Intel Q3 2015 Earnings Call Discussion

[Daniel Nenni]

I would be interested in a meaningful discussion here. Intel is a semiconductor bellwether and one of the most interesting companies to follow. As soon as the transcript of the call is up I will post it here. Until then please share your thoughts...

The audio is up:

Intel (INTC) Q3 2015 Results - Earnings Call Webcast | Seeking Alpha

• CAPEX: Intel has cut its 2015 capex budget for the fourth time: It's now at $7.3B (+/- $500M), down from $7.7B in July and $8.7B in April, and below a 2014 level of $10.1B.
• Gross margin: Q3 gross margin was 63%, +50 bps Q/Q and -200 bps Y/Y, and in-line with guidance. Q4 GM guidance is at 62% (+/- 2%). Higher ASPs lifted Q3 GM, while higher platform unit costs weighed. Higher unit costs and factory start-up costs are expected to weigh on Q4 GM, partly offset by lower write-offs.
• PC/mobile CPUs: Client Computing Group revenue -7% Y/Y to $8.5B, thanks largely to a weak PC market. Op. profit fell 20% to $2.4B. Volumes +3% Q/Q and -19% Y/Y. ASPs +9% Q/Q and +15% Y/Y. Desktop volumes -15% Y/Y, notebooks -14%, tablets -39%; all three segments had higher ASPs.
• Server CPUs: With Web/cloud demand and the Grantley Xeon CPU launch still acting as tailwinds, Data Center Group revenue rose 12% Y/Y to $4.1B. Op. profit rose 9% to $2.1B. Volumes +7% Q/Q and +6% Y/Y. ASPs +1% Q/Q and +6% Y/Y.
• Other segments: IoT Group revenue +10% Y/Y to $581M; op. profit +4% to $151M. Software/services revenue flat at $556M; op. profit up over 3x to $102M. Other (NOR/NAND flash, devices, one-time costs, etc.) revenue +19% to $682M; op. loss of $621M.
• Financials: $1B was spent on buybacks, up from $697M in Q2. Thanks to lower R&D spend, R&D/MG&A spend was flat Y/Y at $4.8B ($100M below expectations). The tax rate was 26.9%, 90 bps above expectations. Intel ended Q3 with $20.8B in cash/investments ($10.3B offshore), and $21.2B in debt.

Transcript:

Intel's (INTC) CEO Brian Krzanich on Q3 2015 Results - Earnings Call Transcript | Seeking Alpha

 

[Robert Maire]

• Intel- good Q - In Line guide- DCG slower growth
• PCs stabilizing- Better ASP mix driving revenues
• $400M capex cut due to tool push into 2016...

Overall good qtr & progress- some pluses & minuses- worst seems behind

Solid Q3, In Line guide..better ASPs offset by slower DCG...
Intel reported EPS of $0.64, a nickle better than street estimates of $0.59. Revenue of $14.47B beat by $250M. Q4 guide is in line at $14.8B.

Overall a good beat. ASP mix was better than expected which led to better revenue on subdued unit volume. On the negative side DCG growth expectations which were a bit high at 15% have been reeled in to low double digits.

We would imagine the stock will trade flattish to slightly down on DCG concerns.

Importantly, tablet losses are shrinking, modem progress continues. The company continues to be driven by servers, the cloud and internet of things. While PCs continue to retreat Intel has done a decent job of mitigating the damage by getting better ASP mixes with specifically targeted products.

$400M capex cut takes capex down to $7.3B for 2015..

Could ASML be reason for the cut???
Intel said the capex reduction was primarily due to the pushout of a tool from 2015 into 2016. They alluded to a "tool" that was being reconfigured (tuned) for more unit output per tool. They also said that most of capex was now going for 10nm ramp and production. Management referred to it as a single tool type and focused on improved throughput which leads us to believe that it was likely ASML EUV tools that were pushed out of 2015 into 2016 with configurations (source power, debris mitigation??) to increase throughput closer to usable levels, which has obviously been the long term issue.

If this is the case it could cause a hole in revenue expectations for ASML for Q4 which we could hear about tomorrow.

The only other potential we could think of might be a KLA tool as they are expensive as well but we think managements reference on the call which was to unit throughput sounds a lot like ASML EUV problems. Maybe this is the first of the alleged 15 unit order slipping as the tools don't make spec....

On the positive side, Intel said capex for 2016 should be higher. Its unclear if its higher primarily due to orders slipping into 2016 or just 2015 being unusually low.

14nm still a problem...
Intel said there will be gross margin pressure in Q4 due to higher expenses (read that as poor yield) associated with starting up 14nm production in Ireland. Some of the gross margin pressure will be offset by lower write offs.

We are a bit surprised as it sounds like Intels copy exact is perhaps not so exact as the idea of copy exact is to work out the bugs at the first fab then "clone" the process everywhere else.
This suggests to us that the 14nm process flow is still flaky and not refined and nailed down enough that minor variations in copy exact can make it go off the tracks.

If Intel still doesn't have 14nm nailed and Samsung and TSMC are pumping out similar geometries we may be back to a closer race to 10nm (even if Samsung parts are power pigs...)

All the right moves...
As compared to where Intel was a while ago we think there is good progress on most fronts. Intel has circled its wagons around the golden goose of DCG, PCs are soft but revenues are stable. Tablet losses are diminishing. Finances are in good shape with a $1B buyback, a nice dividend.

As long as Intel doesn't screw up the Altera merger they could continue to see nice improvements and positive compares.

The stock is not expensive and has had good momentum over the last few months. If we had a reasonable macro recovery, it would be all the better for Intel given the potential upside in server/cloud spending. We would continue to own the stock and be positive on it. If there were too much of a negative reaction we would likely buy into that.

Robert Maire
Semiconductor Advisors LLC

 

[user nl]

$400M capex cut takes capex down to $7.3B for 2015..

Could ASML be reason for the cut???
Intel said the capex reduction was primarily due to the pushout of a tool from 2015 into 2016. They alluded to a "tool" that was being reconfigured (tuned) for more unit output per tool. They also said that most of capex was now going for 10nm ramp and production. Management referred to it as a single tool type and focused on improved throughput which leads us to believe that it was likely ASML EUV tools that were pushed out of 2015 into 2016 with configurations (source power, debris mitigation??) to increase throughput closer to usable levels, which has obviously been the long term issue.

If this is the case it could cause a hole in revenue expectations for ASML for Q4 which we could hear about tomorrow.

This morning ASML reported sales exactly as guided:

VELDHOVEN, the Netherlands, October 14, 2015 - ASML Holding N.V. (ASML) today publishes its 2015 third-quarter results.

• Q3 net sales of EUR 1.55 billion, gross margin 45.4 percent, in line with guidance
• ASML guides Q4 2015 net sales at approximately € 1.4 billion and a gross margin of around 45%
• Both the TWINSCAN NXT immersion lithography platform and the NXE Extreme Ultraviolet (EUV) platform upgraded for next-generation chip production

https://www.asml.com/asml/show.do?lang=EN&ctx=5869&rid=52453


The point is that ASML makes iArF tools that take some 3-6 months before being shipped. The EUV tools take even longer (6-12 months) between `real order` and shipment. So, ASML always has a very clear view on the next quarter of revenue regarding new tools being shipped. The iArF revenue is directly booked when the tool is shipped. The new EUV tools (NXE3350) have much more complex revenue recognition depending on certain milestones to be met, so that revenue comes much later in the books, up to a 6-12 months later or so it seems presently for the EUV NXE3350. They can play with that also a little to make pretty predictable next quarter guidance figures is my impression.

In July 2015 they expected to ship 5 NXE3350 before the end of year. Now they stated only 3 NXE3350 will be shipped due to customers (like INTEL?) pushing out the delivery of the tool. I guess at the same time this gives ASML more time to equip the tool with the latest improvements regarding availability that is currently the main issue for the EUV tool.

ASML said logic customers are somewhat delaying their shipments, so they expect Q4 of 2015 to be with sales around 1.4 BEuro (Q3 was 1.55 BEuro), so no big holes, they still have a backlog of orders of some 2.88 BEuro.

User nl

 

[user nl]

Just a followup from listening to the ASML conference call this afternoon CET:

1) the first shipment of the EUV NXE3350 (that is done now, I would guess to TSMC as they ordered first in fall of 2014) was delayed by some 2 months due to integration-delay problems with the drive laser (coming from Trumpf via Cymer). That is the big CO2 laser that fires on the Sn droplets to generate the EUV in that plasma.

2) the push-out of the 400 M$ Intel capex seems to be due to the fact that the 10 nm ramp is taking some more time and resources of people (Wennink said in ASML conference call). So there seem to be simply not enough people at Intel that can concentrate on using the NXE3350 due to issues with the complexity of 10 nm and yield. They need all resources (people) to fix that problem, in the heat of the competition around 10 nm with Samsung and TSMC (for prestigious reason one would think). Intel didn`t want ASML to ship those NXE3350 tools in 2015 just to sit around in their fab idle, while the Intel people were working hard on the 10 nm multi-patterning/yield issues.

So, the complexity of 10 nm seems to stretch out a little the use of NXE3350 at Intel and saving some capex in 2015 pushing it to 2016.

User nl

 

[Daniel Nenni]

It appears from the transcripts that Intel mobile is slowly winding down. Mark Hibben did a nice job summarizing:

Intel Cuts Contra Revenue, Tablet Sales Plummet - Intel Corporation (NASDAQ:INTC) | Seeking Alpha

Intel needs to stop propping up the antiquated x86 architecture and adopt ARM for mobile devices. The irony of Intel's current approach to mobile is that the combination of Intel's process leadership and the ARM v8 architecture could make Intel the unquestioned leader in mobile devices.

 

[Brianhayes]

"Intel needs to stop propping up the antiquated x86 architecture and adopt ARM for mobile devices. The irony of Intel's current approach to mobile is that the combination of Intel's process leadership and the ARM v8 architecture could make Intel the unquestioned leader in mobile devices."

Mark Hibben is a most entertaining writer and makes you want to agree with most of what he says, but I have just finished listening to the TSMC earnings conference and their description of their roadmap for producing Apple's products down to 7nm. Intel could become the unquestioned leader in ARM mobile SoCs? I wouldn't place a bet at 50:1.

Mark is right, however, as anyone who thinks Intel's management is first rate should read or listen to both presentations. No prizes for guessing which team gave a clear, informative, transparent and honest explanation of their company's position in what now looks like an industry recession.

 

[astilo]

14nm still a problem...
Intel said there will be gross margin pressure in Q4 due to higher expenses (read that as poor yield) associated with starting up 14nm production in Ireland. Some of the gross margin pressure will be offset by lower write offs.

We are a bit surprised as it sounds like Intels copy exact is perhaps not so exact as the idea of copy exact is to work out the bugs at the first fab then "clone" the process everywhere else.
This suggests to us that the 14nm process flow is still flaky and not refined and nailed down enough that minor variations in copy exact can make it go off the tracks.

They still do copy exact. Anyway, since the 22nm, they let the HVM sites doing some process tuning work on the fly (but I also think that 14nm was released a bit earlier to the HVM sites this time). They need to bring a process in volume as soon as they can, otherwise the cost of every node is simply too big. Not a side note. The Ireland bet is for sure huge. They had to move from a robust and state of the art 65nm node, straight to a tricky and not yet mastered 14nm process, skipping 45, 32 and 22nm. That's not a joke, and a clear management mistake if you ask me.

 

[benb]

Adding a couple of thoughts to this excellent discussion:
- Maybe Fab 68 and/or 11X will be put to use in IMFT for Crosspoint memory
- The modem wins are, for the first time, wins for Intel fabs and not TSMC fabs. Those are volume parts previously made on a foundry process. When those parts land inside Intel, that site becomes "the Intel foundry site". An interesting development.
- On a negative note: We heard for 3 quarters "PC demand will be flat...blah blah backloaded blah." That ended up -7%ish. It appears that "backloaded" is code for "probably don't listen to my wishful thinking".