Wednesday, December 4, 2019

5 Thoughts on China Memory Progress, December Edition

Recent headlines highlight ChangXin Memory (CXMT) plans to ramp manufacturing and also referenced YMTC plans in NAND. Lets “fact check” the status and announcements.

Chinese and all memory companies make optimistic claims stating 10’s of Billions have been spent. They show cartoons of factories that don’t exist yet or are empty. These claims are easy to challenge as companies track shipments to China and China reports equipment delivered by area. So it is very easy to verify the actual amount of capex and model where it is going. Lots of money is going into Chinese owned memory companies but it is no where close to $30B yet and there are not significant shipments yet.

  • Report: CXMT is running 19nm DRAM in production at 20K wafer per month and will ramp to 60K by end of 2020.  Production was started in September
    •   Facts: We know based on shipping reports that CXMT is spending $200-250M per month on tools. We know that this is roughly on pace to tool out 1/3 of a large fab. CXMT will have tools to run 20K wafer per month soon and is on pace to ramp those tools to a 40K per month pace at least. The hype over the last 2 years is becoming a reality… it’s a real fab with real tools now!
  • Report: YMTC is running 10-20K wafer per month of NAND and will quadruple output in 2020.
    • Facts: YMTC/Tsinghua is among the leaders in hype in the past. They were supposed to have technologies shipping for the past two years and have shipped little so far.
    • Facts: YMTC is also taking tool shipments at a high pace. 2018 showed tool deliveries of ~$1B dropping dramatically at the end before resuming in June 2019 at a $200M/month pace. We will revisit the plans to quadruple output by end of 2020 but it is possible with current spending level if it continues through 2020
  • Report CXMT has production 19nm technology being used for LPDDR4.
    • Facts: CXMT has made a number of public statement so we don’t need to guess. The technology is based on Qimonda, they hired former Qimonda people as well as people with DRAM experience from Samsung, Micron, and Hynix presumably. So they have a technology and some expertise
    • Rumor (or Alternative Facts!): A recent report online stated that the yields on this technology are very low and this was confirmed with other sources. This is to be expected but they might have some work to do
  •  Report: YMTC Technology is 64L Xstacking and will ramp to 128L in 2020.
    • Facts: We know they are installing equipment and sources state that have low volume and they are sampling some 64L. The technology might not be competitive but if they get shipments out and supply local customers, then profits don’t matter at this time. 
  • SUMMARY PREDICTION (Facts, Alternative Facts, Rumors):
    • CXMT and YMTC have their initial fab line set up and are ramping with large tool deliveries. This is new! They plan to ramp to 60K+ wafers per month by end of 2020. Both are running technologies that are not mature and are probably 1 year plus from maturity. We also have no public samples of either technology (despite my attempts to buy them!)
    • CXMT and YMTC Combined will be less than 2% of the bits shipped end of 2020
    • CXMT and YMTC Combined will be 5% of the bits shipped end of 2025
Help needed: Samples from either technology! THANKS!

Mark Webb

Monday, December 2, 2019

Status of Emerging Memories and Cost Update

Status of Emerging Memory Technologies. No more guessing, Reality is here!
Mark Webb, MKW Ventures Consulting

The time for discussing possible or theoretical emerging memories is past. We have new technologies that are ready for implementation TODAY. We also have technologies being developed that could reach implementation soon. The key is to know the reality and trade-offs of available technologies today and to track the status and time to market of other technologies.

Cost for Memory Technologies in 2019 and 2020 (LOG Scale required!)

We have costs, revenue, cell sizes, performance metrics for ReRAM, PCM (3D Xpoint) and MRAM technologies today. Plus, we have a Product Lifecycle Model that shows where additional technologies are today, when they can be expected to sell in volume and what the milestones are along the way. Nanotube RAM, Ferro-electric RAM, ReRAM modifications and MRAM modifications are technologies that are being developed.

Some topics to discuss:
  • There is no universal memory in existence or even in development. There are always tradeoffs
  • Emerging memories being implemented today are not at the theoretical performance, cost, or density shown in marketing presentations. The reality is still useful, but we need to understand the actual performance to determine applications. We have details on this
  • Nothing is proposed to replace DRAM or NAND in the next 5-10 years so far. DRAM will not be replaced by MRAM and NAND will not be replaced by ReRAM. In 2020, More bits will be added to both NAND and DRAM supply than all emerging memories combined.
  • Embedded memory and stand alone memory have very different solutions and business models. Some technologies work well for one versus another.
  • The financial model of how to bring emerging memories to volume production has changed dramatically in the past two years. Foundries and Wafer fab equipment manufacturers have changed the environment so that we can achieve the scale needed.
  • Which technology from which company will win? We have evaluations of each on our Product Life Cycle and show how you can monitor them over time.

We will be at IEDM in San Francisco/Bay Area Dec 7-13.  Call or Email to set up meeting to discuss how these technologies will impact markets and how some other projections are based on myths, not reality.

Mark Webb

Thursday, October 31, 2019

Memory Market Results for Oct 31 (Trick or Treat) and What’s Normal

Memory Market Results for Oct 31 (Trick or Treat) and What’s Normal
We have the results in from a variety of manufacturers and can state what is going on in the market. We have all the details on individual companies and why they are behaving like they are

DRAM: All companies reported higher bit sales and much lower ASPs. Average ASP dropped mid teens  (remember when people thought prices would go up in Q3?). inventory is still pretty high

Reason: manufacturers are trying to get rid of bits, customers are taking advantage of low prices to build inventory. Datacenter/enterprise is back to growth and mobile looked pretty good. Prices dropped much more than costs so margins continue to shrink.

So DRAM is heading to 20% annual bit growth like predicted by everyone. But price drops will mean margins are continuing to drop. But at least DRAM is still making money despite lower margins.

NAND: Micron and Samsung reported higher bit sales on continued price drops. NAND is not a profitable business so prices must rise or flatten to actually make money. WDC reported flat pricing (they were claiming 10% higher prices in July) and Hynix reported mid single digit price increase on modest bits sales. And Inventory is still pretty high

Reason: Companies are behaving differently with respect to how they manage this and how aggressive to be given continued high inventory. You need to lower prices to move inventory. But you need to raise prices to be profitable. You need to move to 96L/12x layer to lower costs. But this increases capex and increase supply. This paradox will all sort out by 2H 2020… but it will be a competition until them.

NAND is a battle between working off inventory and trying to get profitable. Hynix and Samsung are behaving very differently on this. As are Micron and WDC. NAND will hit 40% bit growth as predicted by everyone but the margins are poor.

So what does balanced and normal look like? (Normal will happen in mid to late 2020)

DRAM price drops by 5% per quarter. Bits shipped increase by 5% per quarter. Costs drop by 5% per quarter. Inventory is flat (target is 4-6 weeks at supplier)
NAND price drops by 7% per quarter. Bits shipped increase by 8-10% per quarter. Costs drop by 7% per quarter. Inventory is flat (target is 6-8 weeks at supplier)

PREDICTION: We have data on who will win in this and who has challenges… And what recent Earnings reports mean…. And when inventory will be worked off…. And when Chinese companies will make a difference.  AND We have costs for everyone, so that will help determine winner.

Mark Webb

Wednesday, October 9, 2019

Preliminary Numbers on YMTC Xstacking 64L 256Gbit Chip

One of our top area of analysis is chip cost for all memory chips and even SOCs/logic chips. We are adding 64L YMTC chip to our report that has costs for all NAND memory from each company from 64L to 1xx layers.

High level Summary:

The chips size takes advantage of xstacking with periphery and array on top of each other. To a first order the chip is similar size to Microns 64L part. The reason YMTC chose chip bonding is to control logic processing and performance separate from array processing on separate wafers and then bonding them together. Having array built on top of the periphery does impact the periphery but Micron and others have been successful doing just that. YMTC claims to get much better NAND performance from the Xstacking process.

The down side is that you have two wafers, the cost to bond them together and the complexity of running a two process system. We will also see how Xstacking works in 8 die and 16 die packages needed for SSDs.

The cost, when mature, is modeled to be 34% higher than Microns 64L CMOS under array part. Since YMTC is not shipping volume yet and the Fab is projected to be less than 10K wafer per month currently, the actual cost is pretty far from maturity. We have the cost today, and how it will track over time as it ramps. We also have 128L projected costs for YMTC (and all other companies). We include die size, wafer costs, wafer yields, die yields, bonding cost and how these change over time.
We also have the business model on how this is still a win for Chinese industry, assuming they can ramp production. And how this impact NAND pricing and markets in the future.

The volume will be very small to start but, assuming the reported information is correct, we should see Xstacking chips in products in the next several months.

Mark Webb

Thursday, September 26, 2019

Top 3 Takeaways from Micron Earnings (and 2 bonus)

Top 3 Takeaways/Questions from Micron Earnings ... and two bonus takeaways from a Intel Presentation

  • Pricing dropped in Micron FQ4/CQ3: DRAM Prices down 20% after Trendforce and various sell side analysts said price was flat. Now there is some offset in timing but flat and down 20% are in way different directions. There is a reason and data to back what happened to Micron in FQ4 on DRAM pricing. They later stated that pricing was starting to climb on NAND although it dropped 8% in FQ4
  • Micron plans very little to no cost reduction in 2020. Repeatedly stated no NAND cost reduction due to conversion to replacement gate. DRAM cost reduction will be less than 10%. DRAM is somewhat expected and not an issue if pricing holds. But no cost reduction on NAND when it is losing money today has some serious implications for Micron in 2020.
  • FQ1 earnings will be less the FQ4. This was a surprise to just about everyone who thought we would get a 2H recovery. Why the drop in a historically good quarter? and this is including 3-5c/share earnings impact of NAND accounting change (BTW there is a good reason Micron is changing NAND depreciation)

2 Bonus items: Intel announced NAND and Optane Roadmap

  • Intel announced they will have new Optane DIMM Modules "Barlow Pass" in 2020 for servers. This seems pretty soon for a follow on given the delays and slow adoption of Apache pass DIMMS. However it is important to keep new generations coming and Intel sometimes delays new products so maybe it wont step on Apache Pass toes too soon. We have cost models for 1st and second gen Optane memory media. Rumor is that Barlow Pass uses Apache pass memory chip (unverified)
  • Intel announced 144 layer NAND technology with floating gate. Intel and Micron are both going to struggle with their first generation apart..... mainly from a financial point of view. but the fact that Intel is announcing the 1xx layer part for 2020 is good for Intel customers and a welcome surprise for me. I was getting worried Intel would skip the first gen 1xx NAND technology. NAND costs will keep coming down in 2020/2021 ... we have summaries of where each company stands in cost with 96L and 1xx layer.

Lots of new technologies, financial challenges, and we have the macroeconomic challenges too. 2020 will not be a dull year.

Mark Webb

Monday, August 19, 2019

Memory Pricing Update and Models August 2019

One of the most common questions these days is “what will happen on pricing”. No one knows for sure and wild reaction to news articles is adding to the speculation

Some thoughts on pricing models

  • DRAM exchange/ Inspectrum is usually directionally correct, but wrong in magnitude and in absolute numbers when comparing to actual company ASP
    • Spot pricing is open market and not related to what major customers pay
    • Customers react differently in magnitude to increases and decreases in exchange price
    • Even contract price is not REAL contract price and does not match company ASPs
  • Over time prices drop … A lot … in a balanced market, we model 25% per year for NAND and 15% per year for DRAM going forward. This is FAR lower than historical memory pricing (35%+) and matches cost expectations.
  • Actual Pricing is a response to market growth, Inventory, and customer relationships. We can discuss how to track these
  • the number one input to memory company profits is pricing.... so anything they can do (legally hopefully) to get pricing to increase will help memory companies. Unfortunately or fortunately, they are just not very good a collusion! but there are legal tactics that they do all the time.
  • Cost for each company matters. One model is that pricing always drops til at least one company loses money. Today, not all memory companies have the same cost. in 2021, there will be even larger differences in cost with different leaders
Where are we today?:

  • We had a shortage of NAND and DRAM through about Q4 2017 and Q3 2018 respectively. Since then both are in major oversupply. Inventory at customers and suppliers skyrocketed. We can explain why this happened and when it will happen again.
  •  Well publicized supply limitations (power outage, Japanese chemicals, Reduced wafer starts) have people optimistic on pricing. All of these require discussion to see if they matter and how much. Remember: lowering wafers starts is a horrible business idea.... unless you tell everyone you are doing it!
  •  Dramexchange has reported spot NAND price increases, contract price decreases, wafer price increases.  This actually all makes sense. We have input from suppliers and customers and where they see the market and strategies. It is a cyclical market, prices will go up WHEN there is a shortage. We have a model on when this will happen across the board as reported by real companies. We also tell you how to monitor it

So lots of people will report where pricing is today (Dramexchange and Inspectrum are good starts). The key is how to monitor this in the next 3 months, 6 Months, 1 year and 2 years and why each company may not see the same effects. Contact us for more information

Mark Webb
MKW Ventures Consulting

Thursday, August 8, 2019

3D Xpoint Revenue and Bit Output

We presented bit output and revenue projections for 3D Xpoint at Flash Memory Summit this week. See details below

To calculate this, we modeled Fab wafer capacity, factory loading percentage, yields, die size, cost, pricing, etc and how those are projected to change over the next few years.

Charts are below, full presentation is on our website and Flash Memory Summit site. The full details on all of the parameters going into the model are available to clients

GB Output Projections:

3D Xpoint Revenue Projections:

See presentations for more info. Call to set up meeting to discuss monthly updates to 3D XPoint output, revenue, and status on 2nd Generation 3D Xpoint.

Mark Webb

See ALL Reports at