I just recorded the video segment of the Top Stock Picks of the Week and my investment idea was Micron (MU Quick QuoteMU - Free Report) , the megalodon of memory.

 

In the video, I shared some of my Friday note to Zacks Ultimate members where I described the tectonic shift that just happened in the semiconductor industry, causing earthquakes and tsunamis in the geology of three stocks: Intel (INTC Quick QuoteINTC - Free Report) , Advanced Micro Devices (AMD Quick QuoteAMD - Free Report) , and Taiwan Semiconductor (TSM Quick QuoteTSM - Free Report) .

 

The bulk of that report follows below -- where I explain the seismic shift and why INTC was down 16% Friday while AMD was up an equal amount -- but what I forgot I wanted to do in today's video was give a big shout to my colleague Dan Laboe.

 

Dan has been extremely bullish on Taiwan Semi (aka TSMC) all year and produced several pieces of content on the company throughout the Coronavirus Crash and subsequent rally in the March through May period -- when shares were still trading near $50.

 

He saw their go-to foundry business as a cash machine, where unique chip designers like NVIDIA (NVDA Quick QuoteNVDA - Free Report) , AMD and Apple would simply contract TSM's advanced semi fab facilities to build their hardware with state-of-the-art 7-nanometer transistor technology.

 

In May, here were several of his reports and insights in articles and in his excellent video blog, The 4th Revolution...

 

 

 

 

 

 

So if you were really listening to Dan in May, you would have been buying TSM shares near $50.

 

I was listening, but not really, seeing TSM's $270 billion market cap as a headwind since it traded at over 6 times forward sales estimates -- vs Intel's $250 billion trading near 3X sales.

 

And most Wall Street analysts weren't that bullish either, probably for the same valuation concerns that ignored the shifts that were actually occurring in major semiconductor industry trends.

 

As recently as July 10, Susquehanna reiterated their "Sell" rating and $40 price target on TSM shares.

 

How wrong we were to miss the gold mine unfolding for TSM -- that only a few analysts saw, like Matt Bryson at Wedbush (more on his April initiation coming up), and Dan.

 

And last Thursday's Intel quarterly report delivered the proof (as you'll see in my story below) when CEO Bob Swan had to admit they were as much as one year behind on their internal roadmap for developing their own 7nm technology in-house.

 

But even before the shocking 10% rally in TSM on Friday -- and another 10% rally today -- Dan was out early last week reiterating his bullish views on TSM with these reports...

 

 

 

When a tech stock gumshoe like Dan has this much conviction on an investment idea, it pays to find out why.

 

So stick with Dan and give him a thumbs up on his Fourth Revolution videos! (link to Zacks YouTube archives)

 

In addition to Dan's excellent research, here was an investment bank view where the analyst employs a detailed revenue and earnings model...

 

Wedbush analyst Matt Bryson initiated coverage of Taiwan Semiconductor (TSM Quick QuoteTSM - Free Report) with an Outperform rating. In his April 27 research report, the analyst cited numerous secular drivers that should result in more demand for semiconductor capacity, including growth in AI related applications.

 

Bryson also sees both China's desire for semiconductor independence and Intel's recent struggles to advance its manufacturing as trends that should push relatively faster growth for foundries and particularly advanced process node capacity, including 7nm and 5nm.

 

He believes TSMC's market share and technological leadership in the foundry space position it to be the primary beneficiary of these trends. While acknowledging that U.S. trade and technology IP policy decisions with regards to Huawei could cloud this year's visibility, the analyst expects both factors will have modest impact on TSMC's longer term prospects as it adapts to a diverse and growing global customer base including NVIDIA, Apple, and AMD.

 

Now here was part of my commentary to my TAZR Trader group on Friday to explain the semi seismic shift...

 

 

Tonight, I want to focus on exactly what the heck happened with Intel that dropped its shares 16% -- and ignited an equal sized rally in AMD!

 

Here's what I began to write you this morning as I almost sent a Buy Alert for more Micron (MU Quick QuoteMU - Free Report) under $50...

 

The amazing storm in Semis today -- with INTC -16% and AMD +16% -- revolves around Intel's mea culpa that they are as much as one year behind on their internal roadmap to roll out significant 7nm (nanometer) technology.

 

This means that the leading semiconductor innovator who makes almost everything in-house will probably have to outsource to Taiwan Semi and others to meet demand next year.

 

In any case, it's great for NVDA as well as AMD. 

 

And even though MU has bigger hurdles with DRAM and NAND dropping down from 10nm to 7nm, they will probably get there before Intel.

 

(end of my draft Buy Alert for MU)

 

The reason I didn't pull the trigger on more MU under $50 is because I wanted to do more homework on just when they will have more visibility on sub-10nm capability. I found some things out as they partner with a little private "chipper" named Achronix. More details to follow there.

 

This was very good to learn so that we know that the Taiwan Semi fab can't own everyone in the space. (TSM shares were up 10% today on this INTC semi-debacle-rotation).

 

So let's look at what Intel promised and how they failed to deliver, causing a rush to the competion.

 

You may recall in our numerous discussions of NVIDIA technology -- and most recently after their May GTC gig -- that they were pushing the envelope of nanoscale architecture in semiconductor engineering.

 

The essence of Moore's Law and how NVDA reinvents it is the ability to dive deeper into the nano-sphere and leverage speed with massively parallel architectures.

 

And the engineering teams at NVIDIA, led by visionary CEO Jensen Huang, are conducting deep R&D in their GPU chips to leverage not just video game advances, but also the bleeding edges of AI.

 

We know that NVIDIA has already contracted Taiwan Semi for various 7nm applications, like their new Ampere A100 GPU board for data centers with an amazing 54 billion transistors. This is the next-gen power level for exascale supercomputers and AI research.

 

According to the gang at TomsHardware.com in a May 14 article, "Nvidia basically couldn't make a larger GPU, as the maximum reticle size for current lithography is around 850mm square. The increase in transistor count comes courtesy of TSMC's 7nm FinFET (fin field-effect transistor) process, which AMD, Apple, and others have been using for a while now. It's a welcome and necessary upgrade to the aging 12nm process behind Volta."

 

And here were whispers in April that NVIDIA might be going deeper yet, courtesy of TechRadar and DigiTimes...

 

By Darren Allan April 24, 2020

 

As it ups orders on 7nm, hopefully indicating RTX 3000 GPUs are still on track

 

Nvidia is making something using a 5nm process, according to the rumor mill, although it’s anyone’s guess what that hardware could be.

 

This comes from a DigiTimes report about how chipmaker TSMC is benefiting from a ramp-up in orders from Nvidia and AMD, and apparently part of Nvidia’s demands pertain to a 5nm chip.

 

A quick review of "nanoscale" terminology in the microscopic universe of integrated circuitry may be in order...

 

A micrometer, or micron, is equal to one millionth of a meter.

 

A nanometer is equal to one billionth of a meter. 

 

For a vital visual reference when we're talking about the nanoscale world, here's a good graphic from Wikimedia Commons to illustrate just how small semiconductor engineering has been able to shrink itself to where transistors are 10nm (nanometers) or less, and smaller than the coronavirus...

 

 

I mentioned Thursday night that we should be buyers of AMD under $60 and we never got the chance today as it soared to nearly $70 because since it already uses TSM for 7nm fab it can fill the gaps in pc, laptop, and gaming hardware while Intel can't.

 

At least NVDA held it's own as the predominant data center geek with 7nm capability.

 

And here was the biting analysis from the chip guys at Raymond James last week after the Intel bomb...

 

 

INTC noted it is developing “contingency plans” to begin outsourcing given the internal roadmap slip -- and our view is that outsourcing has now become inevitable. By outsourcing leading edge technology, presumably to TSMC, INTC would give up what has been its main source of competitive advantage for 50 years and compete only on architecture, which we don’t think is enough to maintain the dominant market share and premium margins that are now expected.

 

In addition, the push out of 7nm (and the associated performance improvement) will provide further incentive for cloud customers to move to custom solutions and accelerated compute platforms from vendors such as NVDA, rather than to use products based on INTC’s inferior transistors.  Nonetheless, we view the roadmap missteps to be stunning failure for a company once known for flawless execution, and could well represent the end of INTC’s computing dominance. 

 

(end of RJ notes)

 

So where is MU in all of this? (and note that the stock symbol for Micron is similar to the international measure for a micrometer)

 

Well, it's complicated. This article will help explain the challenges for DRAM and NAND suppliers to go sub-10nm...

 

By Chris Mellor -April 13, 2020

 

Why is DRAM confined in a 10nm semiconductor process prison when microprocessors and the like are being built using 7nm processes, with 5nm on the horizon? If DRAM could be fabricated with a 7nm process, costs per GB would go down.

 

However, for the next few years 7nm DRAM is fantasy, due to capacitor and other electrical limitations at the sub-10nm level.

 

DRAM is more expensive and more tricky to manufacture than processor silicon, due to its critical nature. It has to hold data over many logic clock cycles. So it will lag in fabrication processes.

 

Process size shrinkage is the key to lower DRAM costs. Broadly speaking, the industry standard 300mm semiconductor wafer has a fixed cost. Therefore the more chips you can fit on the wafer the lower the cost per chip.

 

(end of excerpt from BlocksAndFiles.com)

 

This doesn't mean that Sanjay & Co. aren't trying. Though they say very little about it in presentations or on the website, they do have a partner they are excited about who is developing next-level architectures and if this company doesn't IPO soon, either Micron or NVIDIA should be buying them.

 

There have been a few updates from both Micron and Achronix in 2019, but this December 2018 article from Electronics Weekly really describes the playing field in detail...

 

Achronix is moving ahead at a terrific clip – its 7nm eFPGA core is out and its 7nm FPGA (field-programmable gate array) chip will be out in Q1'19.

 

It had revenues of $100 million last year and has 120 employees. It looks like the setting for a liquidity event.

 

CEO Robert Blake agrees that all the makings of a successful IPO are in place but would not be drawn on when, or even if. 

 

That might be because he’s found a winning formula and wants to play it. 

 

The formula can be encapsulated as “the only path left to improve energy-performance-cost is specialisation.”

 

And what can deliver specialisation better than a base technology which can be tailored to any application?

 

“It used to be easier to get gains,” says Blake, “now it takes more soul-searching to get improvements.”

 

Moore’s Law was the old easy route to gains, now gains come from architectures for efficient data acceleration.

 

More specifically, the best compute performance comes from architectures that target specific applications and data sets.

 

Achronix claims that its 4th generation eFPGA core, Speedcore 7t, put on the market today, delivers 300% better performance for AI and machine learning applications than the Gen 3 core.

 

Achronix claims the generational improvement is 60% faster performance, 50% lower power and 65% smaller die size.

 

The big architectural improvement is adding the Machine Learning Processor (MLP) optimised for efficient large-scale matrix-vector multiplication for AI/ML applications.

 

“It allows data to be moved in much bigger chunks,” simplifies Blake.

 

MLP blocks are flexible compute engines tightly coupled with embedded memories to give the highest performance/watt and lowest cost solution for AI/ML applications.

 

“Achronix was the first company to deliver production eFPGA IP to companies developing SoCs, enabling them to create programmable hardware data accelerators supporting new applications,” says Blake, “the new Speedcore Gen4 eFPGA architecture provides an optimal balance of hardware acceleration previously found only in ASIC implementations and adds the flexibility and reprogrammability of our production-proven FPGA technology to support increasing demand for new AI/ML and high data bandwidth applications.”

 

(end of Electronics Weekly interview excerpt with Achronix CEO)

 

And here was the latest update from the Achronix-Micron partnership.

 

The machine learning/deep learning weeds get thick and deep when you are listening to semiconductor engineers, eh?

 

We'll leave them for now and know that we own the right companies of the future space in NVDA and MU, and maybe a few of our smaller plays that could get accelerated by the exponential transformation of data centers on AI steroids -- like optical networking magician Ciena (CIEN Quick QuoteCIEN - Free Report) who is creating the bandwidth and super low-latency required for 5G data loads.

 

Have a great weekend and I will talk to you soon!

 

Cooker

 

(end of TAZR Friday excerpt)

 

Now, of course, I feel like I only own half of the right companies (TSM and AMD are missing) in the coming exponential acceleration of 5G + AI. 

 

But we know who to buy on the dips.

 

Disclosure: I own NVDA, MU, and CIEN for the Zacks TAZR Trader portfolio.