2025 Annual Letter

Dear shareholders,

This is the first letter to you. As the year draws to a close and the holiday season approaches, Anthony and I would like to share our reflection on what has been a transformative year for our firm. In addition, we aim to report candidly on what we have done, how we think, and where we believe our advantages—and limitations—lie. 2025 was a pivotal year for us. We made a deliberate decision to pivot from software consultant and outsourcing business to investment and acquired businesses. In conjunction with the strategic shift, we also changed our name from AImazing Innovation, LLC to Umbria Asset Management, LLC. This decision was not driven any short-term opportunities. Over time, we realized that while we enjoy building businesses, we are more gravitating toward discussing and evaluating new ideas, assessing strengths and weaknesses, and allocating capital across multiple compelling opportunities than concentrating all resources on a single operating line. We believe our effort will be best rewarded by identifying, acquiring, and investing into attractive ideas than solely focusing on consult company. Our vision for Umbria is to invest in high-quality businesses and ventures across public and private market. We evaluated the traditional fundraising model—taking outside capital investors. However, attractive opportunities will not present to us as frequently, and we shouldn’t be pressured to deploy capitals or change the allocation frequently to match with any benchmarks. We concluded that external capital would introduce incentives that are misaligned with our disciplined and decision-making process. Therefore, Umbria should be built to become a self-sustainable operating company. That means the capital should be compounded internally, reinvesting excess cash flows from strong businesses into the next opportunities when prices and conditions are favorable.

Investment Thesis and Philosophy

Over the course of the year, we refined the investment thesis that will guide our decision-making in the years ahead. Buy-right-and-hold strategy stood out to us. There are three-key ingredients to buy right. First, the company should have durable competitive advantages or so-called moats. The durability of the competitive advantage will give the companies the power to increase its price above inflation, which will translate to stronger earning power. Moreover, we would like to own companies possess multiple competitive advantages as the defense. Multiple competitive advantages will enhance the durability further due to higher reproduction cost for new entrance, or it will less likely to face with disruption. The second ingredient is the predictability in the recurring revenue. Any good companies should be able to captivate customers’ demands. The return in purchases should be compulsory. Last but not least, we should be able to buy or acquire the company at the reasonable price, which will give us margin of safety over our analysis. It is worth to note that the reasonable price is relative to whether we invest into the companies as the passive investors or acquire the companies to improve the operations. When these conditions are met, we are comfortable to hold for extended periods. We believe long-term ownership will reduce the impact of short-term market noise and allow underlying business economics to drive the result.

Portfolio Updates

Market dislocation during mid-March to mid-April due to President Trump’s Tariff news put our framework to practice. We initiated three new positions with GE Aerospace (NYSE: GE), Advance Micro Device, Inc. (NASDAf: AMD) and Hims and Hers Health, Inc (NYSE: HIMS). While these businesses operate in very different industries, they share common traits such as possessing multiple competitive advantages, and reasonable stock price.

GE Aerospace (NYSE: GE)

General Electric has undergone a significant transformation in recent years, completing spin-offs of GE Healthcare in January 2023 and GE Vernova in April 2024, after which the remaining entity was renamed GE Aerospace. We will use GE Aerospace and GE interchangeably. GE Aerospace focuses on the design and manufacture of aircraft engines used across commercial aviation like Boeing and Airbus, business jets like Gulfstream, and U.S. defense applications. Our analysis on GE’s competitive advantages has concentrated primarily on the commercial sector where disclosure is more comprehensive and economic dynamics are easier to evaluate. The commercial aircraft engine industry that power roughly 40-million flights every year and carries millions of people around the world is dominated by four players: GE, Pratt and Whitney (PW), Roll Royces and CFM International (50-50 joint venture between GE and Safran). These manufacturers delivered on average around 2,000 to 2,500 engines every year. GE holds an estimated 60% installed-base market share including half of the market share from their venture with Air Safran. The market share is measured as the installed base which means number of engines had installed and are actively being used. We believe the industry will grow at mid-to-high single digit (i.e., 4-7% growth rate) until 2030. Although the growth rate is not as strong as technology company, this industry’s century-long history and extreme concentration—over 99% market share held by four firms—highlight formidable barriers to entry. The barrier to entry into the aircraft engine market is laid on intellectual property, decades of manufacturing expertise. Turbofan engine is being used the most in the commercial space. Each engine is comprised of more than 20,000 different parts for fans, compressors, etc. To make a reliable engine, the manufacturer needs to balance among aerodynamics and thermodynamics. Each fan is being spun at high resolutions around 10,000 rounds per minute. It will create the discrepancy in temperature between inside and outside the engine. Manufacturers need to have deal with thermodynamics to prevent the engine from exploding. Any new entrant must spend billions upfront for the R&D and thousands of hours to test the engine on the ground before the engine can be sold. Eventually, the newcomer will have to establish the brand name to ensure the purchasers that their engine is reliable enough. In 2024, GE reported approximately 70,000 installed engines comprised 45,000 engines for commercial and the rest for U.S. defense. Its narrow-body engines comprised of CFM56 and LEAP dominate roughly 75% of the segment. The narrow body is benefiting from superior fuel efficiency and lower maintenance requirements, which makes it good for domestic flights or shorter flights. In the wide body sector, their engine power 55% of the fleet while the second player is Roll Royces. GE’s financial performance reflects these advantages. In 2024, the company generated 20.4% return on invested capital and 24.9% return on equity. Their backlog grew to $175 billions compared to $125 billions of PW1. The key driver of this performance is GE’s emphasis on the maintenance, repair and obligation (MRO), which provides the steady stream of revenue with high margin. In addition, GE operates the largest global MRO network in which they either own or through a partnership. In the annual letter, the CEO committed to spend $1 billion to increase the MRO capacity. Last but not least, they have started to track the part record explicitly to control the key parts in the engine in real-time so that they can control their inventory tighter. At the time of our initial purchases, GE’s market capitalization approximated its required performance obligation at $200B with average life at 5 years. Despite strong conviction, we didn’t get into the position on reasonable price. Therefore, we didn’t have sizable GE position. Advanced Micro Device, Inc. (NASDAf: AMD) Advanced Micro Device designs CPUs and GPUs for personal computer (PC) and data-center applications, including its EPYC processor line and GPU stack. Since CEO Lisa Su assumed leadership in 2016, she has decided to close their fabrication and partnered with Taiwanese Semiconductor to manufacture their chips. This shift allowed AMD to focus on design innovation and accelerated product iteration. The results have been substantial in the race against Intel. Before 2019, AMD was always never considered as the competitor to Intel in both PC and data center for CPU. In 2016, they introduced the first generation of Ryzen. However, until 2019, with the third generation of Ryzen architecture, AMD was able to outperform Intel not only on technical benchmark but also the users can purchase their chip with fraction of Intel price. On the data center sector, EPYC architecture was introduced in 2017, and EPYC started to outperform Intel in the period 2019-2021. We can see the clear trend in revenues of both companies. AMD’s revenue in the Data Center sector has outgrown Intel since 2022. However, as 2025, Intel still retains meaningful share in data-center sector due to slow deprecation of old technology in data center so that the data center can sell to the customers at different tier price. We expect Intel market share will slowly be replaced by AMD chips. This story marked the strategy of AMD as they are trying to navigate themselves as the cost leader in the CPU and GPU space. The story was similar between AMD and Nvidia in the GPU for PC sector. Until today, Nvidia GPU always commands the higher premium price due to the programable capability 1 We used the total backlog of RTX, the parent company of PW through CUDA, Nvidia’s parallel computing platform and API. That means the users can change the way GPU is being used from only rendering the image or video for gaming, they can use the GPU for cryptocurrency mining or Artificial Intelligence (AI). During 2022, with the release of ChatGPT, Nvidia has become the most valuable company. To extend their monopoly, in 2024 they started to sell the data center rack which is a full-stack solution including CPU, GPU and memories to their customers to captivate them into Nvidia’s ecosystem. Since 2020, AMD has been strategically acquiring multiple companies with the ambiguous to offer full-stack solution for data center sector. This approach took longer to materialize because of the friction in integrating all acquired businesses into the same ecosystem. In addition, their ROCm system (CUDA-like product) was not ready until late 2024. That means even the customers can purchase AMD’s rack, they wouldn’t utilize all the power of AMD computing system. In mid to late 2024, we started to see the big lift in performance of AMD rack. Their MI300x had comparable performance to Nvidia’s GH200. Our analysis doesn’t require AMD to displace Nvidia as the market leader. Rather, we believe the performance gap between them will narrow as the hardware improvements become incremental and software abstraction layers mature. In addition, customers will want to diversify vendors so that they don’t have to pay for Nvidia’s premium. In recent year, there is initiative to build a universal language to interact with all the GPUs on the market. Although the performance of generalist language will never as close to native language, we believe they can help developers to get most of the tasks done unless they need a customization solution. As the cost leader, AMD will spend lots of money to enhance the R&D, integration and take the loss on selling the rack to widen the adaptation. When we think about the GPU industry, we believe both AMD and Nvidia are protected from multiple competitive advantages. For the new entrants, it took both companies years of R&D to produce the general GPU that can be used for different purposes through. If the new entrant can’t offer significantly better performance or ecosystem, they can’t enter the industry due to the solid brand name of both AMD and Nvidia. Within the industry, the switching cost is still high due to widely adaptation for Nvidia. However, it will reduce through time as we mention above. There are two substitution risks. First, many companies started moving into producing their own special chip to do AI training and inference. In our opinion, this will be short-term threat because AI training technique and architecture will be involving as time passes. To support this flexibility in training and experiment architecture, researchers must be able to have a general GPU like AMD and Nvidia. Second, it is possible that GPU may be obsoleted if researchers discover the new way of training the AI models with comparable or even better performance. The main risk of investing into AMD is the growth rate of AI industry may not be as we predict above. We agree with the statement as we don’t have a clear vision on how large AI will contribute to the growth of the economy as well as the demand for AI. We can only see the improvement of AI on micro-level at the current state. However, when the productivity is improved, there should be a channel to absorbed it; otherwise, it will become wasted. However, in the long-term we are optimistic that many new use cases will be introduced to utilize the power of AI. We started the purchase in early of March to early April. The average market cap for AMD when we purchased was around $160B (average $96/share). We believe this is about the fair value of AMD. Our assumption is that GPU market growth at 10% until 2030, and CPU market growth at 5% until 2030. In 2030, their market share will grow to 10% of global GPU market and 60% domination in CPU market. That will leave us at $94/share. If we calculate AMD fair value without the advancement in ROCm software, and they can’t capture more market share in the AI data center, the fair price should be $80/share.

Hims and Hers Health, Inc (NYSE: HIMS)

Hims and Hers is a direct-to-consumer telehealth company offering affordable solutions across sexual health, dermatology, mental health and weight loss. The company growth rate from 2021, 2022, 2023 and 2024, their growth rate compared to year before was 82.77%, 93.81%, 65.49% and 69.33% respectively. Although their operating margin was not meaningful (i.e., operating loss) for 2023 and around 4% in 2024, the return on invested capital of 2023 and 2024 is 22.29% and 50.31% respectively which is above the telehealth and healthcare industry 8% and 5% respectively. Notably, their growth was solely based on the discipline of the management without any long-term debt or borrowing from revolve credit. This is one the aspect that we like HIMS. We believe that the good management is not only the operators but also good capital allocator. As the good operators, the company can run efficient. However, without the capital allocation skills, the company may pay more tax than they should which will take away their capital to increase the value of their assets or redeploy into meaningful project. They are able to achieve this astonishing achievement because in the last two years, HIMS hit the blockbuster in weight loss where they sold the compounding version which is significant cheaper than Ozempic from Novo Nordic and Zepbound from Eli Lilly. Since 2022, the FDA put GLP-1 into its shortage list due to high demand, it triggered the 503A and 503B allowance so that many compounding pharmacies can start to produce the compounded versions. In February 2025, FDA decided to remove the shortage for Ozempic. This leads to the decline in HIMS stock price as they only had 90 days to off boarding customers who are offered compounded GLP-1. There is a loophole in this law that for prescription drug, the physicians are allowed to prescribe the compounded version. However, as of the Q3 2025, only 20% -25% of HIMS revenue comes from compounded GLP-1. We identified the first competitive advantage they had is their brand name. Each year, they have spent around 45-50% of their gross profit on the marketing budget to build the trust in the customers. During the Superbowl this year, they took the spotlight by raising the awareness about the obesity in the US and offered their weight loss program. The evident is that after the Superbowl ads, the engagement or search of HIMS surged, and their revenue in the first quarter was astonishing. However, it is not the short-term effect because despite the weakening in the weight loss market due to FDA regulations, their deferred revenue increases 57% as Q3 2025 compared to Q4 2024. The second competitive advantage is the vertical integration. In 2024, they acquired 2 different distribution centers to ship the drug. I learned that one needs license to be able to deliver drugs over mail. In late 2024, they acquired two compounding centers that have license for 503A and 503B. In early 2025, they acquired the peptide manufacturing facility. With this move, they extend the vertical integration where not only they can manage the inventory better but also cutting the cost compared to the competitors. The benefit is two folds. First, HIMS can easily introduce more offers which will captivate the existing customers longer on the platform. Second, they can easily become the cost leader in the sector where they introduce the offer because of vertical integration. As we can see that in Q3 2025, they are able to lower the cost of weight loss offer 20%. This may suggest due to the intense of competition. However, if their margin is consistently maintained, which suggested that their cost is reduced faster than the revenue. Again, the weight loss is only about 20-25% of the total revenue. We will need the longer time horizon to observe this effect. Looking forward, HIMS’ most compelling competitive advantage will be its personalization treatment platform. They acquired the new home testing lab and collaborated with fuest Diagnostics to offer their customers full blood test. Combined with all the competitive advantages above, they will be able to easily extend more offers and create better personalization treatment plan for the patient. Although the personalization tool is currently heavily depended on the physician’s prescription, the next investment is to collect more data and use AI as the assistant. This year, HIMS announced that they hired a CTO with AI-focus, and June 2025, Chief of Product to build more ergonomic system for their customers. In the long run, we believe the personalization health will be a strong competitive force as they understand the needs of patients compared to the competitors who enter the market as the generic drugs. There are two main risks about HIMS. First, they are operating in the generic drug market, which is the grey area of FDA. Therefore, at any moment, HIMS’ offer can be retracted due to regulations. They will also have to be extremely cautious about their advertising and language on those ads. Second, besides smaller competitors who are trying to replicate HIMS’ playbook, Amazon One Medical is the main competitor because they have enough resource to compete. However, we believe that with their current model, the company is more focusing on insurance-based patient than HIMS.

Looking ahead

In this coming year, we will begin the search for our first acquisition. We are evaluating both high-quality businesses aligned with investment thesis above and select distressed opportunities where asset values support turnaround potential. Next year is the good time to us to start the search since the interest rate started to decline, which will help us to acquire businesses at a more favorable rate. Acquiring a business with durable moats will give us latitude to make mistakes since the revenue of the business depends mostly on the competitive advantages. Moreover, in our investment thesis, the company should already have steady stream of recurring revenue where the customers will come back to us in predictable interval. The downside with this type of businesses is they will demand higher premium to acquire. With the distress company, the difficulty is laid in identifying whether the asset can turnaround the business. In addition, since we will operate the company, there will be smaller room for the errors in this case. Although it may seem that this type of acquisition doesn’t align with our investment thesis, the beauty of this acquisition is that we can acquire them at big discount to their intrinsic value. Ideally, the asset should demonstrate the potential of having competitive advantages. We believe that this type of search parameters does exist because any good businesses can become extremely bad if they are overleveraged, or unfortunate events happened to the businesses where it can be fixed or will go away in the short-term. While looking at the public market and private market, most of the companies that we understand and meet our criteria are currently overvalue. In addition, the market ahead is still remaining uncertainty for us as the weakening in the labor market, inflations and geo-politics. While macroeconomic remain uncertain, our investment decisions are driven by valuation and business fundamentals rather than macro forecasts. Thank you for all the hard work, and we’re looking forward to reporting our exciting progress years ahead.