Keystone sits inside the optical modules linking AI servers, cleaning the signal before the laser. The rest of MaxLinear sells broadband and Wi-Fi chips.
The company · 2 · What exactly does it sell?
Data-centre chips are now half
Each $1 of sales, second quarter of 2026
50%27%14%9%
Data-centre and networkBroadbandWi-FiIndustrial
The data-centre half grows fast; the broadband and Wi-Fi half rises and falls with the cycle.
The company · 3 · How does it make money?
Chips sold per module and per box
$250M-$270M2026 optical sales forecast
15.6%of revenue paid in shares
$13.6Mcash used last quarter
It earns per chip shipped, and paid chip makers ahead for 2027 orders.
Is it a good business? · 4 · What gives it an edge?
A hold on one chip generation
800G
links won; switching means a redesign
1.6T
the next speed, a new contest
Almost 40%
less power, management says
Winning 800G links does not win 1.6T, and the power advantage rests on management's word.
Built-in optics could drop the plug-in module, and the chip inside it, altogether.
Is it a good business? · 6 · What does its record show?
Sales fell by two-thirds, then rebounded
Sales, $B
0.9
1.1
0.7
0.4
0.5
0.7
Dec 2021Dec 2022Dec 2023Dec 2024Dec 2025last 12 mo
The fall$1.12B → $361Msales, 2022 to 2024
The rebound$361M → $661Mto the past year
Profit−62% → −0.9%operating profit per $1
Costs per dollar fell as revenue recovered, but share pay stays near a sixth of revenue.
Is it a good business? · 7 · How much room is there to grow?
A large market, an unknown slice
Almost 20 millionfast optical modules shipped, Q1 2026
Above $2B1.6T module chip sales expected, 2026
Most modules need a chip like Keystone, but MaxLinear's share is not public.
Is it a good business? · 8 · Who is in charge?
The co-founder still runs it
2003Kishore Seendripu co-founded it
About $50Mof stock he sold in August
7.7%owned by insiders in April
He sold about a ninth of his stake, outside a pre-set plan.
The story right now · 9 · Why do AI servers need optics?
AI servers talk over light
+72%MaxLinear's sales growth
+210%its data-centre sales
First half 2027first 1.6T chip sales
AI servers pass huge amounts of data between them, faster than copper wire can carry, so they use lasers and fibre. Each link needs chips to keep the signal clean.
The story right now · 10 · Why might the wins not last?
Each faster speed reopens the contest
Wins the 800G linksOrders arrive.
Market moves to 1.6TA new contest.
Built-in optics comeMay drop the chip.
Market growth slowsFrom 2027.
LightCounting wrote in September that optical market growth must be at its peak.
The story right now · 11 · What changed lately?
Sales up 72% and the optical forecast raised
$218.1Msales in the third quarter
$124.9Mdata-centre sales, up 210%
−8%broadband sales
25%share fall after the results
Solid quarter
Cash went out as it paid chip makers ahead for 2027 orders.
The story right now · 12 · What is everyone arguing about?
An AI optics winner, or one generation?
Many investors think
An AI optics winner compounding through 2028
We think
Real wins in one generation beside a cyclical half
We are wrong if quarterly sales reach $300M with 25 cents of profit per $1 after share pay.
The price and our view · 13 · What is a share worth?
The price pays for years of fast growth
$10.74what we think a share is worth · the last close is of that
$10.74our value
Attractive below $8.59
Neutral
Unattractive above $12.89
$4.64gloomy: growth halves, less kept per $1
$10.74our estimate: experts' path, then slowing
$15.12sunny: 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 · What if it all works?
As a bet on AI optics, it is speculative
Speculationthe kind of bet this is
3.0×if it works
42%kept if it fails
Multiples of the company's value when the bet was last struck, 4 Nov 2026.
Real upside if optical keeps growing, and a large loss if it stalls.
The price and our view · 15 · How could we be wrong?
Optics could drop the chip
Optics without the chipBiggest
Built-in optics.
Next generation slipsWatch
1.6T due 2027.
Broadband fallsShowing now
Down 8%.
Its market shareUnknown
Not public.
Share pay near a sixth of revenue keeps shrinking each owner's slice.
The price and our view · 16 · How will we know?
February tests optical sales
Optical sales at the raised forecastGood sign
Sales of $300M a quarterWe turn neutral
1.6T chips slipBad sign
FEB 20274
Fourth-quarter 2026 results, 4 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 · 17 · So, what do we think?
Optical chips revived it, but the price needs more
$10.74our value
Attractive below $8.59
Neutral
Unattractive above $12.89
UnattractiveMedium conviction
MaxLinear's new optical chips are winning real orders at AI data centres, but broadband and Wi-Fi, the other half, is cyclical. The price needs sales above its best year at nearly twice that year's profit per dollar.
↑ ↓ to move · C for contents
What it does
More about the company
Who it is
MaxLinear designs chips and has them made by TSMC and UMC. Its fastest-growing chip, Keystone, sits inside the plug-in optical modules that link servers in AI data centres: it cleans up the fast electrical signal before a laser turns it into light.
Module makers and the cloud companies behind them buy it, management says, because it uses almost 40% less power than rival chips from Broadcom and Marvell. No outside source confirms that yet. The rest of the company sells chips for internet boxes, Wi-Fi routers and industrial equipment.
How it makes money
MaxLinear sells chips one purchase order at a time, mostly to equipment and module makers, with nearly half going through distributors. In Q2 2026 half of sales came from Infrastructure, the optical chips plus radio and analog parts, and half from broadband, Wi-Fi and industrial chips.
No customer signs a long contract. What keeps money coming is that a chip, once designed into a product, usually stays for that product's life, two to seven years.
Growth and profit
What growth costs
Growth is making the company cheaper to run per sale: research fell from 63 cents of every sales dollar in 2024 to 38 cents over the last year. Stock paid to staff still costs about 16 cents per dollar, and factories cost little, because others make the chips.
Its record
The numbers
Is it growing?Fast
Sales grew 56% over the twelve months to Sep 2026, faster than the 30% in the year to Dec 2025.
Is it profitable?Losing
Operating margin is −0.9%, up from −5.5% in the year to Dec 2023.
Does growth pay off?Pays off
Running costs take 59¢ of each sales dollar, up from 48¢ in the year to Dec 2021.
Does it earn good returns?Poor
Shareholders' money lost 5.9% a year, down from 9.5% in the year to Dec 2021.
How long can it keep going?Long runway
It used $20.6M of cash over the last twelve months, so the $81.1M it holds lasts about 3.9 years at that pace.
Is the balance sheet safe?Strained
It owes $83.0M more than the cash it holds, and its operations bring in no cash to repay it.
Are shares being diluted?Diluting
The share count rose 23% in 4.8 years, and pay in shares costs 13% of sales.
Why it wins
The evidence
Once a module maker builds Keystone into a product and its cloud customer approves it, switching chips means redesigning and re-testing, so the chip tends to stay for years. That is MaxLinear's hold on its customers.
It is a hold on one generation. Each faster speed is a new design contest, so winning 800G links does not win 1.6T. And outside analysts do not name MaxLinear at all, so its share and its power advantage rest on management's word.
Who could beat it
The rivals in full
In optical chips MaxLinear fights Broadcom and Marvell, the incumbents, and Credo. The larger threat is a design that drops this chip altogether: NVIDIA already sells switches with the optics built in, and LightCounting, an industry research firm, expects such designs to slow sales of chips like Keystone from 2027. In broadband, Broadcom, Realtek and MediaTek sell the same kinds of chips.
cable gateway, fibre and Wi-Fi chips at operators and box makers
Realtek and MediaTek2379.TW2454.TW
Wi-Fi, Ethernet and fibre gateway chips, mostly in Asian-built equipment
How big it could get
The market and who has signed
The market
Cignal AI, an industry research firm, counted almost 20 million fast optical modules shipped in Q1 2026 alone, and most need a chip like Keystone. LightCounting puts 2026 sales of chips for the newest 1.6T modules above $2B.
Against that, MaxLinear expects $210M-$230M of optical sales this year, at what management's own figures suggest is about $20-$25 a chip. The market is large; MaxLinear's slice of it is not public.
Who has signed
Nobody has signed a long contract. MaxLinear has design wins at cloud companies in the US and Asia, through module makers, and about 28 weeks of orders ahead. Orders can be cancelled, and management says a handful of customers will drive most of the volume.
Who runs it
Leaders, pay and the cash
Who runs it
Kishore Seendripu co-founded MaxLinear in 2003 and has run it since. Steven Litchfield has been finance chief since 2018. In April the chief executive owned 6.0% and insiders 7.7% together. Long-term pay rides on sales and an adjusted profit measure, not on profit after stock pay or per share.
In August the chief executive sold about $50M of stock, about a ninth of his stake, outside any pre-arranged plan.
What they did with the money
In 2023 MaxLinear agreed to buy Silicon Motion, then walked away. Silicon Motion is still pursuing it in arbitration for a $160M fee plus damages, more than the company's cash. Since then it has paid down debt to $125M and bought back $20M of stock, but drew $40M on its credit line this year to pay for chips ahead.
The last report
The latest report
MaxLinear's Q3 2026 sales rose 72% to $218.1M and its optical forecast went up
What it did to the thesis Solid quarter
It met the test we set: sales above the guide's midpoint, the optical forecast raised and a profit after stock pay. It did not change what the price asks, which is sales far beyond the company's best year at a margin it has never earned.
The report
Sales for Q3 2026 were $218.1M, up 72% and inside the $210M-$220M guide, above its midpoint. Infrastructure, led by optical chips, rose 210% to $124.9M and is now 57% of sales, while broadband fell 8%. Counting the stock it pays staff, the company made a 13.9% operating profit; on its own adjusted measure the margin was 29.5%.
Management raised its 2026 optical forecast to $250M-$270M and guided Q4 sales to $240M-$260M. Cash from operations was negative $13.6M, because it paid chip makers ahead for 2027 orders.
What's happening now
The background
The 22 October results met the test we set. The next tests are the Q4 results in February and whether the first 1.6T chip sales arrive in the first half of 2027, as management now says. LightCounting wrote in September that optical market growth must be at its peak.
Where we disagree
Both sides in full
The market
An AI optics winner whose optical sales triple this year and keep compounding through 2028.
We believe
The optical wins are real but cover one chip generation, in a market near its fastest growth, beside a cyclical broadband half.
The argument
The market is paying for several more years of fast growth and a much fatter profit margin. The evidence covers one chip generation, and analysts expect the market's growth to slow from 2027.
We are wrong if quarterly sales reach $300M with a 25% operating margin after stock pay, and the 1.6T chip starts selling on time.
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.
The three cases
Bear case$4.64
Growth runs at half the analysts' path, 30%, 21% and 1%, the margin settles a fifth lower at 5.4%, and new capital earns only the 8% rate: the downturn test.
Our value$10.74
Revenue grows 60%, 43% and 2% over three years, the analysts' path (C41), held to at most 60% a year, then fades toward 4%, closing 30% of the gap each year; the margin moves to 6.8%, the analysts' year-three figure after stock pay and on the filed basis (15.8% less).
Bull case$15.12
Growth runs a quarter above the analysts' path, 75%, 53% and 2%, the margin reaches 7.5%, and new capital earns 15%.
The bet
Seen as a bet
This bet is on one business: Infrastructure, mostly the optical chips. Broadband, Wi-Fi and industrial are counted separately, at the modest price listed chip makers in those markets fetch for each dollar of sales.
If Infrastructure grows with its market from today's pace, valued like its optical rivals, the stock is worth only modestly more than today: an expensive bet. If optical stalls, shareholders keep mostly the slower half, a small part of today's price.
What could go wrong
Every risk
The risks
An arbitration bill
Silicon Motion seeks a $160M fee plus damages, more than MaxLinear's cash, in one undecided ruling.
Optics without the chip
Designs that drop the DSP could slow sales of Keystone-type chips from 2027.
The next generation slips
The 1.6T chip moved from late 2026 into 2027, and each speed is a new contest.
Broadband falls again
This half of the company fell by two-thirds from 2022 to 2024.
Shipped through Asia
Three-quarters of products ship to Hong Kong, China and Vietnam, exposed to trade rules.
The founder is selling
The chief executive sold about $50M of stock in August, outside a pre-arranged plan.
What we still do not know
MaxLinear's share of optical DSP chips, and what each sells for. Only a paid industry report says.
When the arbitration will be decided, and for how much. The proceedings are confidential.
How much of Infrastructure is optical. The company reports only the total.
What changes our view
Every route to a new view
Every route to a new view
→ Neutral
The share price falls to about $12.89, back within 20% of what we think the business is worth.
→ Attractive
The share price falls to about $8.59, 80% of what we think the business is worth.
→ Attractive
Quarterly sales of $300M or more. A quarter with an operating profit of 25 cents per sales dollar, after stock pay. The 1.6T chip selling in volume at two or more customers.
The next test
It supports the case if
Revenue at or above $250M, the optical range held or raised, a GAAP operating profit, and first-half 2027 Rushmore shipments reaffirmed.
It pressures the case if
Revenue below $240M, the optical range cut, a GAAP operating loss, or Rushmore's first shipments moved later than the first half of 2027.
We would also be wrong if
Optical sales for 2026 fall below $250M, the bottom of management's October range.
LightCounting cuts its 2027 growth forecast for optical modules below 30%.
The arbitration awards Silicon Motion more than MaxLinear's cash.
The verdict
Scores and how we weighed it
The scores
Growth
3 of 4
Business quality
3 of 4
Financial position
2 of 4
Execution record
3 of 4
Per share capture
2 of 4
Potential
2 of 4
Risk profile
2 of 4
Valuation
1 of 4
Evidence quality
2 of 4
How we weighed it
ValueNegative
The price sits far above our value and above the bull case.
BusinessMixed
A mixed business: a promising optical line beside cyclical broadband.
GrowthPositive
The optical forecast rose and sales beat the guide.
RisksNegative
An arbitration bill exceeds its cash, designs may leave its chip out and the next generation is unproven.
RivalsMixed
Cheaper than optical rivals, but half the company is slower broadband.
The groupMixed
Optical names move together on AI spending news.
Latest newsPositive
The last report met our test and turned a profit after stock pay.
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.