Weekend Update #130
Welcome to Blue Room's Weekend Update. Each week, we're sharing what companies we're researching and the what, the who and the how that we think makes the companies interesting and unique. This roundup is brought to you weekly by a group of interns, creative minds, artists and investors who believe that through best in class investing along with the democratization of financial education we can do great things together. Enjoy, Explore and Share.
Markets managed to eke out small gains for the week, mostly led by tech as Tesla closed out its 11th-consecutive daily gain to reach an 8-month high. Oil stocks also rose after Saudi Arabia made a surprise announcement to cut 1 million-barrels in July following the recent OPEC+ meeting. After gaining over 20% from its October lows, the S&P 500 is now in a technical bull run.
Despite resilience in equities, economic data remains cautious as US mortgage applications fell for a fourth week, with 30-year fixed rates held near an almost seven-month high. Additionally, the MBA index dropped 1.4% last week to 151.7, the second-lowest level since 1995.
The week was also filled with litigation as the SEC brought enforcement actions against both Coinbase and Binance for operating an unregistered securities exchange, among other charges, and Former President Trump got indicted on seven federal crime counts related to classified documents he kept after his time in office.
Weekly Performance
S&P 500 4,298.86 +0.39%
Dow Jones 33,876.78 +0.34%
Nasdaq 13,259.14 +0.14%
Key Economic Readings Next Week
Tuesday, June 13 — NFIB Small Business Optimism, CPI, Real Average Weekly Earnings
Wednesday, June 14 — PPI, FOMC Rate Decision
Thursday, June 15 — Retail Sales
Friday, June 16 — Univ. of Michigan Consumer Sentiment
Thank you Blue Room Analyst NICK PEART
S&P Global US Services PMI™
According to the latest data from S&P Global’s US Services PMI release, output rose at the fastest rate in just over a year, supported by a strong expansion in new business. The upturn in new orders was driven by improved demand conditions in both domestic and export markets. At the same time, firms stepped up their hiring activity again, with employment increasing at a solid pace. Sufficient capacity to process incoming new business allowed companies to reduce backlogs of work for the first time in four months. Meanwhile, business confidence picked up to the highest level in a year, albeit still weaker than the series average. On the price front, rates of both input price and output charge inflation softened since April though remained elevated.
The seasonally-adjusted final S&P Global US Services PMI Business Activity Index registered 54.9 in May, up from 53.6 in April and broadly in line with the earlier released ‘flash’ estimate of 55.1. The latest upturn in business activity was the fourth successive monthly increase, with the pace of expansion accelerating to the steepest since April 2022. Greater output reportedly stemmed from stronger demand conditions driving a sharper rise in new orders.
Growth in new business quickened again, with the rate of increase the steepest in just over a year. Companies noted that greater client confidence supported the expansion in new orders as customers—especially in consumer markets— were more inclined to spend. Some also highlighted a broader client base and the acquisition of new customers.
Takeaways:
The AI market TAM is large and expanding, despite the commentary made by accelerator competitors in reference to CPU/GPU market share shifts. Intel is looking to segment its core competencies and maximize efficiencies. Skipping out on hot lots and providing better transparency around internal manufacturing capacity vs external, are two ways in which Intel believes it can increase efficiency.
Like most other competitors in the AI processor and accelerator markets right now, Intel posits that software development is the key to increasing hardware performance. This effort will be a function of working with customers to design and deploy their workloads to best run on Intel hardware. Intel also prides itself on integrating AI capabilities (at various levels) into all of its computing platforms, from cloud, to enterprise, to edge, and then into client. After several delays, the fourth generation Xeon CPUs are shipping with the target unit shipment of 1 million units by the end of 1H23 for enterprise and cloud scale clients. On the client side, Intel has the integrated VPU to provide AI capabilities to PC computing devices ramping later this year. On the accelerator side, Intel has Gaudi and GPU Max for LLM capabilities (training and inference). Lastly, for enterprise level distributed inference and retraining, Intel offers the GPU Flex.
So overall, Intel actually has a broad product portfolio to serve these new and developing end-markets, the currency headwind is that the product development and pipeline is not causing the same types of revenue inflections that we saw with competitors, causing investors to question the true performance of some of Intel’s products, especially on accelerators for training and inference.
Rich Repetto — Piper Sandler
Okay, everyone. Welcome back to Piper Sandler's Global Exchange and FinTech Conference.
It's my pleasure to introduce Brian Armstrong of Coinbase, who has not shied away from going to the public and expressing his opinion. Too bad you weren't here the last session, Doug Cifu of Virtu is pretty strong in his views.
So first, I'll just turn it over to you, Brian, if you have any comments in general you want to make about Coinbase and sort of what has transpired over the last couple of days, I guess.
Brian Armstrong — Co-Founder and Chief Executive Officer
Yes. Sure. So obviously, we got this complaint from the SEC yesterday, and it was not unexpected. We've gotten the Wells Notice before that. And actually, for the last year or so, we've been in dialogue with the SEC about our business, and they asked us to come in and answer a bunch of questions. We're happy to do that with regulators we work with all over the world. And so we met with them about 30 times over the last year. Every time they asked us a number of questions, we gave them answers. We told them, do you have any feedback for us? When you say come in and register, what does that mean? What license can we go get? We even went out and acquired a broker-dealer license hoping to activate it with the SEC, so we could trade crypto securities. We haven't found a path to do that with them yet. We formally filed a petition with the SEC to get clarity around rule making. They have yet to respond to that.
BLUE ROOM Analysis
In Schrödinger’s first investor conference following the Q1 2023 earnings call, Schrödinger previewed a few exciting catalysts for the future of the company and had a detailed discussion around certain proprietary programs at the 2023 Jefferies Global Biopharma Conference.
Beginning the Q&A session with the potential for AI advancements to further propel Schrödinger’s platform to new innovations, investors were clearly excited to learn about the readthrough from recent excitement in the AI space, fueled by ChatGPT and large language model advancements. President of R&D and Therapeutics Karen Akinsanya and CEO Ramy Farid both pointed out that such AI/machine learning models could be applied specifically to the biology side of drug discovery and noted that it is an area Schrödinger is very interested in exploring right now.
For years, management has made the distinction that, to synthesize molecules with better properties, Schrödinger’s physics-based methods are key primary inputs that any company would need to use before running any machine learning techniques — as the library that machine learning models are trained on is built by Schrödinger’s physics platform. However, management is clearly exploring new ways for AI technology to be deployed in the area of identifying biological targets, in which they can subsequently use the physics-based platform to identify superior molecules. So far, no specific investments or innovations on the biology side have been disclosed, but Schrödinger maintains that they are committed to investing in the platform and continuing to innovate in new areas of drug discovery and molecular design.
Investors will be looking forward to Schrödinger’s R&D Day in September, where CEO Ramy Farid previewed they will be disclosing some of the currently undisclosed proprietary programs. On Friday, June 9th, Janssen will also be providing a data update on their MALT1 candidate, which could provide some readthrough into potential efficacy for Schrödinger’s SGR-1505 program.
Piano Recital, Age 10
Chances are, you or someone you know was forced to learn how to play piano, or another instrument at a young age (yes, the recorder counts too). For me, it was something I dreaded. Every day I knew I would be scolded for not practicing my songs, and my mother would have to repeat the same lecture. At first, I would make snow angels on the carpet, wailing on the floor so I didn’t have to practice. Then, I would cry, scream, every course of action to convince my mother to quit. But everyday it was the same. Tears raced down my cheeks, causing a muffled *plunk* as it hit the ivory keys. When one thinks about music, expression, creativity, art probably comes to mind. Except, in those moments, I was just a slave to the page in front of me. Reigned by a teacher and the invention of a now dead genius. How was this something beautiful? What betterment did apply to me? I squeezed my eyes shut and began to play.
Ten years later
Last quarter, Ambarella posted revenue of $83.3 million, meeting their guidance and beating BLUE ROOM expectations. Beginning in the back half of last fiscal year, Ambarella began seeing sell-in challenges as customers adjusted their end-demand forecasts. For Ambarella, this was most notable in the IoT end market. The stock fell following guidance, which came in below expectations on slower rates of customer inventory digestion.
Like many other companies along the semiconductor supply chain, there’s an anticipation for a rebound in demand after the industry-wide inventory correction subsides, however, the market expectation for that has been pushed back. We will be listening on the call to see whether, and by how much, Ambarella’s previous forecast gets pushed back.
With the inventory correction, Ambarella is seeing the most weakness in customer end demand within the IoT markets relative to the automotive markets. Although some customers may still be seeing their bookings expand, the actual sellthrough of Ambarella products is dependent on the timing of shipments. The company continues to indicate 1Q as the bottom for revenue this year. Ambarella indicated that there is a slight chance that IoT revenue can rebound before the end of this fiscal year, but they “are not calling for it”.
On the automotive side, some customers are pushing out orders, but the size and scale of these headwinds is much smaller than it is for IoT. According to Fermi, 60% of the AI (Computer Vision) revenue is coming from CV2, and a majority of the investment is going into CV3 development and demoing. The CV72 chip is also making solid headway in China in L2, where the design win cycle is much more rapid than it is globally by up to 30 months.
Regarding competition in the automotive business, Ambarella believes that the scale issue is essentially a non-factor at this point as they have proven their ability to get Tier 1 design wins. Relative to the main competitors Ambarella had this to say:
For Nvidia the GPU engine is too power intensive, and will likely only serve a purpose in high-end vehicles, where customers are more willing to tolerate the system costs.
For Qualcomm, Ambarella believes that they have better video quality, processing, compression and power efficiency. They cite their 10 year competitive history with Qualcomm in IoT markets as proof of this.
For Mobileye, Ambarella believes that their “Black Box solution”, or inability to open their software solutions to be customizable for OEMs, is the biggest disadvantage. So, Ambarella should secure wins with OEMs that want a lower-end, ADAS focused, open architecture.
When it comes to leveraging their Tier 1 manufacturing partners, there are two different go-to-market strategies for Ambarella:
The first pertains to L2+ OEMs who want to make the hardware decision based on specs and manufacturing capacity. So in this market subsegment, Ambarella has to do direct marketing with those OEMs to convince them of the value proposition.
The second strategy regards the ADAS subsegment, where the Tier 1 has more control over the choice. In ADAS the Tier 1 can prescribe any solution they believe is the best fit for an OEM. Ambarella relies on its product portfolio to do the marketing work on its own.
DocuSign, the provider of electronic signature solutions, reported earnings Thursday after market close. The company has been undergoing a significant amount of change—the former CEO Dan Springer agreed to step aside last summer, and outgoing CFO Cynthia Gaylor announced this past March that she would be leaving the company following the Q1 earnings release, with incoming CFO Blake Grayson assuming the role beginning next week. The company has implemented a reduction of about 10% in its salesforce which primarily impacts the sales and marketing team and the resulting savings will be redirected towards product development.
In Q1, the company reported $661 million in revenue, a 12% year-over-year increase that beat Street estimates of $642 million but also represents a deceleration in growth (the same period last year saw a 26% increase). Adjusted diluted EPS came in at 72 cents per share, exceeding Street estimates of 55 cents per share and nearly doubling the prior year period’s 38 cents per share. The quarter’s positive performance was aided by an expansion of existing customers and the addition of new customers as well as an increase in sales to their commercial and enterprise customers through their direct and indirect go-to-market initiatives. The company also raised its revenue and billings guidance, which investors had a very positive reaction to immediately following the earnings announcement. However, as the earnings call progressed and market participants digested the news, they realized there was more to the story.
My parents always tell me that I was born with slow-twitch muscles, which is why distance running was a sport that made sense for me. I was more willing than the average person to endure long runs, and I enjoyed being around my closest friends throughout practice. Beyond that, I was never really forced to consider the purpose of my running. Life was simple – meet at the track for 3:15 practice, follow the coaches instructions, show up for meets and try to win. The only context I knew running in was this organized team operation, so when I graduated high school, I thought my running career was effectively over.
I spent my freshman year struggling to find purpose in running, and would fill my weeks with a few leisurely jogs, just to stay in decent shape. At many points during the year, running became more of an obligatory health assignment, and I found myself questioning why I even did it in the first place. My second nature from high school of training 6 days a week was wearing off, and I feared I was going to fall into the pithole of inactive Americans who sit at a desk from dawn till dusk and have lost track of their active youthful days.
On Wednesday, June 7th, 4D Molecular Therapeutics released interim clinical data from the Phase 1/2 AEROW study of 4D-710 in cystic fibrosis and hosted a clinical update call for investors. The data update centered around the 3 patients in Cohort 1, the lower 4D-710 dose at 1E15 vg, and an additional update on the higher-dose Cohort 2 at 2E15 vg has been announced for December.
Encouragingly, even at the lower dose, 4D Molecular Therapeutics achieved a clean safety profile while providing initial signals of efficacy and statistically significant improvements in quality of life measures in CFTR modulator-ineligible or -intolerant patients. The company highlighted that, in all three patients, the gene therapy shipped on 4D’s novel and proprietary vector, A101, achieved over 90% of CFTR protein expression of airway cells. This is the first data of any program to date to show such a benefit — albeit an initial look at a small sample of patients.
Cystic fibrosis could be a large market opportunity for 4D Molecular Therapeutics if 4D-710 can continue to demonstrate a clean safety profile and efficacy in a larger Phase 2 trial. 40,000 people in the U.S. alone suffer from cystic fibrosis, and worldwide, the genetic disease affects over 100,000 people. Currently, Vertex’s Trikafta and CF franchise dominate the market, and CFTR modulators generated ~$9 billion in sales in 2022.
4D Molecular Therapeutics believes that 4D-710 has the potential to show efficacy in both the modulator-ineligible (15%) as well as modulator-eligible CF populations. Focusing first on the high-risk, high-unmet-need group of modulator-ineligible patients, who lack treatment options, 4D has a roadmap to expanding into the more general population if the program continues to produce positive data. The November data update will be a determining factor in which dose level 4D will pursue in Phase 2 trials based on safety and preliminary efficacy signals, although the low-dose efficacy signals in Cohort 1 combined with a good safety profile seem promising enough to progress if the higher-dose Cohort 2 shows safety concerns.
On May 9th, 4D also disclosed the closing of an upsized public offering which raised $138 million in cash, bringing the company’s total cash to $331 million, and extended their cash runway into H1 2026. With clinical data validating 4D’s directed evolution platform as well as additional capital, FDMT stock opened Thursday, June 8th up 15% following the update.
Takeaways:
Nvidia’s been on a 20 year journey through AI beginning with making their GPUs programmable in 2006 and investing further into the architectures and ecosystems around their products. As the AI market progressed, from neural networks in 2012 to generative AI today, Nvidia has kept pace with the innovation to have offerings in the space. For context the evolution of modern machine learning (ML) is as follows:
Tangible neural networks are expressed in proof-of-concept around 2012 and were used for image recognition.
Neural networks evolve to recognize tokens and produce responses to text, speech, video, etc. Those are now called transformer models. The original BERT model was the first.
Generative AI not only accepts inputs and transforms them into basic outputs, but can broadly understand context.
The next step for AI is likely to be reasoning, according to Nvidia.
Regarding scale, the “large” in LLMs is likely to remain, due to these Generative AI models requiring as many parameters as possible to asymptotically approach perfect inference. Some people in the industry are wondering how long the trend of parameter complexity doubling every half-year will continue. For this reason, Generative AI models will become commoditized, as using someone else's model can help increase the time it takes to build your own. People are going to want to monetize that, and we’re seeing that with AI startups like OpenAI and AdeptAI. For actually running these models, Nvidia works with the developers to improve or modify the GPU architecture, drivers, parameter size, bits, software, etc. in order to increase the performance of the GPUs. Nvidia stated that its A100 product got 2.5x faster over its lifecycle simply due to these constant upgrades to the ecosystem efficiency.
Takeaways:
Although Nvidia’s growth has been a data center Cinderella story as of late, their strides in the Gaming market should not be overlooked. Nvidia has invested in AI for gaming to a similar degree as they have on the data center segment. Nvidia’s initial AI investment for gaming was DLSS, which uses generative AI to create 7 “theoretical pixels” for every 1 “actual” pixel rendered. This technology is meant to substitute any delayed or missing pixels while gaming with what the AI thinks should be there, such that the gamer doesn’t notice that a pixel was not rendered in real time. The second introduction they are making on the development side is ACE for games, the combination of Nvidia Reva, for speech recognition and processing, Nvidia NeMo, the larger language generative AI chat model, and Nvidia Audio-to-Face, which is a component of Omniverse that turns language inputs into animated character outputs. Those three AI efforts of Nvidia all come together under ACE for Games, and this allows game developers to implement generative AI into their gaming processes.
Really interesting demonstration of this (Link to Video Demo)
Nvidia has added over 100 million clients to their GPU installed base, bringing the total somewhere above 200 million. The growth is driven by new gamers, as well as existing console gamers making the shift to PC. Nvidia estimates that the growth in GeForce gaming interest matches the growth in revenue, or about a 20% CAGR from 2019 through 2022.
Nvidia has also benefited from the demand for GeForce GPUs by increasing the average selling price per GPU.
Company speculated that the average “higher end” customer spent between $400 to $600 above what they previously paid for newer generation 40 series GPUs, while the average “mid-to-low end” customer spent between $300 to $400.
Nvidia is expecting Diablo IV, Starfield and Call of Duty to be the next drivers of the GPU upgrade cycle.
Chief Medical Officer Alan List and Co-Founder & Chief Research Officer Jeff Smith from Precision BioSciences go in-depth on azer-cel, 19B Stealth Cell, and DMD programs at the Jefferies Healthcare Conference.
Key points of emphasis for azer-cel are that the single gene edit and controlling the CD4:CD8 ratio in the updated manufacturing process give azer-cel strong scientific backing for the positive CAR T data update in May 2023. Alan List described that a Phase 2 study they believe will be registrational and could include around 75 CAR T relapsed patients, with no more than 5 total prior lines of therapy and no more than 2 lines of therapy post-CAR T relapse. Commentary on market size and high unmet need should also be supportive of favorable pricing for an azer-cel commercialized product.
The Duchenne muscular dystrophy program also had positive discussion around how the edit achieves an end genetic result of a dystrophin gene that resembles wild-type even more than the Becker’s mutation. In addition, due to ARCUS’ attributes, the gene editing therapy could potentially persist throughout a patient’s entire lifetime and continually provide benefit. Compared to recent competitor data and micro-dystrophin therapies, Precision’s preclinical data and the surrounding scientific theory make it one of Precision’s most exciting in vivo programs disclosed.
FDA Accepts Biologics License Applications for exagamglogene autotemcel (exa-cel) for Severe Sickle Cell Disease and Transfusion-Dependent Beta Thalassemia
United States
Exa-cel is the first CRISPR gene-editing filings to be accepted for review by FDA.
Sickle-Cell Disease (SCD) — 25,000 immediately addressable patients
The FDA has granted Priority Review for exa-cel’s use in SCD and provided a PDUFA target action date of December 8, 2023.
Transfusion-Dependent Beta Thalassemia (TDT) — 7,000 immediately addressable patients
The FDA has granted Standard Review for exa-cel in TDT and provided a PDUFA target action date of March 30, 2024.
With the PDUFA dates set, Vertex is preparing for a commercial launch as near to approval as possible, and is working with payers, as well as 50 authorized treatment centers to serve eligible patients. Following approval, there will be a multi-month process for treating patients.
Pre-treatment period where patients decide a genetic therapy is right for them, and are referred to an authorized treatment center by their hematologists and begin the cell collection process.
Manufacturing period where a patient's cells are edited and become the exa-cel drug product.
Treatment period where a patient receives myeloablative conditioning, then their edited cells, and is followed for successful engraftment — multi-week hospital stay
Europe
In Europe, the Marketing Authorization Applications (MAAs) for exa-cel were submitted in December 2022 and validated by the European Medicines Agency (EMA) and the Medicines and Healthcare products Regulatory Agency (MHRA) in January 2023. In the EU, exa-cel has been granted Orphan Drug Designation from the European Commission, as well as Priority Medicines (PRIME) designation from the EMA, for both SCD and TDT. In the U.K., exa-cel has also been granted an Innovation Passport under the Innovative Licensing and Access Pathway (ILAP) from the MHRA.
WELCOME
SUMMER INTERN CLASS
+++2023+++
10% OF ALL BLUE ROOM REVENUES GO DIRECTLY TO FUND OUR NON PROFIT TOGETHERISM.
WE CAN ACCOMPLISH ANYTHING TOGETHER.
These materials do not purport to be all-inclusive or to contain all the information that a prospective investor may desire in considering an investment. These materials are intended merely for preliminary discussion only and may not be relied upon for making any investment decision. Any discussion or information contained in this presentation does not serve as a receipt of, or as a substitute for, personalized investment advice from Blueroom or your advisor.
This publication does not constitute an offer to sell or a solicitation to buy any securities in any fund, market sector, strategy or any other product. Investing is speculative and involves substantial risks (including, the risk of loss of the investor’s entire investment). Past performance is not indicative of future results, and there can be no assurance that the future performance of any specific investment, investment strategy, or product will be profitable.
For more information about us and our general disclosures contact us directly.