Value In Complexity
Thesis
Silicon Graphics International (
SGI
)
is a compelling turnaround story with exceptional momentum that has
been misunderstood due to a hard-to-understand product (and product
mix), a complex accounting change, and an acquisition in FY2011
(June FYE). Share price weakness has created a compelling entry
point both ahead of this quarter's earnings and for the long term.
I believe the stock could double in the next 6 months and has the
potentially be a 3x (or greater) in the next 20 months.
Editor's note - For a recap of the short's thesis from
May,
click here
.
Business Description
The current Silicon Graphics International (here "SGI") has a
complex story, but it needs to be fully understood to appreciate
the thesis. SGI is a combination of the old Rackable Systems
business which was in decline when it bought the old Silicon
Graphics out of bankruptcy in the depths of the recent financial
meltdown in 2009 (the old SGI was in BK court literally minutes
before [[GM]]!). When Rackable bought the old Silicon Graphics (and
changed the combined company's name to SGI), the old Silicon
Graphics had an outdated product set and was still some $30 million
in R&D and 9 months away from having a viable (or profitable)
product - hence the bankruptcy. Fast forward to now, the combined
SGI has a strong new management team, led by Mark Barrenechea
(formerly CTO of Computer Associates and SVP of Oracle) who joined
in 2009, and a truly cutting edge line of products coming out of
the technology purchased from the old Silicon Graphics. As a quick
glance at the financials will reveal the business has been growing
well (even without the accounting change... more on this later) and
is just turning the corner into profitability
A quick note on what exactly they sell
SGI has 3 basic segments:
- 1) Old Rackable systems business selling commodity servers
and hardware into large data centers at very low margins (est.
5%-10% contribution margin). This is a legacy business, which is
unattractive except for providing fixed cost and purchasing power
leverage. Management has the stated goal of growing this segment
much more slowly than any other the other pieces of the business.
This segment is roughly 1/3 of SGI's revenue.
- 2) Data storage and processing solutions. This is the real
engine behind SGI: industry best technology around
high-performance computing ("HPC") architectures called
distributed memory (or "clustered computing") and Shared Memory
Architecture ("SMA"). Both of these products (based on Intel x86
chipset) have exposure to the growing prevalence of Big Data and
the increasing prevalence of advanced data analytics. This
segment is roughly 1/3 of SGI's revenue.
- 2.1) Distributed memory allows a computer to break up larges
tasks among many different machines, a necessity in
supercomputing. The SGI product line is very well regarded at the
high end, but this remains - in all candor - a competitive (if
attractive and high-growth) market with worthy competition
([[HPQ]], [[IBM]], [[CRAY]], etc.) although margins still come
out in the mid-low 30%s. This is where the Altix ICE product
competes.
- 2.2) Shared Memory Architecture ("SMA") has a different use
case than distributed memory. While distributed memory can scale
massively (to >1,000,000 cores on some systems), it cannot run
some complex analyses, multi-vector correlation analysis in
particular which SMA (despite maxing out at 2,048 cores) can
handle. While the HPC SMA use case was at one point restricted to
places like NASA, university research labs, national
intelligence/cyber-security and the like, with the proliferation
of Big Data, the high-end is becoming more ubiquitous. For
example, PayPal and the USPS use this for real-time fraud
detection (postal fraud cost the USPS over $130 million in
2009!). SGI's Altix UV product has been out for only about 5
quarters and is gaining exceptional traction as the runaway
leader in SMA computing performance. The 2 closest competitors
are literally 1/16
th
or ½ the capacity but on a "dead" Intel (
INTC
) chipset (Itanium chipset, which HP is the only company that
still OEMs) speed. SGI is also about to launch an upgrade Altix
UV 2000 that will 2x the number of cores to 4,096 4x the capacity
to 64 terabytes. The TAM here is huge and growing, with HP alone
currently doing ~$2.0 billion in annual revenue.
- 3) Services: When SGI sells a big solution (other than the
mostly low-end Rackable products), they also sell a relatively
expensive ongoing services contract on that solution, generally
multi-year at a percentage of the of purchase price. They also
provide IT services as an add-on to their basic product offering.
This makes up much of their deferred revenue (and deferred cost)
post accounting change.
Why the Business Is Mis-Priced Today
I'm sure many of you have read the well-publicized SGI short
thesis which states that without the accounting change and
acquisition, the core SGI business is growing organically at 2% or
less per year. Having gone through the Ks and Qs myself, I will
absolutely agree the company did a very poor job clearly defining
what the impact of the accounting change and the acquisition was on
Revenue and GM in each quarter. That said, the short thesis takes
an unrealistic (and at times simply inaccurate) view of a business
which, as articulated below, is growing quickly and has huge runway
ahead of it. Below is a sampling of refutations to the short
thesis:
Acquiring a customer:
SGI acquired SGI Japan. SGI Japan was a customer of SGI before the
acquisition. SGI reported explicitly that it would have recognized
$9.6 million of revenue in FY2011 from sales to SGI Japan had SGI
Japan not been acquired. It seems the short thesis ignores this
revenue.
Organic Services Growth in Japan:
Following the tragedy in Japan, management moved 60-70 service
oriented employees to Japan to help with IT services required in
the recovery, which led to the $22.5 million in SGI Japan services
revenue and a jump in product revenue to $23.7 million in FY 11 4Q
from a run rate of ~$17.5 million. As this revenue was generated by
employees of SGI pre-acquisition (who would have likely been
generating revenue elsewhere had they not been in Japan), it should
not be characterized as "acquired." My analysis includes 75% of the
incremental services revenue and 0% of the incremental product
revenue as organic. This is verified by the drop in non-SGI JP
services revenue, which would be unusual in the largely-recurring
services business.
Amazon:
The short thesis points to decreases in Amazon (
AMZN
) revenue as a negative; however, management has stated repeatedly
the desire to transition away from low-margin contracts. The
Rackable business got started building huge data centers for the
Silicon Valley titans and this customer concentration is a vestige
of that model.
Accounts Receivable:
The short thesis rather strongly pointed to a decrease in A/R as a
cause for the business being cash flow positive in FY11 3Q and that
this was indicative of a crumbling business. It is no secret that
A/R is driven by revenue and revenue was down in FY11 3Q vs.2Q. The
question is: was this a bad thing? What the short thesis misses is
that
management stated on their 2Q call that they were expecting
this drop in revenue
due to revenue pull-forward. Referring to the second half quarterly
revenue guidance: "at the high end [of guidance] would be something
- if you were just to do it mathematically, you'd be around, I
don't know, close to $140 million per quarter" (vs. $178 million in
2Q). The business had organic (adjusting for SGI JP) revenue of
$140 million in 3Q and $175 million in 4Q. Business lumpiness is
not necessarily business weakness...
Federal spending:
Within federal, the vast majority of SGI's sales go to NASA and the
Intelligence and defense communities which at first seem like scary
places to be. Within NASA, the majority of the Company's work is
either long-term contracted or covered in a multi-year funding
grant that has three years remaining. On the defense side, I have
personally worked very closely with a company selling high-tech
solutions to the Intelligence community, and I have seen first hand
that these budgets are among the last to get cut (especially in a
post wiki-leaks world). This is true for two reasons: 1) Simply
put, it is easy enough to assess the impact of cutting a tank
program (fewer tanks), but it's harder to pinpoint when reducing
10% of computing power has a real negative impact on our
Intelligence capabilities. And 2) budget allocation for technology
is done based on annual budgets where "what you don't use is taken
away" and the individuals making actual the purchasing decision
have only one goal: make sure it "just always works" for their
bosses. For better or worse, these dynamics (and the obvious
growing importance of cyber security and warfare) lead Intel and
defense technology budgets to be quite resilient. Further, channel
checks with government VARs all confirm health in this channel.
Skipping to the Crux: The Business Is Growing...
Fast!
The SGI story hinges on the growth of the high-margin,
high-value HPC SGI solutions, and therefore I have focused on that
in the below analysis. Management reiterated their FY2012 guidance
for $740 million - $780 million (roughly 50% organic and 50%
acquired) off of $630 million in FY 2011 (18%-24% growth), and the
debate is largely around whether this is requires an unrealistic
acceleration of a stagnant business (short thesis) or represents
reasonable guidance for a business with very strong momentum, which
I think is the case here. A few key data points:
Bookings ("Unfakeable") Growth:
While an accounting change and other "smoke and mirrors" can
obscure an income statement, barring outright fraud, the balance
sheet is much tougher to manipulate. Therefore, given the
accounting change, Bookings (Revenue + change in Deferred Revenue)
is a very good indicator of how much "new business" the company is
actually winning. My assumptions - and adjusting out SGI Japan -
get to 9.5% organic bookings growth for FY2011 (the bull case gets
to >14% organic growth!), with
35.7% Y-o-Y growth in the fourth quarter
. Admittedly, the business is lumpy by nature, but the key takeaway
is clearly strong growth, and a very strong end to the year.
Simply put, had the accounting change not occurred although
recognized revenue would have been $92.1 million in 4Q FY2011 (down
19% sequentially and 9% Y-o-Y), but deferred revenue would have
increased by $73.8 million in the fourth quarter (vs. an average of
$27.3 million in the prior 4 quarters) and therefore
Bookings for 4Q 2011 was $165.9 million, 20% and 36% up
sequentially and Y-o-Y.
Revenue from Rackable is effectively never deferred, so the growth
in the deferred component is indicative of growth in the HPC
business.
It is also worth noting that accounting rules do not allow an
acquiring company to book the deferred revenue of the target
company, so SGI Japan would not have meaningfully influenced these
figures.
(click to enlarge)
Above their Own Bar:
Pages 40 & 41 of SGI's latest 14A Proxy filed Oct 21 lays out
management's variable performance based compensation and their
internal goals, as set by the Board (who approved the accounting
change etc.). In the fourth quarter, the Company
beat their internal revenue plan by
7.2%. For the FY 2011, they were 6.6% ahead of the internal revenue
plan (after beating in Q1 and Q2 as well, before the SGI Japan
acquisition). The company did miss the GM% target in 4Q by 330 bps,
but achieved their GM% target for the year. The key takeaway here,
however, is that the Company has been consistently beating the
Board approved, internal plan (which for obvious reasons is
unlikely to be fooled by accounting shenanigans).
Hiring Correlates with (Real) Growth:
The Company has stated they are hiring, including growing their
sales force by 12%. A quick trip to Monster.com, Linkedin.com, or
Dice.com will corroborate that SGI has multiple postings for
positions across the organization. Company's rarely hire when they
are not growing (for real).
We Already Know This Was a Good Quarter:
Management's analyst day was
3 days before the end of the first quarter
. They reaffirmed their guidance (they only give full year guidance
as big deals can be lumpy) and were very optimistic. A couple
direct quotes from 9/27:
- Our U.S. Federal business is strong. We expect it to grow in
FY 12
- We expect in FY 12 all geographies to grow, all right
A Quick Note on Margins
While the above analysis focused on top line organic growth
which has been complicated by the accounting change and
acquisition, the margin growth story is much more straightforward
and is very well covered by the various research analysts following
the stock.
Long Term Margin Story:
As the very low margin Rackable business becomes an increasingly
smaller component of overall mix, margins will naturally
appreciate. Over the last 6 quarters, the Company has grown gross
margin vs. comparable quarter prior year in each quarter and by an
average of 779 bps (FY11 GM% was 27.0% vs. 22.2% in FY2010).
Altix UV:
As mentioned above, the Alitx UV has a huge technology advantage in
shared memory architecture, allowing the Company to earn >40%
product gross margins.
Guidance Expects Solid Margin Growth
: Management reiterated 100-300 bps of margin growth in FY2012 at
their analyst day. The long term goal of non-GAAP EPS of
$1.75-$2.00 on $1.0 billion of revenue (potentially by FY 2014)
also implies ongoing margin growth.
Valuation
The stock currently trades at roughly 0.5x sales. Many of the
peers are valued at >2x revenue, especially the smaller high
growth peers (with M&A comps
4x-10x
revenue). Even if you exclude all of the Rackable revenue (~1/3 of
sales), the business is cheap at .75x sales of the high-growth HPC
products and the accompanying high-margin services.
As the Company was the combination of two formerly distressed
businesses, it is still on the verge of consistent profitability
and therefore trailing multiples of earnings or EBITDA or earnings
don't really apply here. That said, if the company hits the long
term plan of $1.75-$2.00/share in EPS in FY2014, and it trades at
20x forward PE, it's a $35-$40 stock by June ((FYE)) 2013 in 20
months, assuming no value for the balance sheet assets... Which is
still only arround 1.0x revenue, implying there may be further
upside from there. That timing is likely more aggressive than what
can be underwritten by a prudent investor, but it gives a sense of
the potential outlined by SGI's long term guidance.
The company has $140 million of cash ($4.34/share) on the
balance sheet and no debt. It also has $466 million in NOLs and 600
patents on its balance sheet. Excluding cash and the value of the
NOLs (roughly $3.03/share by my math), the business trades at just
9.9x the midpoint of FY 2012 Earnings - below the revenue multiple
paid for some comps!
M&A:
The Company's technology and patents make them a clear target for
any of the large technology firms (HP, Dell, EMC etc.) or Intel,
who could directly use the ability to leverage additional cores in
each system. Additionally the CEO clearly knows ORCL management
from his time there. It goes without saying that M&A could be a
very compelling outcome; even a 1.0x revenue takeout would be
roughly double the current share price.
Other Relevant Data Points
New partnership with Microsoft:
SGI is partnering with MSFT to access the lower end of the HPC
market by integrating with MSFT's Windows Server 2008 R2 including
SQL Server. This was a long time in the making and could be a nice
boost in the next few quarters.
Short Squeeze Potential:
Given the popularity of the bear thesis, 24.2% of the stock's float
is sold short implying 14.5 days to cover and creating the
potential for a huge short squeeze on after solid earnings (as
happened on the prior 3 earnings announcements).
Audit Committee:
Given the chatter around the accounting change, it is worth noting
that Chairman of the Company's Audit Committee is the former
Managing Partner of the Earnst & Young San Francisco office.
The CFO of HyperRoll (a high-performance database software company)
is also on the Audit Committee.
Potential Risks to the Downside
- The Company sources some of their HDD hardware from WDC. The
factories in Thailand that were impacted by the floor make
primarily thin HDDs for notebook rather than data center form
factors and SGI should not see a direct impact. Still, HDD market
constraints could have negative implications in the near
term
- Revenue is lumpy, so deal slippage creates very real headline
risk in any quarter
- Along the same lines as lumpy revenue, reports indicate that
the Company has delivered a few Alitix ICE Romley based systems
already and will sell more in the coming quarter, but accounting
rules will prevent revenue recognition until Intel makes the
Romley "Generally Available" despite SGI having the product
already delivered to the client
- The Company is still in the early innings of selling into
commercial enterprises (as the high end of HPC was traditionally
limited to Federal/Intelligence, laboratories, and massive data
users like weather forecasters, etc.) and the ongoing success of
their sales force is yet unproven.
Catalysts
- Strong FY12 1Q earnings (reported Nov 8 after market)
- Ongoing strong Altix UV sales called out
- M&A Activity
- Roll out of Intel Romley processor systems with Altix ICE X
and Altix UV 2000 which are just beginning to roll out in FY12
Q2
Disclosure:
I am long
SGI
.
See also
Colgate May Be Worth Considering As An Income
Play
on seekingalpha.com