Follow me on Twitter @microcapreturns
Atomera is an Intellectual Property licensing company that plans to generate royalties from three unique patent portfolio’s. Each portfolio can be licensed by itself or sold to the highest bidder. With fixed costs around $15m/year 95% of future royalties will fall to the bottom line and with only 23m shares outstanding the earnings per share can be substantial. Atomera is engaged in the business of developing, commercializing and licensing proprietary materials, processes and technologies for the $450+ billion semiconductor industry that is growing to $750b by 2027. By incorporating MST, transistors can be smaller, with increased speed, reliability and energy efficiency. In legacy nodes, by adopting MST, performance can be increased and die shrunk so capacity can increase over 30% and help solve the current industry shortage issues along with reducing power and water usage that improves the industries carbon footprint. Recent data demonstrated the applicability of MST to leading edge 5/3/2nm fabs meaning MST can be adopted across the entire $750b TAM. Atomera believes that MST can be widely incorporated into the most common types of semiconductor products, including analog, logic, optical and memory integrated circuits.
There are three unique patent portfolios that can generate royalties or be sold off. The first portfolio is the one with the the existing JDA in Phase 4 and also covers the additional 10 customers in Phase 3. The goal is for these Phase 3 customers to move into Phase 4/5 and then into production. The second portfolio which will start to come into play with Phase 4 customers will be licensing of specific part designs for production. As an example, the first JDA customer, which we know will be multiple nodes/customer product lines could well also license the second portfolio for unique designs. The third portfolio, which became clearer on its potential with the recent whitepaper, is for next generation fabs in the 5nm and smaller variety. While it is too early to put final valuations on portfolios 2 and 3 it is safe to say that portfolio 2 will likely follow the piece part(Integrated Device Manufacturer – IDM) royalty path with each part earning a royalty and portfolio 3 will be utilized in next generation fabs that cost in excess of $20b each. A $20b fab that may require MST to achieve diffusion profiles on the highest performing processors and memories that have ever been produced should yield very high royalties.
- 3 unique patent portfolios to license
- Core MST Method and Device (Wafer and IDM)
- Core MST Technology for $450b space
- 500 nodes potential $64m-$255m/year in royalties per node
- MST Enabled Devices/Architecture
- Near term parts using MST (DRAM, CMOS Image Sensors, Varactors, Optical Waveguides etc..)
- Next-Gen Architectures using MST
- Parts in 2nm, 3nm and 5nm fabs
- Nanowire, Nanosheet FETs and GAA
- Happy Earth Day!
- By Robert Mears, Atomera Founder and CTO
Portfolio #1 Core MST Method and Device consists of two unique licensing opportunities for MST. The first is the wafer manufacturers who will supply MST wafers for companies to build parts that can take advantage of all of its unique properties. The second is to semiconductor companies or IDMs who make unique piece parts and in this case it will be a piece part royalty stream. Both will be estimated for royalties per node. In the case of the JDA, that is in Phase 4, that will be IDM or piece part royalties as they have already said it will be used by multiple business units and product lines once approved for usage.
- Assumptions for portfoilio #1 royalties
- First adopters will be with largest semiconductor companies
- There are 370 fabs and around 500 potential nodes
- Royalties will be 2% which is right in the middle of the 1-3% that company has stated as typical
Foundry Wafer royalties. Large wafer manufacturers, like TSMC/Global Foundries/SMIC, and others shown below, will pay a 2% per wafer royalty to manufacture wafers with MST technology. These wafers will then be used to manufacture parts that can take advantage of the low resistance super high performance wafers. As soon as one of the large wafer manufacturers start selling these unique wafers the rest must follow quickly in order to stay competitive. Also fabless companies like Qualcomm and Apple will demand multiple sources for these wafers. As shown below each of the worlds largest wafer fabs will generate $58m/node in royalties for Atomera once running at 100%.
Individual piece part (IDM) royalties. These will be for companies like Samsung, SK Hynix, STMicro and AsahiKasei(AKM) who are designing parts that will take advantage of the MST properties for lower power, size, voltages or die shrinks that can boost capacity. As can be seen in the slide below for a large customer, like our first JDA, the estimate for royalties is .09/part. A high volume part could generate $255m/node in royalties per year for Atomera. The first JDA is expected to be multiple nodes.
Portfolio #2. MST Enabled Devices. The company has patented many specific parts in the areas of Image Sensors, memories and varactors. Companies have the option of licensing specific designs. I would expect these royalties to follow the IDM model as discussed above.
Portfolio #3. Next Generation Architectures. The recent MST blog and whitepaper discuss how MST will be beneficial for next generation fabs. Specifically how it can do a much better job blocking dopants than the industry standards. Carbon has been the industry standard for over 20 years and has severe limitations that the industry has had to live with. Now with MST they can get 3x the performance without the limitations. As the whitepaper discusses and as seen below there is no reason it won’t be adopted in next generation fabs making it a must have for the industry. While hard to put an exact number on if even a small 5 cent royalty per part on all future fabs outputs will be a huge number every year.
Future Valuation of 3 Unique Patent Portfolios
Trying to figure out the future valuation for Atomera’s IP all 3 portfolios must be taken into account. Portfolio 1 covers the current 370 fabs that cover over 500 nodes. Calculations above have each high volume node in full production valued at between $58m-$255m/year in royalties. Since they are working with half the largest semiconductor companies and have 11 customers in either Phase 3/4 lets say 2 nodes per customer. So 22 nodes and using an average per node of $156m/node brings a yearly estimate of $3.4b/year in royalties. While I think we will start to see royalties in 2022 full running royalties will not kick in until 2023. The 22 nodes estimate is roughly a 5% fab penetration which is on the very conservative side. By 2025 I expect roughly a 50% fab penetration but over time the fab size will come down as smaller fabs come online.
Portfolio 2 which covers unique piece part royalties will much more likely be by a family by family basis. Since these royalties will be part by part and much more likely to be lower volume I will use the wafer fab royalty of $58m/node for each part. By 2023 I could see 5-10 parts fitting into this category or roughly $250m/year in royalties. There is also a strong possibility that certain parts might be sold outright so companies don’t have to share designs.
Portfolio 3 has come much more into focus with the recent whitepaper and data showing how all future fabs will likely adopt MST from the start. First fabs won’t be online until 22/23 at the earliest. While these fabs will produce the bleeding edge parts that typically command the highest royalties I am going to estimate that royalties will only be 5 cent per part. The reason for this is I think this becomes the industry standard, like carbon has been for the last 20 years, going forward. I will use a conservative $1b/year in royalties starting in 2023 but really not seeing mainstream volume until 2025 but going for years to come. Industry standards have to have compelling reasons to be changed as can be seen with the MST data. Of course there are likely to be many other uses for MST to be found as the company continues to find applications.
Looking at the three portfolios in full running production in 2023 the company could see over $3b/year in royalties. By 2025 that number could be well over $5b/year. With current cash position of over $30m there is no need to dilute the stock which leaves share count around 23m. With limited overhead of $15m/year almost 100% of the royalties will fall to the bottom line making Atomera the potential CASH COW for the next decade. Unfortunately these numbers are well known since most come out of their presentations. What many people don’t know is every large company has a Mergers and Acquistion (M+A) department within their company. Their job is to find potential targets and determine what they are worth and decide if they should acquire the other company. In Atomera’s case every company they are currently working with has done exactly that. But Atomera management has been working this for a long time and they also know what the numbers are. Until the first Phase 5 is announced I can’t see any offers being made but once the first one hits I think you have 19 customers who all have already done their due diligence. Below I list several companies who could eventually acquire Atomera. Companies like Synopsys and any of the equipment suppliers are high on my list especially with their stock prices so high that they can use shares to purchase with minimal dilution.
Who is most likely to buy Atomera
Many people who look at Atomera ask the question if the technology is so great a large semiconductor company would have bought them years ago. This is the least likely scenario. Just like we are seeing with SoftBank trying to sell ARM to Nvidia none of Nvidia’s competitors who use ARM are excited about licensing from a direct competitor. This is exactly the same problem a large fab would have if they owned the MST technology and another large fab wanted to license. Can you imagine Samsung doing installation on a TSMC node? Not going to happen. But companies like Apple insist that multiple companies be able to produce their parts so they force companies like Qualcomm to develop multiple companies to produce their parts. So while I expect a large numbers of fabs to adopt MST over the next several years the installations will need to be done by companies that are not direct competitors to the fabs themselves in terms of making semiconductors. The list below is likely a good starting point of who would be interested in acquiring the company and the first Phase 5 announcement will certainly bring interest once it becomes obvious that industry adoption is starting. The first Phase 5 qualification erases any doubts about the viability of the MST technology.
Who would be interested in owning Atomera and why
Why would companies like Tesla and Apple be interested in an IP company? It wouldn’t primarily be because of the royalty stream but instead be leverage over the large fabs to make sure they get their parts first. The industry is seeing a shortage in fab capacity and major automobile and electronics manufacturers are seeing shortages in electronic part deliveries. Since this is probably not going to change in the future having this leverage over the fabs would insure their parts are delivered on time.
Synopsys, Inc. provides electronic design automation software products used to design and test integrated circuits. With the release of MST Cad, which is a speciality program to allow designers to design parts taking advantage of the unique properties of MST. Given they are one of the industries top simulation companies they have the unique ability to understand the value of the MST technology. They are in perfect position to purchase the IP and license to other semiconductor companies given their long relationships with everyone in the industry.
Applied Materials, Inc. provides manufacturing equipment, services, and software to the semiconductor, display, and related industries. They have been a long time partner and are critical for modifying the semiconductor equipment that is used to make wafers to implement MST. They are in the unique position, working with all the fabs, to take advantage of selling equipment specifically to take advantage of MST. This would give them a proprietary advantage in the space. Any of the semiconductor equipment partners they are working with would also be potential acquirers.
SoftBank Group Corp. is a Japanese multinational conglomerate holding company headquartered in Minato, Tokyo. SoftBank owns stakes in many technology, energy, and financial companies. The are best known in the IP licensing space having purchased ARM holding, a processor company, for $32b in 2016 and is now selling it for $40b to Nvidia. ARM only creates and licenses its technology as intellectual property (IP), rather than manufacturing and selling its own physical CPUs, GPUs, SoCs or microcontrollers. This is the perfect company to buy or invest in Atomera. While they could just purchase they may decide to take an equity stake in the company and then help speed up adoption by providing resources and contacts within the space. I would guess they could speed adoption to 50% penetration by several years.
Large fabless companies Qualcomm and Broadcom will also be possible acquirers. While not considered in the original analysis since I did not think an individual semiconductor company can own this technology since it needs to be adopted across the industry it is possible that a fabless semiconductor company could. Since these fabless companies would not be competing with other fabs they could license the MST technology without directly competing with the fabs. In fact the fabless semiconductor companies are in the unique position of needing multiple fabs to adopt as soon as possible and this would help get fab priority for parts and also control their costs by earning royalties.
- Whitepaper summary
- MST® Dopant Blocking for Advanced 3D Applications
- Whitepaper by Robert Mears, released 6/29/21
- The paper puts the new MST data in context for advanced 3D architectures for the 2/3nm node, such as nanosheets or gate-all-around (GAA) devices.
The reduced diffusion FinFETs analyzed above also exhibit reduced RDF and overall reduced Vt variability, as seen in the table in Fig. 6. Reducing the overall variability allows a tighter specification over the manufacturing range. Since leakage depends exponentially on Vt, the net result is an ability to lower the median Vt, allowing yet higher median current. Conversely, the same current can be achieved for a lower supply voltage. Since power dissipation depends on the square of the supply voltage, this reduction in variability can be used to reduce power.
Boron and phosphorus are well known to diffuse readily in silicon, particularly when they combine with silicon
self-interstitials to form dopant-interstitial pairs. The MST oxygen layers are doubly effective at reducing diffusion,
because they trap and localize the interstitials as well as trapping boron and phosphorus
The main technique currently used by the semiconductor industry to reduce dopant diffusion, is to co-dope silicon
with carbon. Carbon is a substitutional dopant (it sits in the same lattice location compared to where a silicon atom
would normally sit). While the industry has tolerated carbon doping, it brings other issues of changing the silicon
stress, and is recognized as a contaminant which can increase scattering and reduce current.
Carbon doping has been known for more than twenty years, and has been well characterized. The classic
experiment/diagnostic for the effectiveness of carbon doping on diffusion blocking is a boron marker experiment.
Boron is introduced during silicon, silicon-germanium or silicongermanium- carbon epitaxial growth, typically as
two thin regions: one without and one with carbon doping. Once grown the epitaxial stack is then subject to various
anneals, and measured by SIMS
Comparing the MST boron marker with published data for SiGe:C (carbon doping) it can be seen, comparing Fig.
9 and 10, that MST is significantly more effective for the same thermal anneal. While the silicon control is similar in
both experiments, for the SiGe:C case, the peak of the lower boron marker is reduced more than 3x, and the boron
marker broadens by a factor of two, compared to almost no discernable change for MST in Fig. 9. Because of the
quantitative nature of these experiments, the diffusion coefficient under carbon doping and MST can be directly
compared. The result is that MST is more than twice as effective as carbon doping. Furthermore, as we shall see,
the diffusion blocking can be achieved in a very thin epitaxially grown MST region, considerably broadening the
application scope of the MST approach.
Our goal for 2021 was to achieve >90% blocking
for a grown MST region of less than 5nm. The Atomera epi team has risen to the challenge and excelled. The latest
characterization data has shown 99% blocking (compared to silicon controls) of both boron and phosphorus for MST
films thinner than 5nm, under 1000C 20s anneals, and 98% for 1050C 20s anneals. Atomera believes these data
significantly increase the epi applications for MST in advanced devices.
Any due diligence from this site is for entertainment only and not a solicitation to buy or sell Atomera stock. Any estimates are just examples of what is possible and should not be considered financial advise. I have not been compensated in any way and will never be compensated for my reports.