You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Intel Q2 Non-GAAP EPS of $0.13 beats by $0.16.
Revenue of $12.9B (-15.7% Y/Y) beats by $760M but down 15% YoY
Forecasting third-quarter 2023 revenue of approximately $12.9 billion to $13.9 billion vs $13.28B consensus
Expecting third-quarter EPS attributable to Intel of $0.04
Non-GAAP EPS attributable to Intel of $0.20 vs. $0.17 consensus
Revenue attributed to its Data Center and AI group fell 15% year-over-year to $4B, while Network and Edge revenue dropped 38% to $1.4B. Intel's foundry unit saw a 307% surge in revenue to $232M.
"We remain focused on operational efficiencies and our Smart Capital strategy to support sustainable growth and financial discipline as we improve our margins and cash generation and drive shareholder value," CFO David Zinsner said in a statement.
“Our Q2 results exceeded the high end of our guidance as we continue to execute on our strategic priorities, including building momentum with our foundry business and delivering on our product and process roadmaps," said Pat Gelsinger, Intel CEO. "We are also well-positioned to capitalize on the significant growth across the AI continuum by championing an open ecosystem and silicon solutions that optimize performance, cost and security to democratize AI from cloud to enterprise, edge and client.”
Looking ahead, Intel said it expects third-quarter revenue to be between $12.9B and $13.9B, with the mid-point above the $13.28B estimate. It also expects adjusted earnings of 20 cents per share, topping the 13-cent-per-share estimate.
A good chunk of interest from AI companies looking at Intel's advanced packaging capabilities. Also, Gaudi and other AI accelerators hit 1 billion in pipeline.
Certainly it was a nice improvement, but it was also against a very weak compare. Note their Q4 2022 numbers, if I remember correctly they were much higher than this quarter's. Also note that Q2 2021 was higher than Q2 2023, I believe. So, it's anything but a pattern for them, it's one quarter.
In the earnings call, they said it was from higher packaging revenue, and higher IMF nanofabrication tools.
Certainly it was a nice improvement, but it was also against a very weak compare. Note their Q4 2022 numbers, if I remember correctly they were much higher than this quarter's. Also note that Q2 2021 was higher than Q2 2023, I believe. So, it's anything but a pattern for them, it's one quarter.
In the earnings call, they said it was from higher packaging revenue, and higher IMF nanofabrication tools.
There was also some talk about IFS having opportunities to sell advanced packaging for chips fabricated at other foundries. Some of that may have already happened, but it seemed more a future opportunity (and perhaps a transient one if IFS really becomes viable on the fab side).
I'm not sure if the change in depreciation accounting on fab equipment at Intel came through yet (I think they were extending the depreciation period from 3 years to 4). That's going to create a one-off bump in the P&L somewhere.
There was also some talk about IFS having opportunities to sell advanced packaging for chips fabricated at other foundries. Some of that may have already happened, but it seemed more a future opportunity (and perhaps a transient one if IFS really becomes viable on the fab side).
We already know about graviton, and I remember on prior earnings calls they talked about revenue being up from automotive customers. Considering MTK is the first wafer customer announced and they said that their intel 16 products would go into production in 2024, I think it is fair to assume those auto customers were there for intel's packaging. It also helps the packaging stuff is WAY faster to qual and design around than a full chip.
I'm not sure if the change in depreciation accounting on fab equipment at Intel came through yet (I think they were extending the depreciation period from 3 years to 4). That's going to create a one-off bump in the P&L somewhere.
You can't change depreciation models mid life. So the change in deprecation would only show up for the new tools that they chose to deprecate on the 8 year curve rather than the current 5 year (the 4 year curve thing was stopped during BS' tenure once they stopped converting fabs every 4 years). This also wouldn't really show up directly on IFS's P&L as IFS is just a customer facing support org. It should only indirectly show up as the price for IFS's services being influenced by the cost of said services. Until we get the unified TD-MSO-IFS P&L, we won't really have direct insight into those kind of finances.
"On Intel 20A, our first node using both RibbonFET and PowerVia, Arrow Lake, a volume client product is currently running its first stepping in the fab."
"For the remaining three nodes, I would highlight Intel 3 met defect density and performance milestones in Q2, released PDK 1.1 and is on track for overall yield and performance targets."
"Reinforcing customer confidence in our road map, Ericsson will be utilizing Intel's 18A process technology for its future custom 5G SoC offerings."
"For both Sierra Forest and Granite Rapids, volume validation with customers is progressing ahead of schedule. Multiple Sierra Forest customers have powered on their boards and silicon is hitting all power and performance targets."
"On Intel 20A, our first node using both RibbonFET and PowerVia, Arrow Lake, a volume client product is currently running its first stepping in the fab."
"For the remaining three nodes, I would highlight Intel 3 met defect density and performance milestones in Q2, released PDK 1.1 and is on track for overall yield and performance targets."
"Reinforcing customer confidence in our road map, Ericsson will be utilizing Intel's 18A process technology for its future custom 5G SoC offerings."
"For both Sierra Forest and Granite Rapids, volume validation with customers is progressing ahead of schedule. Multiple Sierra Forest customers have powered on their boards and silicon is hitting all power and performance targets."
Yeah, but one of the negatives of it all was, Intel didn't say much at all about customer's moving to their Intel 3 process, which by all indications is becoming a dud. Even so, there's apparently a lot of activity on 18A, so it's overall very good. Based on how Intel's narrative has changed over the last couple of years, I'd be pretty shocked to see significant activity before 18A, which pushes back their horizon for significant revenue for IFS. But, let's hope I'm wrong.
What I liked most is they said they continued to gain market share in CCG. But, how much can they take, when AMD has so little to begin with? I'd like to have heard some progress in GPU market share, but I guess given their limited range of products in that area, and the natural reluctance of people to buy Intel GPUs so early, it's not entirely unexpected.
Accounting is something of an art, and companies always save some accounting tricks – perfectly legitimate items that meet the discerning eye of financial standards – to goose their numbers when they really need it. When they really need to tell a turnaround story. And so it is with Intel in the...
"On Intel 20A, our first node using both RibbonFET and PowerVia, Arrow Lake, a volume client product is currently running its first stepping in the fab."
"For the remaining three nodes, I would highlight Intel 3 met defect density and performance milestones in Q2, released PDK 1.1 and is on track for overall yield and performance targets."
"Reinforcing customer confidence in our road map, Ericsson will be utilizing Intel's 18A process technology for its future custom 5G SoC offerings."
"For both Sierra Forest and Granite Rapids, volume validation with customers is progressing ahead of schedule. Multiple Sierra Forest customers have powered on their boards and silicon is hitting all power and performance targets."
"Looking specifically at each node, Intel 7 is done and with the second half launch of Meteor Lake, Intel 4, our first EUV node is essentially complete with production ramping. For the remaining three nodes, I would highlight Intel 3 met defect density and performance milestones in Q2, released PDK 1.1 and is on track for overall yield and performance targets." So, too early to tell for 20A.
"Looking specifically at each node, Intel 7 is done and with the second half launch of Meteor Lake, Intel 4, our first EUV node is essentially complete with production ramping. For the remaining three nodes, I would highlight Intel 3 met defect density and performance milestones in Q2, released PDK 1.1 and is on track for overall yield and performance targets." So, too early to tell for 20A.
I would consider Intel 3 and Samsung 3nm done as well but they missed the foundry window so there is no real ecosystem to support them. TSMC N3 already has everything you need and companies are taping out dozens of chips this year and hundreds will follow. My guess is that TSMC N3 will have 90% foundry market share when all is said and done. TSMC N2, Intel 18A, and Samsung 2nm is another story but today TSMC N3 is a much better process in all regards including maturity.
I would consider Intel 3 and Samsung 3nm done as well but they missed the foundry window so there is no real ecosystem to support them. TSMC N3 already has everything you need and companies are taping out dozens of chips this year and hundreds will follow. My guess is that TSMC N3 will have 90% foundry market share when all is said and done. TSMC N2, Intel 18A, and Samsung 2nm is another story but today TSMC N3 is a much better process in all regards including maturity.
Based on this chart, just about the only company that could love Intel 3, is Intel. It should give them a nice edge in performance over AMD, but it's difficult to see a lot of scenarios for it for other companies.
18A looks very attractive though.
Of course, this chart doesn't show the power efficiency, and that could be a really big factor as well, for both categories of processes.
But, clearly, Intel is not talking about 3nm as a foundry node much anymore, and pointing much more attention to 18A, so that surely indicates exactly what you are saying.
Power efficiency usually tracts with density but there is also a performance/power trade-off that must be made. It is difficult to get power numbers out of Intel but I can tell you from ecosystem experience TSMC is the power efficiency king thanks mostly to Apple. Power is critical to mobile SoC designs and Apple is critical to TSMC, absolutely.
I don't know if that is true anymore. Smaller gate pitch means worse subthreshold leakage, and lower active power. My understanding is that for mobile SOCs subthreshold leakage is the largest as well as the least desirable source of battery drain. After all nobody likes their phone dying when they aren't using it. My potentially flawed understanding is that most of the power reduction items these days are from the underlying changes to the transistor rather than pitch shrinking. Evidence for this can be found back when intel's old nodes had a density and technology lead vs TSMC/Samsung's best, and yet they could do lower power than intel's ATOM SOCs. Granted this isn't totally apples to apples because of the different archs and intel not having monolithic mobile SOCs during their big mobile push, but I think it gets the point across.
I'd go a step further. Low power is TSMC's legacy and is what most of their customers have always wanted. Just like the fabless industry they enabled, they started off in embedded, then moved to mobile, and relatively recently HPC became TSMC's growth engine.
The Intel foundry numbers include IMS mask making (More than 50% of revenue) and packaging (20%?) . The actual wafer foundry is quite small. I would assume the packaging is packaging of Intel chips. I believe the classic foundry (wafer sales) is still less than 100M per quarter. Is intel going to report external wafer sales foundry in the future or blur it into one number for manufacturing?
The Intel foundry numbers include IMS mask making (More than 50% of revenue) and packaging (20%?) . The actual wafer foundry is quite small. I would assume the packaging is packaging of Intel chips. I believe the classic foundry (wafer sales) is still less than 100M per quarter. Is intel going to report external wafer sales foundry in the future or blur it into one number for manufacturing?
Accounting changes, that makes sense. Significant external wafer sales are still years out.
Have you ever looked at Samsung's financials? Do they break out the foundry business? Internal versus external customers? I'm guessing Intel will do something similar. IDM foundries, lots of wiggle room in the financials.
Samsung's foundry business has got to be shrinking with the loss of QCOM and NVIDIA at 5/4/3nm. Quite a few other Samsung customers jumped as well. The yield issues at 5/4 killed them and Samsung 3nm is just not competitive with TSMC N3 yet.
Also, by pushing out the useful life of equipment from five to eight years, they will be saving $4.2 billion in depreciation expense in 2023. So, there's some accounting changes that make this quarter look profitable.
But, the biggest number is revenue, and that's up from Q1 and should be up again in Q3. So, at least that's positive, irrespective of paperwork (we hope, anyway).
Also, by pushing out the useful life of equipment from five to eight years, they will be saving $4.2 billion in depreciation expense in 2023. So, there's some accounting changes that make this quarter look profitable.
But, the biggest number is revenue, and that's up from Q1 and should be up again in Q3. So, at least that's positive, irrespective of paperwork (we hope, anyway).
"Also, by pushing out the useful life of equipment from five to eight years, they will be saving $4.2 billion in depreciation expense in 2023. "
Basically Intel is kicking the can down the road.
Changing the equipment depreciation schedule from five years to eight years will have immediate effects on increasing Intel's tax liability. On the other hand most companies that are making good profits like to shorten/accelerate equipment depreciation schedule. Doing it can increase their depreciation expenses and consequently reduce their tax liabilities in short term
Intel probably chose to extend its depreciation schedule because they don't see a great profitability potential in the next several years. Intel needs every dollar to make its accounting book pretty. Intel seems to be desperate.
www.intc.com
seekingalpha.com
