Every processor needs about one volt at huge currents. Monolithic Power sells the converters and modules beside it, to server, car, phone and industrial makers, mostly through distributors.
The company · 2 · What exactly does it sell?
Data centres are the biggest market
Each $1 of sales, second quarter of 2026
39%20%16%13%6%6%
Data centresStorage and PCsCarsNetworksConsumerIndustrial
Data-centre sales grew fastest, on demand from AI servers.
The company · 3 · How does it make money?
Chips sold per server, car and phone
$506.7Mdata-centre sales last quarter
55%kept per $1 after making it
About 85%of 2025 sales through distributors
Most sales go through distributors, and orders can still be cancelled.
Is it a good business? · 4 · What gives it an edge?
Its own secret chip-making recipe
1 of 10
power-chip makers in NVIDIA's 800-volt design
2024
a big AI customer moved work away
Its own chip-making recipe, installed in suppliers' factories, fits more of the power circuit on one chip, but the edge is not permanent.
Is it a good business? · 5 · What is the catch?
Infineon leads and is sold out
InfineonAI power leader, prepaid deals
TI, onsemisell the same kind of part
Vicorpower modules for the same racks
Listed rivals: IFX.DE · TXN · ON · ADI · VICR · AOSL · NVTS · 2577.HK
Racks move to 800 volts from 2027, and Monolithic is only sampling those parts.
Is it a good business? · 6 · What does its record show?
Sales tripled with profit rising
Sales, $B
1.2
1.8
1.8
2.2
2.8
3.7
Dec 2021Dec 2022Dec 2023Dec 2024Dec 2025last 12 mo
Sales$1.21B → $3.74Bsales, 2021 to the past year
Profit22% → 29%operating profit per $1
Kept57% → 55%kept per $1 after making it
Chip costs track sales but operating costs do not, so growth widens profit.
Is it a good business? · 7 · How much room is there to grow?
Room exists; the question is its share
$86.5Banalog and power chip market, 2025
161 GWdata-centre power capacity, 2026
$2.79Bits total 2025 sales
Nobody publishes the exact market for AI power chips, so it can only be bounded.
Is it a good business? · 8 · Who is in charge?
The founder still runs it
1997Michael Hsing founded it
27 Octobera new finance chief named
Hsing invented its chip-making recipe and still runs the company as chair and chief executive.
The story right now · 9 · Why do AI chips need power chips?
AI processors draw enormous current
+165%its data-centre sales growth
57%share rise over the year
800 voltsracks move there from 2027
AI processors draw far more power than older chips, so each needs more and better power converters beside it. Suppliers of those chips are growing fast.
The story right now · 10 · Why does the order type matter?
Prepaid supply is firmer than open orders
AI demand surgesEveryone wants power chips.
Infineon sells outCustomers prepay.
Monolithic's orders openCan be cancelled.
Demand pausesOrders vanish first.
The price already pays for sales nearly doubling from 2025 to 2027 with profit per dollar held.
The story right now · 11 · What changed lately?
Sales up 63% and a higher forecast
$1,202.4Msales in the September quarter
+165%data-centre sales
−3.8%storage and computing on the quarter
$1.315B to $1.385BDecember forecast
Strong quarter
In September GlobalFoundries agreed to make its chips in Singapore.
The story right now · 12 · What is everyone arguing about?
The cleanest AI play, or priced?
Many investors think
One of the cleanest ways to own AI power
We think
Real growth, but the price pays through 2027
Against us: three straight beats, and it has always cost more than peers.
The price and our view · 13 · What is a share worth?
The price sits between our value and sunny case
$932what we think a share is worth · the last close is of that
$932our value
Attractive below $746
Neutral
Unattractive above $1,119
$376gloomy: growth halves, less kept per $1
$932our estimate: experts' path, then slowing
$1,411sunny: growth beats the forecast by a quarter
Value per share in three futures. None is a prediction of the share price.
We estimate what a share is worth from the cash the business should earn over many years. Our rule: buy only when the price is at least a fifth below our estimate, as a cushion for being wrong.
The price and our view · 14 · How could we be wrong?
AI orders could slow
AI orders slowBiggest
Orders can be cancelled.
Infineon takes shareShowing now
Adding capacity.
Late to 800 voltsWatch
Sales late in 2027.
Its AI shareUnknown
Nobody publishes it.
A repeat quality slip, its accounts and a lawsuit add to the risk.
The price and our view · 15 · How will we know?
February tests the December forecast
Sales within the December forecastGood sign
800-volt parts win ordersGood sign
AI orders cancelledBad sign
FEB 20275
Fourth-quarter 2026 results, 5 February. Companies report every three months; each report is a test of the story.
Our view also changes on price alone: if the shares fall to within a fifth of our estimate, we move to Neutral.
The price and our view · 16 · So, what do we think?
An excellent power-chip maker at a full price
$932our value
Attractive below $746
Neutral
Unattractive above $1,119
UnattractiveMedium conviction
Monolithic Power designs the chips that feed processors steady low-voltage power, and has grown fourteen years running with no debt. But the price already pays for sales nearly doubling from 2025 to 2027.
↑ ↓ to move · C for contents
What it does
More about the company
Who it is
Every processor needs about one volt, delivered steadily at huge currents. MPS sells the chips that make that voltage from the power a board receives: converters, controllers and modules that sit right beside the processor.
Its buyers are server, car, phone and industrial makers, most reached through distributors. Unlike most chip designers, MPS writes its own manufacturing recipe and installs it in its suppliers' factories. NVIDIA lists it among ten approved power suppliers for its next racks.
How it makes money
MPS sells chips by the unit, mostly through distributors, and keeps about 55 cents of each sales dollar after the cost of making them. Its 2025 sales of $2.79B came from six markets: data centres, storage and computing, cars, communications, consumer and industrial.
Profits are high. In Q2 2026 it kept 31 cents of each dollar as operating profit, or 37.5 cents before stock pay to staff.
Growth and profit
What growth costs
Growth has made MPS more profitable: research, sales and admin costs rise more slowly than sales, so each new dollar of sales keeps more profit. The chips themselves cost about the same share of every sale. A flat year reverses it, as 2023 showed.
Its record
The numbers
Is it growing?Fast
Sales grew 40% over the twelve months to Sep 2026, faster than the 26% in the year to Dec 2025.
Is it profitable?Strong
Operating margin is 29%, up from 26% in the year to Dec 2023.
Does growth pay off?Pays off
Running costs take 26¢ of each sales dollar, down from 35¢ in the year to Dec 2021.
Does it earn good returns?Strong
Each dollar put into the business earns 24% a year after tax, about where it was (21%) in the year to Dec 2021.
Is it making cash?Strong
It turned 16% of sales into free cash, $609M over the last twelve months.
How steady is it?Swingy
Yearly sales growth has ranged from +1.5% to +49% over the last 5 years.
Are shares being diluted?Stable
The share count rose 2.7% in 4.8 years, and pay in shares costs 6.8% of sales.
Why it wins
The evidence
Most chip designers use whatever process their factory offers. MPS wrote its own and keeps it secret, so it can put more of the power circuit on one chip. That makes parts smaller and more efficient, which wins the spot beside a new processor.
Outsiders back this: GlobalFoundries is installing the MPS process in a Singapore plant, and NVIDIA lists MPS as an approved supplier. But the edge is not permanent. In 2024 a big AI customer moved work to rivals over quality problems.
Who could beat it
The rivals in full
Infineon is the leader in AI power, sold out and signing prepaid deals. Texas Instruments, onsemi and Vicor sell the same kind of part into the same racks.
The bigger shift is racks moving to 800 volts from 2027. MPS is only sampling those parts and expects sales late in 2027, behind rivals already shipping.
Each rival
Infineon AI and data-centre powerIFX.DE
MPS's multiphase controllers and power stages on AI accelerator and server boards; Infineon has prepaid capacity deals and says it is on allocation.
Texas Instruments, onsemi and Analog Devices power ICsTXNONADI
MPS's DC-DC regulators and multiphase controllers in servers, cars and industrial designs; onsemi bought processor power technology in 2025.
MPS's power stages and modules beside the GPU, by delivering current from under the processor.
Alpha and Omega multiphase controllers and smart power stagesAOSL
MPS's server and graphics power stages, at the cheaper end of the same socket.
Navitas and Innoscience gallium nitride for 800-volt racksNVTS2577.HK
The high-voltage conversion stage in 800-volt racks that MPS is only sampling and expects revenue from late 2027.
Renesas multiphase server power6723.T
MPS's server CPU and accelerator power in the same platforms; named by MPS and listed by NVIDIA.
How big it could get
The market and who has signed
The market
Nobody publishes the exact market, so it can only be bounded. Analog and power chips were an $86.5B market in 2025. Data-centre power capacity is growing about 31% in 2026, to 161 gigawatts. Infineon alone expects over EUR 1.6B of AI power sales this year.
Against that, MPS sold $2.79B in total in 2025. Room exists; the question is how much of it MPS keeps.
Who runs it
Leaders, pay and the cash
Who runs it
Michael Hsing founded MPS in 1997, invented its chip-making process and still runs it as chief executive and chair. On 27 October the board named Tony Balow, a finance vice president since 2023 and 23 years at Intel, as finance chief, ending the interim period; the long-time controller became chief accounting officer. Insiders own under 4%, most of it Hsing's. Their yearly bonus rides entirely on adjusted operating profit, and their shares on sales growth and the stock's return against other chipmakers; nothing rewards cash or clean books.
What they did with the money
MPS has no debt and $1.41B of cash. In the first half of 2026 it paid $179.5M in dividends and bought back only $4.0M of shares, while stock pay to staff keeps adding about 2.6% to the share count a year.
The last report
The latest report
MPS's quarterly sales hit $1,202M, up 63% in a year, and it guided higher
What it did to the thesis Strong quarter
It confirmed the AI data-centre line is still driving growth with margins held, and the next guide runs above what we thought 2027 would bring. It left open whether customers are buying ahead, which the next quarters must show.
The report
Q3 2026 was another record. Sales rose to $1,202.4M, up 63% on a year and 3.6% above the top of the company's own range, and data-centre sales reached $506.7M, up 165%. Margins held: about 56 cents of each dollar after making the chips.
Management guided the December quarter to $1.315B to $1.385B, a yearly pace of about $5.4B. Elsewhere growth was mixed: storage and computing fell 3.8% on the quarter and cars rose 19% on a year.
What's happening now
The background
The shares fell from $1,689.89 on 3 June to about $1,313 within the month, as the whole chip sector sold off, then recovered after the June-quarter beat. Half of the past year's 57% rise came from buyers paying more per sales dollar, not from sales.
In September GlobalFoundries agreed to make MPS chips in Singapore. The September-quarter report on 29 October cleared its guide; the next test is the December-quarter report in early February.
Where we disagree
Both sides in full
The market
MPS is one of the cleanest ways to own AI power, with a decade of growth behind it and data-centre sales still speeding up.
We believe
The business is excellent and the AI growth real, but the price already pays for it lasting through 2027.
The argument
The market reads management's rising growth floors. We also read Infineon's report: the leader is sold out and has customers prepaying for years of supply, while MPS's orders can still be cancelled. MPS has lost AI work once before.
Against us: MPS has beaten forecasts in each of its last three reports and has always cost more than its peers, and kept growing anyway.
What it's worth
How we value it
Valued on the analysts' revenue path for three years, each later year closing 30% of the gap to 4%, on official operating profit with stock pay a cost and tax paid in full; the thesis's own evidence is the check on that path.
What the price assumes
Sales growth in the first year, then slowing as ours does, at the same profit per sale.
The price assumed
54.2%
Our estimate
58%
Measured at the close of
6 Nov 2026
The three cases
Bear case$376
Growth runs at half the analysts' path, 29%, 15% and 9%, the margin settles a fifth lower at 25.9%, and new capital earns only the 8% rate: the downturn test.
Our value$932
Revenue grows 58%, 31% and 19% over three years, the analysts' path (C54), then fades toward 4%, closing 30% of the gap each year; the margin moves to 32.4%, the analysts' year-three figure after stock pay and on the filed basis (0.9% less).
Bull case$1,411
Growth runs a quarter above the analysts' path, 72%, 38% and 23%, the margin reaches 35.6%, and new capital earns 15%.
What could go wrong
Every risk
The risks
AI orders slow
Orders can be cancelled; if data-centre buyers pause, sales fall while the price still assumes growth.
Infineon takes share
The leader is adding capacity with prepaid customer deals and could price to win sockets.
Late to 800 volts
Racks move to 800 volts from 2027, and MPS expects revenue from those parts only late that year.
Quality slip again
A key AI customer moved work to rivals over quality in 2024; it could happen again.
Accounts and a lawsuit
MPS restated 2024 accounts, has an unfixed control weakness, and faces a fraud suit a judge let proceed.
Asia and one founder
94% of sales go to Asia, the founder is chief executive and chair, and the finance chief is interim.
What we still do not know
How much of AI power MPS supplies against Infineon; nobody outside publishes a share.
How much of data-centre sales is AI accelerators rather than ordinary servers and storage.
Whether the business lost in 2024 has fully come back, and on what terms.
When the accounting control weakness will be fixed, and who will be the permanent finance chief.
What changes our view
Every route to a new view
Every route to a new view
→ Neutral
The share price falls to about $1,119, back within 20% of what we think the business is worth.
→ Attractive
The share price falls to about $746, 80% of what we think the business is worth.
→ Attractive
Quarterly sales reach $1.5B, up from $980.6M.
The next test
It supports the case if
Revenue at or above $1,315M, gross margin at or above 55.5%, enterprise data still growing on the quarter, and a March guide at or above $1.45B.
It pressures the case if
Revenue below $1,315M, gross margin below 55.5%, enterprise data down on the quarter, or a March guide below $1.35B, flat on the December guide.
We would also be wrong if
MPS signs or files a multi-year AI supply deal with a stated volume.
The 2026 annual report again finds the accounting control weakness.
The verdict
Scores and how we weighed it
The scores
Growth
4 of 4
Business quality
3 of 4
Financial position
4 of 4
Execution record
3 of 4
Per share capture
3 of 4
Potential
2 of 4
Risk profile
3 of 4
Valuation
2 of 4
Evidence quality
3 of 4
How we weighed it
ValueNegative
The price sits above our value, which already takes sales nearly doubling.
BusinessPositive
Long growth with no debt, and steady power chips for processors.
GrowthPositive
The AI line drives growth, and the next guide runs above what we expected.
RisksNegative
AI orders can be cancelled, Infineon may price to win, it is late to higher-voltage racks and quality has slipped before.
RivalsNegative
The high price per sales dollar is its own, not the group's.
The groupMixed
AI chip names move together on spending news.
Latest newsPositive
The last report confirmed the AI line with margins held.
In plain words
Our estimate of value
We forecast how much cash the business can earn over the next ten years and beyond, then ask what that stream is worth today to an investor who wants an 8% yearly return. We do it three times: a gloomy future, an expected one and a sunny one. We call a share Attractive only when its price is at least 20% below the expected estimate, and Unattractive when it is 20% above.