REPORT FROM BARS--WHY THEY KEEP POUNDING THE TABLE!:
ASCEND COMMUNICATIONS
April Update: Why we continue to strongly recommend ASND
Paul Johnson, CFA (212) 407-0415 paul_johnson@rsco.com
Tim Weingarten (212) 407-0444 tim_weingarten@rsco.com
Ara Mizrakjian (212) 407-0406 ara_mizrakjian@rsco.com
B a n c A m e r i c a R O B E R T S O N S T E P H E N S BancAmerica ROBERTSON STEPHENS
Ascend Communications, Inc. ASND $43.44 4/22/98
Industry: Networking
Paul Johnson, CFA 212 407 0415
Change in Yes/No Was Is Ara Mizrakjian 212 407 0406
...Rating: No BUY
...EPS 1997: Actual $1.06 FY DEC 1997A 1998E 1999E
...EPS 1998E: No $1.18 EPS: 1Q $0.31 $0.26 A $0.35
...EPS 1999E: No $1.46 2Q $0.31 $0.28 $0.36
52-Week Range: $60-22 3Q $0.20 $0.30 $0.37
Shares Outstanding(MM) 192.8 4Q $0.25 $0.34 $0.38
Market Cap ($MM) $8,377 Year $1.06 $1.18 $1.46
Avg Daily Trading Vol (000) 7,136 P/E 40.8x 36.9x 29.8x
3/98 Bk Value/Sh $5.47
3/98 Tot Debt/Tot Cap 0%
LTM ROIC: 40% Revs($M): 1997A 1998E 1999E
Price/Book Value: 7.9x 1Q $292.7 $305.1 A $420.0
EBITD/Sh: NM 2Q $311.7 $335.0 $435.0
Div/Yld: $0.00 NM 3Q $270.4 $360.0 $450.0
3-Yr Sec Growth Rate: 25% 4Q $292.5 $405.0 $465.0
Year $1,167.4 $1,405.1 $1,770.0
MktCap/Rev 596% 473%
We believe current expectations for Ascend's ATM revenue stream in 1998 lie in the
range of $300m to $350m and expectations for core switching for the same period range
from $500m-$560m. We believe these expectations to be considerably below the
realistic potential of Ascend's ATM and total core switching revenue in 1998. We
continue to believe that published consensus estimates will rise throughout 1998 and
1999 as carriers ramp deployment of the STDX9000, the CBX500, and the GX550.
Following are some examples of service provider contracts that we believe are not fully
recognized in today's consensus estimate for core switching at Ascend.
BancAmerica Robertson Stephens 1
ú Williams $150m-$200m
When the Williams non-compete agreement with Worldcom ended back in early January, Williams
announced their attentions to build a new nation wide data network for the purpose of wholesaling
ATM and frame relay bandwidth. The core network will be based on Ascend 550s populated with the
new OC48 cards (higher margin interfaces) and the edge network will be based on the Ascend 500 and
9000. In conversations with Williams, they indicated to us their intentions to build this network as
aggressively as they could and anticipated possibly spending another $50m-$100m on Ascend
switching equipment if they were able to keep on plan. As Williams ramps their wholesale offerings,
we believe they will aggressively move into retail data and will continue to place incremental orders for
Ascend switching equipment.
ú GTE/BBN $100-150m
There are a number of issues currently underway that are benefiting Ascend core switching at both the
ILEC side (GTE) and CLEC side (BBN) of GTE. At the traditional GTE unit, we believe that their 5
year $1 billion strategic agreement with Cisco has been canceled. We believe that going forward the
run rate of this relationship will be a maximum of $50m per year. In conversations with GTE, they
have indicated to us their desire to buy their core switching equipment from Ascend. Additionally, the
BBN unit recently announced a contract with Ascend for approximately 30 frame and ATM switches.
After conversations with them, we believe a more accurate representation of the contract size is in the
range of greater than 200 frame and ATM switches from Ascend and that a significant portion of these
are 550s.
ú Qwest minimum $100m 1998
While still remaining IP-centric, Qwest recently announced its intentions to deploy nation-wide ATM
and frame relay services based on the Ascend 9000 and 500 by this summer. Although the release did
not quantify the size of the deployment, we believe that their intentions are to offer the services in 125
cities by the end of the year. In order to roll-out 125 PoPs of frame and ATM services, Qwest will have
to build a nation-wide ATM backbone as well. This will be based on Ascend 500s and 550s as well.
ú Level 3 $50m-150m
Level 3, the other IP-centric carrier and very similar to Qwest, is aggressively lighting fiber and
building a data and voice network. While the name Level 3 indicates the company's desire to base their
network on the IP protocol, we have learned that Level 3 plans to start a large-scale ATM and frame
relay deployment in 1998. We believe that the switched portion of Level 3's network will be based on
Ascend equipment. Furthermore, Level 3 recently announced a $150m acquisition of XCOM - a
CLEC based in Boston. XCOM has a number of strategic agreements with Ascend and their entire
voice access and data remote access offerings are based on proprietary XCOM software that runs on
Ascend TNTs.
ú Bell South $100m
The CLEC unit of BLS is currently in the middle of designing a new inter-LATA data network in their
unregulated territories. We believe Cisco was recently awarded the layer 3 router portion of this build-out.
We believe the layer 2 switching RFP is sized at approximately $100m for ATM and frame
switches. Our understanding is that as of last week the short list has been narrowed down to Ascend
and Newbridge and the decision will be finalized very shortly. We believe the majority of the contract
2 BancAmerica Robertson Stephens
will be awarded to Ascend for a number of reasons. In recent conversations with BLS, the frame relay
and ATM networks in the regulated territory are based entirely on Ascend equipment and BLS has been
very happy with their choice and has no complaints. Additionally, the severity of the Cisco BPX failure
during last week's AT&T frame relay outage plays very well into Ascend's hands. The frame relay
network went down due to the BPX's inability to reroute traffic and handle signaling between switches.
Ascend has a distributed signaling architecture and is known as having the best rerouting capability
among ATM switch vendors. The Newbridge 36170 has a centralized architecture and has trouble with
both signaling and rerouting and the BPX has trouble with rerouting due to software issues. Last
week's events at AT&T (especially the tens of millions of dollars lost as they repay their customer's
SLA agreements) were very fresh in the minds of the operations guys at BLS when we talked to them
earlier this week.
ú AT&T minimum $50m 1998
We believe AT&T's intentions for 1998 are to build an ATM network with dimensions of a magnitude
that will stun the telecom community as AT&T's begins to realize its plan to migrate all of its services
to an ATM core. We believe T is currently in the midst of designing this network, but is waiting for
Lucent's 80 channel WDM product before they begin to deploy significant amounts of hardware. We
believe the ATM network will include Lucent Globeviews in the deep core, surrounded by a core of
Ascend 550s, surrounded by Ascend 500s and 9000s at the edge. We also believe the current Cisco
BPXs and Newbridge 36170s in the network will be pushed to the edge in order to support some of
AT&T's frame relay services.
ú Bell Atlantic $50m
Bell Atlantic recently announced their intentions to build an inter-LATA data network. To this end,
Bell Atlantic expects to spend $1.5B in 1998 on transmission equipment. Currently, a large majority of
the switching installed base at Bell Atlantic (and the original Nynex territory) is based on Ascend ATM
switches. In conversations with BEL, they have indicated to us their satisfaction with Ascend and have
hinted at future ATM deployment based on Ascend 500s and 550s. However, no contracts have been
awarded yet.
ú PSINet $50m
PSINet has a network comprised of 300 PoPs and approximately half of these have ATM and frame
relay switches from Ascend in them. Additionally, PSIX just announced the completion of a $600m
debt offering in order to raise cash for further network build-outs. In conversations with PSINet, they
indicated their intent to fill their other 150 PoPs with frame relay and ATM switches from Ascend.
Finally, PSI recently signed a contract with IXC Communications to receive 10,000 route miles of
OC48 fiber and we believe part of the upgrade at PSI will be for 550s in the core.
ú MCI minimum $30m 1998
MCI and AT&T have the biggest frame relay networks in the world. MCI is currently in the process of
upgrading their frame relay network which is comprised of 130,000 ports and more than 600 Bay BCN
routers. Because Bay never invested in their frame relay technology, MCI has decided to look
elsewhere for their upgrade - our understanding is that the upgrade will be based on Ascend 9000s and
500s. Additionally, we believe MCI is currently building out its ATM backbone. While there are
currently 22 Newbridge 36170s in place for ATM we believe future ATM upgrades will be at a
minimum of 50% Ascend equipment.
BancAmerica Robertson Stephens 3
ú LCI minimum $30m 1998
After completing an extensive evaluation process between Ascend, Newbridge, Cisco, and Northern
Telecom, LCI has standardized on Ascend for their ATM and frame relay deployments in 1998. After
conversations with the operations staff at LCI, we anticipate that the initial upgrade will be based on 30
550s and 28 9000s. We expect to see further announcements between Ascend and LCI in the Spring
Interop timeframe that will detail additional deployments of Ascend equipment.
Ascend is currently at the forefront a strong new product cycle in core switching that is leading to a
tornado in demand for their entire core product line. There is no other portfolio of ATM and frame
relay hardware and software that can compete - in fact, one large carrier described the current
competitive environment as "a field day" for Ascend. While systems vendors are always in a state of
`leapfrog' with their speeds and feeds on these platforms, we feel that Ascend has other means of
differentiating their products outside of a numbers-by-numbers comparison. Additionally, we believe
that an examination of the competitive landscape yields little in 1998 that could be considered
threatening to Ascend's core switching position within the industry.
As described in our ATM report, carriers are installing ATM for more than just a simple transport
mechanism. We believe that this benefits Ascend extremely well as Ascend is known among the
legions of service providers as having not only some of the best hardware available, but also the best
ATM software in the industry. As more of ATM's feature set is utilized, the software on a switch
becomes that much more important. For complex switching products such as these, and as other core
switching vendors have discovered, software can be the Achilles Heel of any hardware product. The
reason Ascend's software is best-in-class is that they very smartly took the code-base from their
original STDX8000 product and ported it first to the STDX9000, then later to both the CBX500 and
GX550 and therefore have gained 8 years of software debug whereas most competing products are still
in the first or second year of debug.
Due to the frame relay outage and the cause of the failure at AT&T last week, we feel that going
forward carriers will look very keenly on the rerouting and fail-over capabilities of any core switch they
are thinking about purchasing. This positions Ascend well, as their switches uniquely support rerouting
and protection capabilities similar to the underlying SONET infrastructure. Every line card has a
signaling and rerouting ASIC on it and there is also a `Smart-PVC' mechanism which polls circuits
every 80ms for error and reroutes on the fly.
Finally, a quick note about the hardware in the GX550. Carriers are extremely limited in the amount of
rack and floor space they have in their COs or PoPs and look favorably at any switch which is
competitive while maintaining a small footprint. We won't name any names, but most of Ascend's
competition have built or are building core switches which claim to have backplane bandwidths of
hundreds of gigabits per second or even terabits, but what they don't tell you is that to obtain this
capacity you have connect sometimes 10 to 20 different smaller capacity switches together!
Furthermore, we know of at least one vendor that is building a core switch which is currently the size of
a refrigerator! There are very few carriers we know of who would choose to waste that kind of precious
rack space. Lastly, it is one thing to build an ATM switch that has tens or hundreds of gigabits of
capacity, but it is extremely difficult to have low enough latency across the switch to yield a throughput
that matches that capacity. Typically, core switches comprised of multiple units trunked together have
eye-popping aggregate capacity, but very low throughput. We talked to many different service
providers who had GX550s in their labs filled with OC48 interfaces that were yielding throughputs in
line with the 25Gbps bandwidth of the switch by maintaining very low latency from input port to output
port.
For further explanation of the dynamics of carrier-class ATM and frame relay deployment as well as
the drivers of demand for ATM switches, please see our report titled "The Role of ATM in Broadband
Service Provider Networks" available later today.
BancAmerica Robertson Stephens maintains a market in the shares of Ascend Communications
and has been a managing or comanaging underwriter for Ascend within the last three years.
FOR ADDITIONAL INFORMATION, CALL YOUR BANCAMERICA ROBERTSON
STEPHENS REPRESENTATIVE AT (415) 781-9700. |
