One Physics Moat, Five End Markets
A CEO interview and a full analysis of a high growth, high margin of safety opportunity
Today’s company just printed a quarter that was a substantial upside surprise and the growth came from every part of the business at once rather than one lucky order.
What makes it interesting is where the demand is coming from. This company sells components into five markets that are all inflecting at the same time. Any one of those end markets could eventually be larger than the entire company is today. The technology is genuinely differentiated at the physics level and it wins in exactly the places where conventional silicon cannot follow.
The robotics and automation angle is one I keep coming back to. Every robot needs to know where its own body is before it can do anything useful, and the more dexterous the machine, the more of those measurements it needs. This company’s parts sit in the hardest version of that problem: the fingers on high DoF hands, the wrists, the ankles, the places where the movement is small and the precision requirement is brutal. Content scales with dexterity, so the same design win is worth more in each successive generation of robot. They already have production design wins here, largely unannounced, alongside a large sample pipeline.
Humanoids get the attention, but management pointed me toward the less glamorous and much larger near-term opportunity sitting next to it. Factory automation and industrial robotics are volume markets today and this company already ships into them. Their components also carry the high-speed data between robots and the controllers running the floor, so they benefit from automation build-outs in two separate ways.
The opportunity exists because almost nobody is looking. There is no investor relations function. There is no forward guidance. There is no sell-side analyst coverage at all. No estimates, no models, no notes. They have design wins in robotics and factory automation that have not clearly been announced because there is no mechanism by which they would be announced.
To understand this business you have to read every filing/transcript, pull contract amendments, and talk to management directly. I did all three, including a conversation with the incoming CEO this past week where he gave me two separate markers on how large he believes this company can become.
Full write-up below: the company, the technology, the end markets, the growth math, what the CEO told me, the potential valuation, and the risks.
The company
There is a semiconductor company in Eden Prairie, Minnesota with a $532 million market cap that just grew revenue 81% year over year, runs an 81% gross margin, holds a net-cash balance sheet, pays a 3.6% dividend, and trades at roughly 20x its current run-rate earnings.
Its quarterly earnings call is hosted on Google Meet and answered mostly by shareholders who have owned the stock for twenty years. The ticker is $NVEC.
NVE Corporation makes spintronic components. Magnetic sensors and digital isolators built on tunneling magnetoresistance, sold into medical devices, factory automation, robotics, and power conversion. Forty-two employees. One fab. More than 50 issued U.S. patents.
For years the story was range-bound. Revenue peaked at $38.3M in fiscal 2023 on the chip shortage, then reverted to $25.9M in FY25 and $26.3M in FY26. Then fiscal Q1 2027, the quarter ended June 30, printed $11.03M in revenue, up 81% year over year. Net income rose 79% to $6.39M, or $1.32 per diluted share. That is a 44% sequential jump off a Q4 that was itself the strongest quarter of FY26.
Two details matter more than the headline.
First, the growth was broad. Not one lumpy defense order, not a single distributor restock. The 10-Q language is specific: increases in both defense and non-defense sales, and increases through both direct and distributor channels. Every vector up at once.
Second, gross margin went up to 81.3%, from 80.6%. A bear argument on $NVEC was that growing through distributors would grind margins down. Volume leverage more than offset the mix. Operating expenses fell to 15% of revenue from 19%. Roughly 70 cents of every incremental revenue dollar landed in operating income.
What the market cannot see
Most small caps are underfollowed. $NVEC is genuinely opaque. Their opacity is the whole reason a business this good with the tailwinds they have trades where it does.
There is no IR function. Management said to me directly: the company is customer focused rather than shareholder focused, and it has not been as open on the shareholder side as shareholders (or I in the future) have wanted. They are evaluating a shift to be more open. The filings run to the legal minimum. No backlog disclosure. No product-line or end-market segmentation. The 10-Q gives you two revenue lines, product sales and contract R&D, and that is it. There is no guidance. There is no analyst coverage.
So when $NVEC lands a design win in robotics or factory automation, there is no press release, no bookings line, and no analyst note. It shows up in the revenue number quarters later, and by then it is mixed into a single aggregate figure with everything else.
The company has design wins in robotics right now, including humanoids. Management confirmed to me they have small design wins in production, alongside “many, many” samples out with customers, concentrated in their newest wafer-level chip-scale sensors. None of that has been announced cleanly. None of it is in a model anywhere, because there are no models.
To get the story on this company you have to read every transcript, read full filings, talk to management directly, and then do your best to connect all the dots. Very few people are doing that on a ~$500M market cap.
Quick note on the chart. It is approaching a monthly breakout after building a base for over a decade above the IPO base breakout level. Really beautiful high timeframe setup.
The technology: why spin beats charge
The semiconductor industry has operated on a single physical idea since the 1950s: pushing, pulling, and containing electrical charge. Electrons get pumped through microscopic channels on silicon wafers and stored in buckets called capacitors. That framework defined the technological age we live in. Most people never notice the inefficiencies built into the structure itself.
The largest of them is power. Pushing charge through silicon generates heat, and stored charge in capacitors leaks over time. To hold data in a state, and to protect the system that runs the digital age, conventional devices demand continuous, draining power.
$NVEC attacks the problem from outside the domain of electrical charge entirely. Instead of manipulating an electron’s charge, it builds chips that manipulate electron spin.
Every electron carries an intrinsic property called spin, which gives it a permanent magnetic orientation. Electric charge is variable and leaks. Spin does not. Semiconductors built on this principle are called spintronics.
A genius sandwich
Under an electron microscope, a conventional semiconductor looks like a city: an intricate lattice of roads (channels), buildings (capacitors), and sites for shuttling electrons around. $NVEC’s chips are built on something far simpler, a sandwich.
Top layer: a conductive material whose magnetic orientation can be freely influenced by external magnetic fields.
Middle layer: an extremely thin film, a few nanometers across.
Bottom layer: a conductive material whose magnetic orientation is fixed.
When the orientations of the top and bottom layers match, electrons pass through freely. When an external force flips the top layer into opposition, the middle layer becomes a barrier and electrical resistance spikes.
$NVEC builds two structures on this concept. Giant Magnetoresistance (GMR) uses a metallic middle barrier. Tunneling Magnetoresistance (TMR) uses an insulating one. TMR rests on one of the stranger results in modern physics, quantum tunneling, where electrons pass straight through an insulating barrier that should classically stop them, provided their spin orientations align.
By measuring the resulting resistance as different forces act on the top layer, $NVEC can sense magnetic fields, isolate high voltages, and store binary data at little to no incremental energy cost.
The product suite
Sensors. Traditional sensors rely on the Hall Effect, where a magnetic field bends an electric current and slightly distorts voltage. Hall sensors need constant power to sustain that current, and their accuracy drifts with temperature. $NVEC’s sensors measure direct resistance changes inside the sandwich, and because TMR exploits quantum tunneling, even tiny magnetic shifts produce enormous deltas in resistance that are trivially easy to read cleanly. Estimates put the performance advantage of TMR over Hall at roughly 40%.
Data transfer isolators. In any high-voltage environment, electric vehicles, industrial power supplies, data centers, low-voltage control electronics must communicate with high-voltage lines without physical contact. Otherwise thousands of volts reach a car’s central display and cook it in seconds. Legacy setups use optical isolators, converting electrical signals into flashes of light to bridge the gap. Those light-emitting diodes are slow, power-hungry, and degrade over time. $NVEC places a microscopic coil over a GMR sandwich instead. A signal entering the coil creates a magnetic pulse, the pulse flips the top layer, and the resulting resistance is read as data on the far side. The result supports speeds well beyond 150 to 200 Mbps with none of the thermal degradation of the incumbent technology.
Memory. Standard memory holds data as electrical charge in capacitors, which leak and therefore require constant power. $NVEC’s MRAM stores data as physical magnetic orientations inside the sandwich, thousands of them per chip. Aligned orientations store a 0, unaligned a 1. Magnetic orientation is a physical state. It cannot leak or decay as long as the chip is kept away from strong external fields, so power can be cut from the system entirely without losing data.
The practical output of all this: their newest sensors measure 0.65mm on a side, roughly one third the area of the conventionally packaged version, and remain stable in fields above 9 Tesla, stronger than the most powerful MRI machines. An implanted device using one will not fail if the patient needs a scan.
NVEC has more than a business model moat, they have a physics moat.
The competitive picture
The average semiconductor company spends 16% to 18% of revenue on R&D. $NVEC spends 8%.
It can do that because it is the only pure-play spintronics vendor funded by U.S. government contract R&D to advance the foundational spintronic architecture. The contracts are for MRAM development in aerospace and defense. But an MRAM chip is nothing more than $NVEC’s signature sandwich replicated thousands of times across a wafer, so those dollars go directly into improving the core structure, and improvements to the core structure keep the isolators and sensors best in class too.
That is the elegance of the model. They are effectively paid to maintain the best spintronic technology in the United States, across every product line.
Let’s look at who they compete with.
Texas Instruments and Analog Devices dominate high-volume capacitive isolators and Hall sensors. That is fine for most customers, and it leaves correspondingly less competition in the segments where $NVEC wins, medical devices and defense. $NVEC has supplied Abbott for two decades and runs a 100% domestic, ITAR-registered fab in Minnesota. Its competitors’ own legal terms explain why that matters.
From TI’s official terms of sale:
“TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS.”
From ADI’s standard product disclaimer:
“Applications or uses as critical components in life support devices or systems are not authorized.”
The largest analog semiconductor companies in the world have contractually excluded themselves from $NVEC’s most valuable market.
Allegro MicroSystems is the largest magnetic sensor pure-play, and it added roughly $30M of GMR/TMR revenue by acquiring Crocus Technologies. Its own 10-K describes where that technology is headed:
“Structural technology shifts in automotive and industrial markets, including xEV, ADAS, AI data centers, and robotics, are driving demand... By aligning our R&D investments with high-growth technology trends and subjecting programs to disciplined return on investment (’ROI’) evaluation, we believe we can support our long-term growth targets.”
Allegro’s new flagship TMR sensor is engineered for high-speed power conversion in EV charging, clean energy, and AI data center power supplies. They will push TMR into the industry, which validates the technology, but the high-volume markets they are chasing sit well outside $NVEC’s core. Allegro is not spending money to develop ultra-low-power switches for implantable medical devices or radiation-hardened memory for defense programs. The line is drawn there. The one place they overlap is AI data center power, which Eames confirmed to me is a real growth vector for $NVEC as well.
Everspin Technologies is the primary publicly traded MRAM competitor, and the pattern repeats. Everspin chases high-volume standalone memory for storage controllers, industrial equipment, and casino gaming machines, applications that need memory density above all. $NVEC competes in custom, embedded spintronic memory for extreme environments: high radiation, anti-tamper security, zero standby draw. They barely overlap.
Sizing the opportunity
$NVEC sits inside several sectors with a sizable SAM, isolating only the high-conversion sub-segments where incumbent technology falls short of what the application actually requires and spintronics becomes the mandatory answer rather than an option. As customers launch next-generation products, a disproportionate share of the cumulative volume routes to $NVEC, and the phase-out of legacy technology that was capping performance acts as a second tailwind.
High-voltage isolation: $200M SAM. Digital power isolators across data centers, EV chargers, and solar microinverters represent a $2.7 billion market, with industrial equipment the largest slice at 28% and high-speed applications above 75 Mbps another 20%. Together that puts $NVEC’s addressable slice near $350M, serving high speed, high transient noise immunity, and zero degradation. Because $NVEC specifically supports up to 200 Mbps, I cut that conservatively to $200M. Legacy optocouplers and capacitive isolators work fine for most customers and hold the mass-scale volume, but they cannot survive where thermal degradation and sensitivity are unacceptable. This figure also excludes the rapid expansion of data center power that Eames confirmed as a direct demand driver, and a recovery in EVs with next-generation architectures would add volume on top. Expect roughly a 10.5% CAGR with upside on AI DC power.
Medical sensors: $75M SAM, and this one can be verified. $NVEC enters medical wherever ultra-precision or sub-microamp power draw is non-negotiable. Pacemakers are the clearest case, since they sit and sense inside a body for decades. Roughly 600,000 pacemakers are implanted annually, and over 1.2M devices when you include neurostimulators and drug pumps. Each typically requires one to two multi-axis sensors or wake-up switches, putting annual sensor demand near 2.5M units. At a standard ASP around $30, the implantable SAM is 2.5M × $30 = $75M.
We know Abbott is 37% of revenue, so sales to Abbott are roughly $27M × 37% = ~$10M. Abbott generates about $2.4B in pacemakers and ICDs following its acquisition of St. Jude Medical, at ASPs of $6,000 to $7,000, implying roughly 350,000 units annually. Multiply 350,000 × $30 and you get ~$10M, which matches the disclosed Abbott revenue almost exactly. Independent estimates putting Abbott’s market share at 18% to 20% land in the same place. The SAM checks out against the filings from two directions.
Medical navigation: $75M+ SAM. When I asked Eames where medical goes beyond Abbott, he named cardiac rhythm management as the anchor, then pointed at three extensions: neurostimulation, catheter navigation, and electromagnetic navigation. The first sits inside the implantable figure above. The other two are a structurally different and arguably better business.
Implantables are a one-sensor-per-patient-per-decade market. Navigation catheters are consumables. Roughly 97% of electrophysiology ablation catheters are single-use for infection-control reasons, so every procedure consumes the sensor content rather than amortizing it across a device lifetime. That converts a device market into a recurring one.
The volume supports it. More than 4.5 million catheter-based electrophysiology procedures are performed annually worldwide, and 3D mapping-guided ablation exceeds 78% adoption in U.S. tertiary centers. Electromagnetic navigation works by tracking a sensor at the catheter tip against an external field to locate it in three dimensions inside the body, which is exactly the measurement $NVEC’s parts exist to make, and the sensor must be small enough to sit in a tip threaded through a blood vessel.
Eames tells me their sensors are much smaller than competing technologies, so physicians can perform less invasive procedures inside small areas of the body. Legacy platforms still use 1930s-era glass reed switches, and the upgrade path runs through two levels: 3T and 7T pacemakers replacing glass switches with packaged GMR/TMR sensors delivering reliability and sub-microamp draw, then micro-implants using sand-grain-sized wafer-level package dies, which $NVEC is the first to offer with its new WLCSP sensor. Industry estimates put growth at an 11% to 15% CAGR as the category moves up the technology stack.
Sizing it conservatively: assume only half of those 4.5M procedures use navigation, one to two sensors per catheter, at a medical-grade ASP of $20 to $30. That produces a $50M to $100M range, and I am carrying the midpoint, $75M, as the SAM. Expect low-double-digit growth as pulsed-field ablation and AI-guided navigation expand procedure volumes.
The caveat worth stating: $NVEC has not disclosed revenue from navigation, and Eames flagged these applications as slower-moving than robotics or automation because of regulatory timelines. Customers are actively looking to use the technology, which is why it earns a place in the SAM, but the conversion runs on a medical clock rather than an industrial one.
Industrial automation and robotics: $250M SAM. Precision position and angle sensors in industrial machinery represent a $625M market, and roughly 40% of industrial robotics uses TMR-based sensing, putting the SAM at $250M. Automated factory motors run above 120 degrees Celsius, which strains conventional Hall sensors. TMR delivers precision without thermal drift or power drain.
Robotic arms are a natural place for $NVEC to win inside this category. Every joint on an articulated arm needs position and angle feedback, the environment is hot and electrically noisy, and the precision requirement rises with every axis added. That is the exact set of conditions where Hall sensors degrade and TMR does not. This is also a market being built out right now rather than someday. FANUC America announced a $90 million investment in March 2026 to acquire property and construct a new 840,000 square foot facility in Michigan for potential expansion of its U.S. robot manufacturing, targeted for completion in late 2027.
Humanoids sit a few years out, but Eames supplied the unit math: a single robot carries 100 to “a couple hundred” sensors, of which $NVEC provides ~10%, at ASPs of $1 to $10 depending on volume (vague I know). At production pricing near the low end, that is $10 to $20 of content per robot as a floor, with meaningful upside as hands gain degrees of freedom and mix shifts toward higher-value parts. $NVEC works through 37 distributors across 42 countries and is converting sampling into production design wins. Expect a conservative 10% to 11% CAGR in industrial automation, and triple digits in humanoids once the flywheel starts.
Aerospace and defense: $100M SAM. The radiation-hardened memory market, specifically magnetoresistive memory for mission-critical systems, is valued near $300M in 2026. $NVEC sits squarely in it, earning high-margin revenue from customers who need maximum performance in punishing environments, demand decades of reliability, and will pay for both. Filtering for applications requiring extreme radiation immunity, sub-microamp consumption, and ITAR compliance, where $NVEC’s domestic fab is a structural advantage, I cut this conservatively to $100M. $NVEC actively receives defense orders for its MRAM chips, and management expects the usual lumpiness to give way to stronger growth in FY27. Expect a predictable 5% CAGR, constrained by the pace of government contracting.
The five pillars total roughly $700M of highly conservative, high-conversion opportunity:
That excludes cross-application sales, any penetration beyond the high-performance niches carved out above, and same-customer growth.
Now put it against the business as it actually stands. $NVEC did $26.3M last year, and the Q1 print annualizes to a $44M run rate. Even on the higher figure, they are capturing roughly 6% of a SAM built specifically to exclude everything speculative. Getting to $150M in revenue means taking about a fifth of it. Getting to $100M means taking one dollar in seven.
The $700M is still deliberately understated in two places, both of which management confirmed are real. The isolator pillar is cut to $200M and takes no credit whatsoever for AI data center power expansion, which Eames named as a direct demand driver. The robotics pillar counts industrial automation only and assigns zero dollars to humanoids, despite management handing me the high-level per-robot unit economics. Both are carried at zero on purpose, because neither can be sized with the precision the rest of the build demands.
One deliberate omission beyond those. The end-of-life wave across TI’s and ADI’s legacy portfolios could be worth hundreds of millions in TAM on its own. The honest question is what share of those orphaned customers actually need $NVEC-level sophistication. It is hard to size. Eames framed these opportunities as strategic and intentional, describing $NVEC as choosy and unwilling to chase rock-bottom pricing.
Modest share in any two of the five pillars gets $NVEC well past $100M in revenue. They do not need to win everything. They need to keep winning the extreme-specification sockets they already win.
The re-rate mechanism
Sequential growth is not the thesis. It is the catalyst most likely to force the market to reprice the thesis it has not yet noticed, and it looks to me like a high-probability event rather than a hopeful one. The reason is that the near-term drivers and the long-term drivers are different things, stacked on top of each other.
What should drive the next three quarters.
Defense is recovering off a trough. Sales fell 67% in FY26 on procurement timing, management expects them to increase significantly in FY27, and Q1 already showed defense adding to growth rather than subtracting from it. Procurement recoveries run multiple quarters, and the comparisons ahead are easy.
Medical is expanding within an installed franchise. Cardiac rhythm management is the anchor, and it extends naturally into neurostimulation, catheter navigation, and electromagnetic navigation, where the advantage is that smaller sensors let physicians work in tighter spaces and perform less invasive procedures. Same customers, same qualification path, new sockets.
New products are ramping with customers who already buy from them. Management told me most of the new-product volume in Q1 came from existing customers designing in the newer parts, with newer customers ramping behind them more slowly. Existing customers convert fastest because the relationship and the qualification are already done.
Factory automation is the nearer-term robotics story. This is the one management pointed me toward directly, and the one the market overlooks in favor of humanoids. Their 10-K states the mechanism plainly: couplers carry high-speed data between robots and central controllers, and as automation expands, the need for higher speed and channel density grows with it. It is already a volume business for them, and it is far larger today than humanoid robotics.
Distributor inventories have normalized after the semiconductor downturn, and the expansion is complete and running. Order-to-shipment runs from two weeks to a few months, so demand converts to revenue inside a quarter or two.
What should drive the next three to five years.
Robotics content compounds as humanoids scale. Their design wins sit in the high precision sockets: the fingers on high degree-of-freedom hands, and the wrists and ankles where a small magnet moving relative to a sensor tells the robot what motion it just completed. Those are small wins today. They are also the sockets that multiply as robots get more dexterous, and their content per unit rises with degrees of freedom. At the $10 to $20 floor per robot, three million humanoids a year is $45M of revenue, larger than the entire company today, from the conservative end of the ASP range. Data center power conversion is one of the largest capex items in the industry and their isolators are aimed at it. The 800V DC transition and the shift to silicon carbide and gallium nitride power stages both favor higher switching frequency and higher isolation voltage, which is where $NVEC’s parts win. They sell through sub-assembly and power supply makers.
Medical navigation and neurostimulation carry long regulatory cycles, which makes them slow to start and very durable once they start. The end-of-life replacement business from Texas Instruments and Analog Devices is ongoing rather than a one-time event, and every socket won there is high margin revenue that lasts the life of the design. The near-term drivers are recoveries and ramps in businesses $NVEC already runs. The long-term drivers are new markets where the physics advantage is structural. Both pull the same direction at once.
The comps nobody is looking at
Here is the part that decides how this reprices. Because $NVEC reports on a March fiscal year, the quarters it is lapping are the weakest in recent memory.
FY2026: Q1 Revenue $6.10M | Q2 Revenue $6.40M | Q3 Revenue $6.20M | Q4 Revenue $7.65M
Those four quarters sum to the $26.3M reported for the year. Q1 FY27 already came in at $11.03M against a $6.10M comp, which is the 81% figure everyone saw.
Given our discussion above and the high probability for sustained and sequential growth, apply 15% to 25% sequential growth, all well below the 44% just printed, and look at what it does to the year-over-year line:
If NVEC can achieve 15% sequential growth, $NVEC prints roughly triple-digit year-over-year revenue growth in every remaining quarter of this fiscal year. Not one quarter. Consecutive quarters, at an 81% gross margin, from a company still running at fractional utilization.
Three forces, not one
The multiple should expand for three independent reasons that happen to be potentially arriving together.
Profitability. $NVEC’s operating margin was ranked better than 99.61% of 1,017 companies in the semiconductor industry, against an industry median near 5%. It runs an 81% gross margin, a 66% operating margin, and a 58% net margin. This is a top-half-of-one-percent operator on the metric that matters most.
Scale. Multiples expand as revenue bases grow, purely on liquidity, index eligibility, and institutional investability. A $26M-revenue company is uninvestable for most funds. A $60M company with a $1B+ market cap is not. Growing the base is itself a re-rating force independent of growth rate.
Growth. Triple-digit year-over-year revenue growth, sustained across a full year, from a profitable company with no debt.
Any one of these argues for a higher multiple. All three arriving simultaneously, into a stock with no analyst coverage and no guidance, is how violent re-ratings happen.
What the market pays for this profile
I went looking for historical precedent: companies under $150M in revenue, growing this fast, at anything close to this profitability. The honest finding is that there is essentially no clean comp.
Universal Display (OLED), 2011 to 2013. The best gross-margin analog in the market. Its gross margin has ranged between 67% and 87%, ranked better than 98.61% of 2,376 hardware companies, on a licensing and materials model. Revenue was in the $80M range in 2012 with rapid growth. Net margin has run in the low-to-mid 30s even at maturity, and the stock carried a multiple far north of 50x through its growth years.
Credo (CRDO), now. Not a scale comp, but the cleanest available read on what today’s market pays for triple-digit growth plus profitability plus a net-cash balance sheet. Fiscal 2026 revenue reached $1.34 billion, up 206%, with non-GAAP net income up more than fivefold, a gross margin near 68%, and a net margin in the low 30s. Its forward P/E sits near 40x. Previous multiples were much higher in 2025.
My point is not to pinpoint a multiple. It is a statement that the current multiple, if NVEC proves this growth is sustainable over multiple quarters, is indefensible against any benchmark you choose to apply.
What a real re-rate looks like
I am not going to pretend to know the correct multiple. What I will do is show what happens if the market eventually treats this the way it has historically treated companies with a fraction of $NVEC’s profitability. Say $NVEC proves over the next two or three prints that this is a durable triple-digit growth story rather than one lucky quarter, which is exactly what I expect. Apply a 40-50x multiple to FY27 earnings at a 60% net margin:
And that values the fiscal year in aggregate, not the exit rate. On the Q4 run-rate at 50x, the same scenarios produce $416, $473, and $535. I am not underwriting those numbers. I am pointing out that they fall out of conservative sequential assumptions, a net margin the company already earns, and a multiple in-line or below what the market has assigned to businesses with materially worse economics.
The more useful way to hold it: at $110, if $NVEC simply holds its current 20x multiple while earnings compound, you make money. Every point of multiple expansion on top of that is the market catching up to what the business already is.
What the CEO told me without telling me
Pete Eames takes over as CEO at the August 6 annual meeting. He is a physicist, at $NVEC since 2003, and he personally built the fab expansion this entire growth question rests on. He does not give guidance and he did not break that rule with me. He did leave three markers.
On capacity. I asked which constraint binds first as new products ramp into production, and what multiple of the current run rate the footprint can absorb. His answer, in his words:
“We run at a fractional utilization, so our capacity is quite a bit higher than our current manufacturing run rate. A factor of 2 is a very conservative estimate for the additional capacity that is available to us.”
He added that they see no binding constraint on headcount or throughput, that their outsourced assembly and test partners can absorb the growth, that they are bringing on additional partners, and that the in-house wafer-level line adds capacity on top of all of it. A factor of two off a $44M run rate is $88M, and he called that very conservative. They have additional floor space inside the building to expand quickly, and management regularly tours other local industrial facilities to stay ready if they need more. The path to $100M+ in revenue does not require a new fab, a capital raise, or a dividend cut. The capex is already spent. Fixed asset purchases were $57,000 last quarter, against $2.19M for all of FY26.
On the dividend. This is the one that stuck with me. $NVEC pays $1.00 per quarter, $4.00 a year, which has recently exceeded earnings. Eames told me nobody plans to cut it, that he is proud of it, and that as revenue grows he expects the payout ratio to normalize toward something traditional, at which point he would like to increase the dividend. A traditional payout ratio is 30% to 50%. For $NVEC to get there while holding a $4.00 dividend, EPS has to reach roughly $8 to $13. At their margins, that implies $65M to $107M in revenue, two and a half to four times last year. He described that as the precondition for raising the dividend, not as a stretch goal.
On the sales model. They run four full-time internal salespeople on top of 37 distributors across 42 countries. He noted, unprompted, that companies typically switch to a direct sales organization at $100M to $300M in revenue, and that they envision doing so when appropriate.
Two different answers, both benchmarking a company several times the current size. When I asked him directly what has to go right for $NVEC to become a $100M company, he said the things that have to go right are going right, that the market’s changes align with their technology and benefit proposition, and that they are seeing explosive growth and adoption of their new products. That is as close to a forward statement as this management team gets.
Pricing power, documented
$NVEC’s Supplier Partnering Agreement with Abbott was extended in December 2025 through December 31, 2027, with scheduled price increases for both 2026 and 2027. I pulled the amendment. It covers the same three part numbers as the prior amendment, with a fresh price step in each year.
Raising price on parts that a twenty-year customer has designed into implantable medical devices, where requalification is expensive and slow, is what pricing power looks like in practice. The amendment also states that both parties will work on aligning pricing beyond 2027, so they are already negotiating the next extension.
There is a second, quieter source of pricing power. When Texas Instruments and Analog Devices end-of-life low-volume analog parts, the designs those parts sit in still need a component. $NVEC markets replacements and commits to long-term support that the large incumbents will not. Eames told me an open socket from a competitor like TI can be a meaningful revenue line for a company their size, that they are choosy and take these only where they can hold margin, and that the current opportunity could already be hitting the books. That business is direct, high margin, and sticky for the life of the design.
What could go wrong
Concentration. Abbott was 37% of revenue in FY2026. The agreement runs to the end of 2027 and the filing states plainly that they cannot predict whether it will be renewed. This is the single largest risk in the story. The offset is that the FY27 broadening should mechanically reduce Abbott’s share even as Abbott dollars grow. Watch next year’s concentration note. A number below 37% confirms diversification is real.
Reversion. They have spiked before. FY23 hit $38.3M and fell back to $25.9M within two years. The counter is that this quarter’s strength was broad rather than concentrated in one order. The October print resolves it. A second consecutive quarter above $10M dampens the reversion argument, and another quarter of sequential growth likely kills it altogether.
Competition. TMR is no longer exotic. Allegro bought Crocus for $420M to own a piece of it. $NVEC holds pricing power where it is sole-sourced into extreme specifications and faces real pressure at the catalog end. The 80% gross margin depends on mix staying weighted toward the former.
Opacity cuts both ways. No guidance and no segment detail means the stock can gap hard on a single print in either direction.
Small company risk. One fab, one site, 42 people, and a first-time CEO. The mitigant on the last point is that Eames built the expansion himself and voluntary turnover runs 7% against a 16.4% semiconductor industry average.
Insider selling, disclosed. Outgoing CEO Dan Baker sold 20,000 shares between July 27 and July 29 at a weighted average of $114.91, gross proceeds of roughly $2.30M. His revocable trust also filed a Form 144. He retires from the CEO role on August 6 and remains Chairman.
The sale came paired with the exercise of 15,000 incentive stock options across six grants, at strikes from $47.20 to $81.96. The footnote on the Form 4 explains the timing: all of those options would expire three months after his retirement. Retiring August 6 puts the deadline in early November. All six exercises were cashless net exercises, so he paid nothing out of pocket and received 6,375 net shares after share withholding covered the strike cost. A portion of the sale covers the tax on that spread, which at combined federal and Minnesota rates runs somewhere north of $300,000. What he kept is the part worth focusing on. He held 59,928 shares going in, added 6,375 from the exercise, sold 20,000, and finished with 46,303 shares. That is a 22.7% reduction, leaving him roughly $5.1M of stock at current prices, close to 1% of the company, plus 2,500 options he did not exercise. He also stays on the board as Chairman. Nobody exiting a thesis keeps a fifth of a percent shy of 1% ownership and the chairmanship. The Form 144 registered 48,371 shares, which reads alarming until you notice that shares outstanding are 4,837,166 and 48,371 is exactly 1.00% of that, the Rule 144 volume ceiling an affiliate may sell in a rolling three-month window. Filing at the cap is routine housekeeping to preserve flexibility. He sold 41% of what he registered. The filing also shows the shares were acquired in November 2012 and that he had sold nothing in the prior three months.
Why this is the moment
The market started to reprice $NVEC off the Q1 print. That move was episodic, driven by a single headline number, and the stock has drifted back since. What the market has not yet processed is why the number happened: a completed capacity expansion, a portfolio of genuinely new products landing in genuinely new markets, existing customers adopting those products, new customers queuing behind them, defense recovering off a depressed base, and pricing power documented in filed contract amendments.
It also has not processed how many separate things have to go wrong for the growth to stop. Defense recovery, medical expansion into navigation and neurostimulation, factory automation, end-of-life socket wins, robotics content, and data center isolators are six independent drivers running on different clocks. A stumble in any one of them slows a quarter. Stopping the multi-year story takes most of them failing at once, and they do not share a common failure point. A business with 81% gross margins, a 58% net margin, no debt, no capex requirement, management-confirmed capacity to likely triple revenue, and five independent end markets that each could exceed its current size, is not a 20x earnings business. It is priced that way because it is small, quiet, and unfollowed.
The margin of safety is real. At 20x run-rate earnings with a 3.6% dividend and $43.9M of net cash, you are not paying for the robotics option, the data center option, or the automation option. You are paying roughly a market multiple for the medical and industrial business that already exists, and getting the rest for free.
Disclosure: I hold a long position in $NVEC. This is not investment advice. Do your own work. All figures are drawn from $NVEC SEC filings, earnings call transcripts, company materials, and a July 2026 conversation with management.
 | A guest post by
|
Well written piece and very interesting company, cheers!