The problem for Intel is that the product mix headwind will remain mostly in place and will get worse as Intel launches its Lunar Lake CPU in late Q3.
Intel feels that Lunar Lake is a strong product and noted that it will power over 80 new Copilot+ PCs across more than 20 OEMs. Note that this compares unfavorably with the “100+ design wins” that Advanced Micro Devices (AMD) claims to have with its Zen 5 laptop CPUs. Never in the PC industry history has Intel gotten fewer design wins for its new product than did AMD. The limited adoption of Lunar Lake likely has to do with the timing of the chip’s release. September availability misses the back-to-school season and was likely a bit late for an aggressive Christmas toll out.
Lunar Lake, which Intel is excited about, is more competitive with AMD Zen 5 but is expensive to make as it uses chiplets manufactured at TSMC (TSM). Furthermore, Lunar Lake integrates a memory chip on package which Intel has to procure from memory suppliers and sell to OEMs with “0%” mark-up.
Unlike Nvidia (NVDA), which is making 80% margins on the HBM it is integrating into GPUs, Intel is getting no margin for integrating memory! Readers may wonder why this is the case. The answer is that AMD’s Zen 5 is a superior architecture and is able to deliver high performance without needing an on-die memory. A PC OEM buying a mobile Zen 5 CPU from AMD can buy memory directly from memory producers. To compete with AMD, Intel will have to sell Meteor Lake with “0” markup on memory. As such, Intel will have to eat the cost of memory in every package that has to be thrown away. This can be disastrous in the event Intel has yield problems.
With Lunar Lake being a major seller for H2 and much of 2025, even with good yields, margins will continue to be under pressure. As a result, Intel shifted its goal of getting back to 60% margins to 2026.
It is not too surprising that Lunar Lake has so few design wins as it has a very limited number of different products. Arrow lake releases soon after Lunar Lake to fill out the product line, and will have many more options and likely many more design wins. Lunar lake has 4 17W products with either 16GB or 32GB of RAM , and only one more premium at 30W and 32GB. All the “H” series performance laptop parts, as well as all the mainstream to low cost product will be arrow lake rather than Lunar Lake.
They are absolutely correct about the financial/margin information and one reason Intel stock tanked so badly. Intel said the product is not cost optimized at all, and with the DRAM on board the margins will be below the already pathetic Intel average. I would think this would be a little concerning to AMD investors in that Intel is willing to sacrifice in order to maintain market share.
The leaked benchmarks so far look pretty impressive for a 17W part, but we will know more when the official product launch happens September 3.
It is not too surprising that Lunar Lake has so few design wins as it has a very limited number of different products. Arrow lake releases soon after Lunar Lake to fill out the product line, and will have many more options and likely many more design wins. Lunar lake has 4 17W products with either 16GB or 32GB of RAM , and only one more premium at 30W and 32GB.
You raise a good point. It sounds like it was never the product you’d build a whole portfolio around, so… sure.
the margins will be below the already pathetic Intel average.
Reminds me of the days when Athlons and Durons were going for embarassingly low ASPs and we used to monitor those prices on … oh, what was it, pricewatch.com or something? that used to have the latest price on every SKU
I think Lunar Lake might be one of the most expensive products out there, but also the most costly to build. It is on the more expensive TSMC N3 node, and has both a fairly beefy GPU and NPU. It also has extra L4 cache to reduce memory access, and hence power. Everything about it is optimized for reduced power rather than cost. It is really a proof of concept that X86 can get better performance at lower power than the ARM contenders.
Intriguing but in the position Intel is in, it’s tough to defend bringing a product like that to market - unless you really, really need to prove the concept.
I had the impression that Lunar Lake was aimed at something like the M3 in Apple’s lineup – with the integrated memory etc… Can they prove they can get comparable performance by implementing such an architecture? What is the x86 tax, really, all else being equal? But even if that lesson needs proving out, do you really need to ship it? I dunno. Perhaps it’s an artifact of Intel thinking they had more money and running room than they actually did.
It certainly fits in the “if we knew then what we know now” category.
However, this could be another “banias” moment, if you recall the shift from Pentium 4 to “core”…
Alan
I certainly do. The Beginning of the Beatdown.
The fact that Intel has backed away from Keller’s Royal Core redesign makes me think we don’t have to worry about that. At least not for a minute.
In all seriousness: If Intel can crack the 18a process, they could become a threat again, no two ways about it. As long as AMD doesn’t pick another Bulldozer design, though, I feel reasonably good about it. And the whole ARM shift in the market suggests that even head-to-head a revived Intel is not going to get into a one-on-one knife fight with AMD in an elevator.