allaboutcircuits.com/news/ampere-altra-max-128-core-processor-challenges-intel-and-amd-in-a-cloud-based-processor-showdown/)

10.   Gizbot (https://www.gizbot.com/computer/news/intel-and-amd-might-face-tough-competition-with-arrival-of-new-players-067321.html)

11.   Standard and Poor’s

12.   Moody’s Investor Services

 


No comments: