So…Intel…for what it is worth. ( Local spot of price of 2 increasingly copper less pennies )

1: Ditch all the foundries. Now. Either sell them for what you can to GloFlo or write them down. Walk away. Quickly. By the end of the year 2021. Go with a combo of TSMC, Samsung and GloFlo.

2: Create a ten year plan to phase out of x86/x64. The time has come to pull down the curtain on that architecture and let it die as so many other architectures that have come and gone after they have outlived their usefulness for a future that never sleeps and demands changes and progress that those archs can no longer promise nor give.

3: RISC is your game now. Pick one. Either join the OpenPOWER foundation and drive further innovation with that Arch and ISA or….

Join the RISC-V Org, outright buy SiFive and Canonical to get Ubuntu and turn yourself into an IBM and become a turnkey, integrated Platform Solution Provider.

4: Forget ARM. You made the mistake of letting that go decades ago. So did AMD. Let AMD go down the drain as the sole remaining, big, hot iron x86/x64 arch designers. Once x86/x64 hits 3nm, it’s over. If anyone can even get there at scale needed to be profitable.

OpenPOWER, or RISC-V and a combo of your efforts in FPGAs such as Altera and your Neuromorphic efforts could make quite the compelling story leading into the 2030’s.

Intel stick to manufacturing claim: National Supply chain safety reasons only. That’s why TSMC and Samsung are evaluating the possibility to get local footprint.

Intel entering public cloud: I would only add the ‘all hyperscalers are doing their own CPU/DPU/AI chips these days anyway’ so time for action is running out

Intel buying ARM?! : train de-trailing sound/mind explodes! 🙂 Since we are living in the Intel outsourcing CPUs to TSMC world, why not!

Financially Intel monopoly mass is unsustainable lacking cost optimization in line with revenue potential and current generation that has nothing to do with design manufacturer but a continued reduction in unnecessary costs, margin drag and the continued elimination of financial leaks that have robbed the enterprise of capital supporting plant, production, equipment, construction, research, development and brain trust.

Similar to AMD 35% annual gross revenue sacrifice supporting bundle deal sales close incentive, reclaiming Intel bundle deal averaging $40 billion annually over the prior decade is the surest method to increase both enterprise s gross revenue overall for enterprise valuation and into future investment.

Sustainable business dictates that give-away-in-sales-package and the unnecessary costs of production go.

Intel reports earning today will certainly be invented. “Throwing off cash”, how about letting fewer components out the door for nothing in sales package. Or eliminating that production cost speaks a lot about area optimized dice that rely on Intel mask shop incredible in-house competitive cost advantage.

Pursuant update on lithography advance Intel has successfully harvested 14/12nm and 193 quad patterned immersion at a similar competitive cost : price / margin to TSMC 7 nm that has nothing to do with physical component performance. This strategy of trailing one node for learning waiting leap frog is not cost competitive trailing at two nodes behind where channel inventory data places Intel 10 nm at 20% of current production and I’ll throw in another + 10% that worldwide channel inventory currently does not see. I thoroughly expect Intel to will remain a harvesting follower trailing one node behind for learning to look out for inordinate factor cost increase, node to node, while aiming to leap frog and most likely in packaging. By trailing one node Intel benefits competitively on back half of run production cost : price / margin while fabrication competitors shoulder the brunt of next factor cost increases. To do this means 10 nm production at Intel must be peak volume production capable now.

The trick is for Intel to cost and profit optimize every time competitors attempt to profit maximize ramping on a more costly new node. There after Intel does what Intel always does when ramping a new node, by following other’s learning and the initial factor cost increase, copy what works for Intel and leave the garbage behind. This may not necessarily be innovative in terms of volume production but on Intel volume keeps the operation close enough to leap frog technically when that opportunity arises.

It’s really just good ole Intel doing what Intel has always done best, borrowing. Plus Intel is a platform business and this conversation neglects those aspects of Intel’s core business essential to systems integration.

This analyst anticipates + 10% volume from what can’t be seen in channel now is Xeon Ice Lake where any lack of 10 nm Xeon volume looses a critical cost : price / margin contributor essential for shoring up enterprise financial structure. If Xeon + 10% on Core 20% of current production means by 2021 end of year 2/3rds of all Intel production is 10 nm. Where without 10 nm volume, Intel Xeon product line price support pulling up all of Core gross margin, Intel’s flagging monopoly house of cards comes tumbling down. Something 14 nm Scalable kibble even if for keeping customer enterprise information tech and the business of compute, as it exists today humming along, cannot shore up Intel by category product line gross margin without that Xeon price support.

Considering Intel’s 10 nm manufacturing plan we’ve seen Big fin Tiger Lake quad, now so said on CPU specification + 39% base, + 7% all core, + 14% max frequency boost with power range less 20% to + 40% higher frequency and the soon anticipated octa implementation. In incremental steps Big fin frequency appears to have improved as much as 30% over Small fin Ice Lake 10 nm where at Ice Lake U up to 4.1 GHz i7 is not produced on channel view however maxes at 3.9 GHz. All about a good layup and always has been.

Average weekly ramp suggests improvement; over the production runs to date Comet Lake U + 20% per week, Ice Lake U + 17% per week, and TL quad + 37% per week except in the 15th week of supply Big fin Tiger channel volume is just 60% of Small fin Ice in the same supply week. Quad + power + area appear to be Intel’s recipe for bringing a well laid-up 10 nm component to market. In relation Ice lake U range 15 to 28W supply volume over the next 10 weeks everyone will know if Tiger Lake quad range 12 to 28W can ramp.

Pursuant Xeon, at 14 nm there is no doubt Intel on yield assessment Intel can produce a 16 core part in very high yield. Where fortunately, for most commercial data applications CPU max frequency wanting system stability and power optimization on utilization is not the top check box; stability, price and cost of power for core/thread optimized applications is, and for applications effectiveness across the entire commercial market 18C processors or less still make up 83% of all Intel production and 45% of all AMD production. Specific to Intel speaks for the need of area optimized mask sets, or disaggregate you chose.

Intel Scalable Skylake, Cascade lake and CL refresh full run by core grade;

4C = 8.13%
6C = 4.77%
8C = 23.76%
10C = 9.69%
12C = 15.17%
14C = 4.89%
16C = 9.81%
18C = 6.6%
20C = 4.93%
22C = 1.42%
24C = 4.06%
26C = 2.05%
28C = 4.71%

AMD Epycs all inclusive

8C = 19.35%
12C = 0.67%
16C = 25.46%
24C = 16.43%
32C = 34.01%
48C = 0.52%
64C = 3.55%

For Epyc Rome only;

8C = 14.51%
12C = 3.39%
16C = 15.67%
24C = 11.88%
32C = 31.92%
48C = 2.93%
64C = 19.71%

Specific commercial dGPU, salvaging Scalable lakes from data processing to dGPU accelerated compute cluster is the largest secondary market refurbishing opportunity ever to come along in the history of compute. At 300 million units deployed total available market, everyone’s got their eye on this including innovation replacements to dGPU accelerated compute.

On farming out, Intel fabrication and production operation is optimized to secure Intel margin that is not TSMC or Samsung margin. Albeit there is the competitive producer margin squeeze. Disablement has never been cost competitive for Intel and since 22 nm places a drag on margin. You can’t be good at producing a many cores component only to disable it for down bin offerings and the full product line card.

At Tiger Lake and for specific Scalable components Intel appears to have recognized this reengaging the mask shop for area optimized designs. AMD is in a similar situation with APUs where the cost of Renoir disablement for down bin availability, 4/2 cores from an 8C die, is prohibitive on the cost of disablement on the overall loss of full run margin. This has been one of the reasons Intel contracts out bottom of the product stack and legacy node components. Personally I think we will see less of this rather than more. The only reason for Intel to contract fabrication is associated with loss of capacity resulting from a total revenue loss, financial leaks in the enterprise, occurring the last decade and escalating in the prior year. One can see the amount of low core and high core count Scalable lakes bundled into sales package and shipped but not as revenue units.

“Think Intel peak happened a few years ago?” On channel supply volume data Intel production peaked between January 2018 and September 2019 at approximately 856 M units annual primarily Xeon is now down into 350 M unit range. The majority of said Xeon volume escaped as sales close in bundle deal and through 2020 as down bin low core count components. Bundle deal revenue recognition, where few recognized said Xeon revenue down the supply chain that if recognized, and paid taxes on, would have more than covered the price of before now addressing Intel’s current capacity conundrum making that issue moot. Which gets back to not throwing off cash, but preserving capital for enterprise investment.

“But as Intel’s might in the datacenter has grown, its paranoia has kicked into overdrive and it has been a bit acquisitive without being as focused as it needs to be”, which should be focused on sustainability of the entity over tying, bridging and loosing production values. On acquisitive, Intel divisions need to stand on their own balance sheets because DCG’s ability to subsidize all operating divisions escaped out the back door. This analyst firmly believes Intel acquisitions going forward need be focused on the core of the business which is design fabrication and component product manufacturing that is not lateral or soft beyond tools vertical to the need of the enterprise, and since down channel Intel already maintains distribution sales tentacles any Intel acquisition should be focused on materials science, manufacturing tools development, packaging is a big movement now and associated research. Not to monopolize the upstream tool chain but to assure equipment deployed in Intel manufacturing facilities works in unison with Intel design tools and Intel design methodology. Intel fabrication process and Intel tools cannot be separated which appears partially responsible for the at 10 nm difficulties. Getting rid of Intel fabrication and/or refocus on foundries resolves none of the Intel core business issues simply extenuates them creating more issues because Intel tools and design fabrication process are so tightly coupled.

Intel Annual R&D as % stated revenue and % change in the value of PE&C net

1993 = 10.93% = base
1994 = 9.64% = + 34.31%
1995 = 8.00% = + 39.20%
1996 = 8.67% = + 13.60%
1997 = 9.36% = + 25.67%
1998 = 9.55% = + 8.84%
1999 = 10.59% = + 0.91%
2000 = 11.55% = + 28.15%
2001 = 14.30% = + 20.70%
2002 = 15.07% = – 1.51%
2003 = 14.47% = – 6.65%
2004 = 13.97% = – 5.36%
2005 = 13.25% = + 8.52%
2006 = 16.60% = + 2.87%
2007 = 15.03% = – 3.89%
2008 = 15.22% = +3.70%
2009 = 16.09% = – 1.82%
2010 = 15.43% = + 3.91%
2011 = 13.92% = + 32.00%
2012 = 19.04% = + 18.44%
2013 = 19.91% = + 12.31%
2014 = 20.65% = + 5.76%
2015 = 21.91% = – 4.15%
2016 = 21.45% = + 13.54%
2017 = 20.87% = + 13.65%
2018 = 19.12% = + 19.14%
2019 = 18.57% = + 13.09%

Intel Plant, Equipment and Construction Y/Y % contribution or decline and % of revenue;

1993 = base
1994 = 25.54% and 11.90%
1995 = 28.16% and 12.99%
1996= 11.97% and 4.87%
1997 = 20.43% and 8.69%
1998 = 8.12% and 3.59%
1999 = 0.90% and 0.36%
2000 = 21.97% and 9.78%
2001 = 17.15% and 11.71%
2002 = – 1.54% and – 1.02%
2003 = – 7.12% and – 3.93%
2004 = – 5.66% and – 2.61%
2005 = 7.85% and 3.46%
2006 = 2.79% and 1.39%
2007 = – 4.04% and – 1.79%
2008 = 3.57% and 1.67%
2009 = – 1.85% and – 0.91%
2010 = 3.77% and 1.58%
2011 = 24.24% and 9.55%
2012 = 15.57% and 8.17%
2013 = 10.96% and 6.46%
2014 = 5.45% and 3.24%
2015 = – 4.33% and – 2.94%
2016 = 11.92% and 7.26%
2017 = 12.01% and 7.87%
2018 = 16.06% and 11.10%
2019 = 11.57% and 8.91%

So what happen? Since 2016 Intel is investing in plant, equipment and production but in what category and where, and how do acquisitions within PE&C line item contribute to design, process and the overall manufacturing core and whole platform competencies? It‘s time for Intel to invest in the core business design, process, and manufacturing similar to 1995, early 2000’s and 2011. On all that product revenue that snuck out the back door, places Intel implementing CEOs 9 month, 18 month, 3 year within 5 year plan on an accelerated basis that, in my opinion, focuses on the core of the business that is not other enterprise’s business. God ole Intel only procures from other sources to get a look inside. And while there are cost efficiencies matching product and process, that’s Intel’s business ultimately. And if Intel foundry is part of that solution a federated fabrication cluster divested from its product divisions opens up a potential Intel fabs will produce other than Intel volumes they will never walk away from. Choosing in addition to Intel and working with customers on what Intel manufacturing can produce profitability within sustainable frameworks.

Certainly Intel customers in hyper scale, cloud and other businesses of compute benefit from Intel vertical platform focus to some degree today and can Intel extend from their own derivatives to produce the wholly unique and original designs of others? Of course they can, tightly coupled with Intel component and within platform solutions.

Pursuant Cloud, well Intel has a cloud, it’s called Intel Cloud. Where just like Intel fabs, Intel cloud is closely coupled to produce Intel design on Intel tools so in my mind, anything else diverts that resource from its core intent and Intel’s core business.

“The key thing for Gelsinger to do – and that Intel’s founders and managers did successfully many decades ago – is to honestly and brutally assess what is key to Intel’s future and what is not”. Where Intel is so bound up and tightly coupled with itself the only answer is to stick Intel. Sell off the fabs and Intel will go broke.

Modifying the manufacturing relationship under a federated operating structure enables individual fabs to do what they know works within the overall Intel volume envelope. “You have to dig in.”

Yea, its all been about construction, outfitting and getting back to work.

Mike Bruzzone, Camp Marketing

2. Outsource chip manufacturing to TSMC, Samsung and etc. Buy their share, or whatever, just abandon that sinking ship.

3. Do not get lured into cloud business. Do not compete with itself, by competing with its own potential customers. Do not enter IAAS cloud market. Try to form alliances instead, or let the open hardware shape the market ahead.