Intel suffers turnover like any company, but its upper
echelons aren’t known for being a revolving door. That’s changed in
recent weeks with the (unconfirmed) departure of Aicha Evans, who joined
the company in July to head its mobile semiconductor division. Now,
longtime veterans Kirk Skaugen and Doug Davis are both leaving as well.
Skaugen was a senior vice president and general manager of Intel’s
client group, while Davis is a senior VP and general manager of the
Internet of Things division. Skaugen is leaving immediately, while Davis
has pledged to stay on through the end of 2016.
These departures all occurred just a few months
after Intel hired former Qualcomm executive Dr. Venkata “Murthy”
Renduchintala, president of the Client and Internet of Things (IoT)
Businesses and Systems Architecture Group.
We don’t talk much about Intel’s market position because the
company is fundamentally healthy; far moreso than AMD, at any rate. The
truth, however, is more nuanced. Intel still dominates the datacenter
and conventional client markets, but the company has had very limited
success in the mobile and Internet of Things markets.
In the graph above, CCG stands for client computing, DCG
means data centers, IOT is the Internet of Things, and SSG is software
and services. Datacenters have been a growth market for Intel, but other
areas have underperformed.
Eight years ago, Intel’s problem was simple: It couldn’t
build chips that were capable of hitting the necessary TDPs to operate
in a smartphone or tablet. Today, that bottleneck is long-since solved;
x86 Android tablets are indistinguishable from their ARM-based
counterparts at the same price points. The larger issue, according to
sources we’ve spoken to, is that Intel hasn’t found a cost structure
that simultaneously supports its margins and delivers performance-equivalent silicon to compete against ARM.
Intel’s traditional plan has been to subsidize low-end
markets with high-end sales of server, workstation, and datacenter
processors. This works well in the existing PC market, but the company
has had no success in duplicating this model in the tablet world. When
Windows 8 launched in 2012, Intel’s stated plan was to capture the high
end of the tablet market (think $700 – $900 devices) with Clover Trail,
while ARM devices running Windows RT would be pushed to the bottom of
the market. This never happened — customers were turned off Windows 8 in
general, and Clover Trail simply wasn’t powerful enough to justify its
own price point. Over the last few years we’ve seen Intel ship millions
of tablets with contra-revenue, launch new partnerships with companies
like Rockchip, and iterate on its Atom processors to deliver lower-cost,
higher-performing variants. Despite these iterations and improvements,
it has yet to take significant share in mobile.
(No one is making significant revenue in the Internet of Things, and
there are no killer devices from any company, so we can’t exactly say
Intel is doing worse than anyone else.)Being an IDM is a blessing — and a curse
4-5 years ago, Intel’s status as an Integrated Device
Manufacturer (IDM) looked like a huge positive for the company compared
to its competitors. An IDM is a company that’s responsible for both the
design of the silicon and the foundry that will build the chip itself.
Intel’s fabs are famous for their design methodologies and “copy
exactly” rule that mandates each and every fab be built to the same
specifications and with the same capabilities. If one Intel fab adopts a
new manufacturing technique using platform XYZ, that technique and
hardware will be adopted by every other Intel foundry tasked with
manufacturing the part.
Intel,
TSMC, and Samsung feature sizes at 14nm. Intel leads its competitors by
multiple metrics, but these metrics haven’t given it the advantage it
hoped for.
The advantage to having extremely specific design rules that
are tightly coupled to your process technology is that the CPU team can
design hardware mapped to the process node’s strengths. The
disadvantage is that it makes fabbing for other customers difficult and
drives up your own costs.
The problem boils down to this: Intel’s CPUs and manufacturing are tuned to deliver exceptional products
that are designed in a particular way. Foundries like TSMC,
GlobalFoundries, and Samsung are tuned to deliver high throughput, low
costs, and flexible manufacturing platforms that can be adjusted to
whatever the customer needs.
Intel is caught between a rock and a hard place. On the one
hand, its continued dominance in traditional PC markets has been partly
driven by adopting philosophies like copy exactly. This guiding
philosophy is also partly responsible for the problem it has today with
scaling its products into new markets, and these aren’t situations that
hiring or firing executives are going to solve. So far, the company
hasn’t demonstrated a coherent strategy for fixing this problems. Its
foundry business remains small, its mobile and client businesses are
still shrinking, and a handful of partnerships with Chinese OEMs aren’t
sufficient reason to overhaul its fabs for mass manufacturing.
Post a Comment