MONOLITHIC POWER SYSTEMS, INC. (MPWR) stock intrinsic value fair value estimate & owner earnings analysis

What the business is. Monolithic Power Systems designs high-performance power-management chips and modules - the parts that convert and regulate electricity inside servers, AI accelerators, memory and storage, notebooks and graphics cards, cars (driver-assistance and infotainment), optical networking gear, home appliances and industrial equipment. It owns its process technology but outsources wafer fabrication to foundries in China, Taiwan, South Korea and Singapore, with assembly and test in Asia. Revenue has grown from $333 million in 2015 to $2.79 billion in 2025 (about 24% a year), falling meaningfully in only one year of the last decade (2023, when revenue was flat and profits dipped). 2025 growth of 26% was broad-based: storage and computing +46%, automotive +43%, communications +37%, industrial +35%, consumer +26%, while enterprise data (AI and cloud servers) actually slipped 2%. That matters: the company is not a single-customer AI story. Gross margin has been steady at about 55% for three years. There is no debt and roughly $1.4 billion of cash and short-term investments at June 30, 2026. About 85% of sales go through distributors, three of which handled 54% of 2025 revenue, and 55% of 2025 revenue shipped to customers in China (largely contract manufacturers), with another 20% to Taiwan. How we measured earnings. We use owner earnings: operating cash flow, with working-capital swings removed, stock compensation subtracted as the real cost it is (about $227 million in 2025), and only the capital spending needed to maintain the business (we estimate growth spending as the company's long-run 7.4% capex-to-revenue ratio applied to each year's revenue increase; the rest counts as maintenance). This deliberately ignores a 2024 accounting gain: the company booked a $1.1 billion deferred-tax benefit (restated down by $195 million in February 2026 after an error was found, which also led to a disclosed control weakness) for a ten-year foreign tax incentive that began in 2025. The restatement did not change cash flow. The incentive does lower actual cash taxes through 2034, and that benefit is inside recent owner earnings, so we remove roughly 8% of owner earnings when it ends. Owner earnings per share were $9.45 (2022), $8.25 (2023), $7.64 (2024) and $11.80 (2025). Earlier years were much smaller and lumpy (2017 and 2019 were depressed by unusually heavy capital spending). Starting point. Chips are cyclical, so we anchor on the median of 2022-2025, a full cycle from the 2022 boom through the 2023-24 inventory correction to the 2025 recovery: $8.85 per share. The business has clearly outgrown that level. Owner earnings rose 54% in 2025, and the first half of 2026 confirmed it over two consecutive quarters: revenue up 37% (second quarter up 48% to $981 million) and owner earnings about $7.23 per share versus roughly $4.90 a year earlier. Our base case puts 2026 at $13.50 per share - 14% above 2025, well below the 54% just achieved, and below a simple doubling of the first half. Base case ($408 per share). Growth slows as the company gets bigger, and two chip downturns are built in (a 10% earnings drop in 2028 and 8% in 2033, similar to 2023-24), with recoveries of 12-22% in other years. Owner earnings reach about $28 per share by 2035 after losing the tax incentive, about 8% a year from the 2026 level. After that, earnings grow 4% a year for ten years. The reason is the company's 28-year record of winning new designs in each new product generation using its own process technology, which does not depend on any single product lasting. Earnings then hold flat for a decade and fade to zero over the following ten years. There is no permanent value beyond that. Bull case ($681 per share). 2026 reaches $14.50, which is just the first-half pace repeated. Share gains keep going in automotive, memory/storage, optical communications and AI power at 18-25% a year for several years, closer to the company's own ten-year revenue record, with only one mild downturn (-5%). Earnings reach about $47 per share by 2035, then grow 5% a year for a decade before holding flat and fading. Bear case ($135 per share). 2026 lands at $12, which is what the business has already earned in the first half plus a weak second half. Then three things the company itself discloses go wrong together. A chip downturn arrives. Sales tied to China (55% of revenue) are hit by US-China trade or export restrictions. And the foreign tax incentive is undercut early by new OECD global minimum-tax guidance, which the company names as a risk. Together they cut earnings 20% in 2027 and 10% in 2028. Growth resumes at only 6% a year, earnings hold flat after 2035 and fade to zero by 2055. Debt is not a risk here because there is none. What would make this wrong. The most important assumption is that design wins keep broadening across end markets. The 2025 filing supports it - growth came even as AI server revenue fell - but a lasting loss of position at the largest AI and data-center platforms together with Chinese domestic substitution would push results toward the bear case. On the upside, the base case may understate 2026-2027 if the first-half pace holds. The payment records reconcile to reported cash flow; the only gaps are small non-cash items the company lists itself (interest accretion on Treasury bills and deferred-compensation gains). Price discipline. Because the range is wide ($135 to $681) and the business is cyclical with heavy China exposure, the margin-of-safety price carries a large discount: roughly $240 per share, about 40% below our base value.