Lattice Semiconductor
A maker of the small programmable chip that switches a server on, watches over it and checks its software is genuine, and now owner of the code that boots it.
The scores
Growth3/4 ↑
Sales up 62% in Q2 2026, but the record is two quarters after a 31% fall in 2024 and narrows to servers.
Business quality3/4 →
Design-in and longevity lock the socket and a rival names it, but the buyers wrote Caliptra and OpenBMC and OCP's reference runs on Altera.
Financial position3/4 ↓
$925M of new floating term debt amortising to 2031, covered by cash flow on management's numbers, but the plan no longer funds itself.
Execution record3/4 ↑
Two 2026 quarters guided and beaten above the high end and channel inventory cut as promised, but seven quarters under this chief executive.
Per share capture2/4 ↓
Stock pay near a fifth of sales, a deal partly paid in shares, buybacks competing with debt repayment; the count stayed flat only via buybacks.
Potential2/4 →
The upside rests on hyperscaler capital budgets and a backlog cancellable on 60 days, nothing signed beyond a call; AMI cross-selling is unquantified.
Risk profile3/4 →
Two distributors carry 69% of sales and 67% ships to Greater China under live export-control risk factors, disclosed and survivable, with no dated binary event.
Valuation2/4 ↑
About 35 times adjusted 2027 earnings and 16 times guided sales annualised pays for the plan and AMI's margin; little is left if growth plateaus.
Evidence quality3/4 →
Central figures are filed and the outside sources reach the customers' own designs, but the backlog and every AMI number rest on management's word.
If it works 0.8× · if it fails you keep about 63%.
The Verdict
We rate Lattice neutral. It makes the small programmable chips that switch a server on and watch it; in July it bought AMI, whose code boots the server. AI data centres have pushed sales up fast, but stock pay equals about a fifth of sales and the purchase added debt, so the business is mixed. The price is full: buyers already count orders booked for 2027 as sales, so cancellations would hurt. As a bet it is priced for success: the chip business covers under two-thirds of the price, and AMI growing as promised adds little. The view hinges on that backlog and AMI's figures reaching the audited accounts.
The long caseEvery new server must boot and be managed remotely; if AI keeps adding servers and chip types, the main independent firmware house collects a fee on each.
The evidence under this - 8 claims
- If the panel's year-two figure of $4.24 (6 analysts) for the year ending January 2029 is right, the multiple falls to about 26 times two years out, which is where a 40% grower with a 40% adjusted operating margin is not obviously dear.
- The 126-session decomposition shows the multiple fell 3.8% while revenue per share rose 21.5%, so the de-rating has already begun and the price is paying less for each dollar of sales than in March.
- Stock pay is settled in shares rather than cash, and the company has offset the dilution with buybacks: diluted shares rose only 1.8% over the year while it spent $115M repurchasing stock, so the cost shows up in cash spent on buybacks rather than in the share count.
- Every listed semiconductor peer reports on the same non-GAAP basis and the panel's estimates are struck on it, so the multiple is comparable across the group even if the profit definition is generous.
- The fiscal 2025 10-K states that backlog has historically not been a predictor of future sales because purchase orders can generally be revised or cancelled up to 60 days before delivery without significant penalty and part of revenue is turns business.
- No backlog figure is filed anywhere, and the customer capacity agreements are described on a call and have not been filed as exhibits.
- No AMI financial statements have been filed: the 8-K/A with audited financials and pro formas was not on EDGAR at 19 September 2026, so every AMI figure is management's description of a private company it bought from a private-equity seller.
- The Q3 guide itself carries the first evidence against the margin story: combined non-GAAP gross margin of 69.5% is below the FPGA business's 70% because of the AMI hardware line.
- Management says 2027 is 'pretty much booked' and the panel expects far more than a plateau (C09, C34), so holding the FPGA business flat understates it if the backlog becomes revenue.
- Q2's margin excludes stock pay of about a fifth of revenue (C05), so on GAAP the same business earns far less.
The company
A server's big processors do the computing. Beside them sits a small chip that turns the power on in the right order, gathers readings from dozens of sensors and checks the firmware is genuine before anything runs. Lattice makes that chip, a programmable one whose wiring the customer sets in software, and supports many parts for more than 20 years.
Since July it also sells AMI's firmware, the code that starts a server and lets an operator run it remotely.
The evidence under this - 4 claims
- AMD's 10-K states the opposite trade from the other side: 'continued competition from the ASIC market, which has been ongoing since the inception of FPGAs', so rising ASIC cost is a rivalry the FPGA has never won outright.
- Caliptra shows the security function the fourth trend relies on moving inside the processor rather than onto a companion chip, and no outside body measures the far-edge AI socket the fifth trend describes.
How it makes money
It sells chips, almost all through distributors, two of which took 69% of 2025 revenue, and keeps about 70 cents of each sales dollar after the cost of making them. AMI adds licence fees and a royalty on each server its code runs.
The evidence under this - 4 claims
- No AMI financial statements have been filed: the 8-K/A with audited financials and pro formas was not on EDGAR at 19 September 2026, so every AMI figure is management's description of a private company it bought from a private-equity seller.
- The Q3 guide itself carries the first evidence against the margin story: combined non-GAAP gross margin of 69.5% is below the FPGA business's 70% because of the AMI hardware line.
Where the money comes from, stream by stream
| Stream | Share of sales (Q2 2026) | How it earns it |
|---|---|---|
| Compute and Communications | 62.6% | FPGAs for servers, storage, networking, client computers and wireless and wireline infrastructure, sold mostly through distributors. |
| Industrial and Embedded | 37.4% | FPGAs for factory automation, robotics, cars, industrial IoT, smart home and consumer devices, also sold mostly through distributors. |
| AMI firmware | not disclosed | Licensed server boot firmware (about 60%) and manageability software (about 40%); management expects over $200M of revenue in 2026. |
Who runs it
Ford Tamer has run Lattice since September 2024; before that he ran Inphi until Marvell bought it. Lorenzo Flores, who came from Intel, has run the finances since February 2025. Insiders own well under 1% of the company; large fund managers are the owners. The yearly bonus rides on sales, adjusted profit and personal goals, and most share awards on sales growth and the share price against the market, none on cash or official profit.
The evidence under this - 3 claims
Why it wins, and who could beat it
Why customers come back
Once the chip is designed onto a board, the board's power, sensor and security logic is written for it. Switching means a new layout, new code and re-testing the security path to save a few dollars, so buyers rarely switch within a product generation.
It shows in the margin: 70.3% in the June 2026 quarter, and 68.2% in 2025 after a year in which sales fell 30.9%. AMD names Lattice as one of its main rivals in these chips.
The evidence under this - 3 claims
The threat
The danger is the socket disappearing. AMD, Google, Microsoft and Nvidia wrote Caliptra, an open design that puts the security check inside their own chips. Axiado sells one chip that does the whole management module's job, Lattice's part included. And the big cloud firms back OpenBMC, free software that competes with AMI.
The evidence under this - 3 claims
- Caliptra's own scope statement is an SoC-internal root of trust for measurement; it does not sequence power, bridge buses or verify a BMC image from outside the SoC, which is the job the discrete control FPGA does.
- Lattice's Q1 2026 call frames its FPGAs as companions to CPUs, GPUs and BMCs rather than rivals, and AMI as silicon-neutral firmware validated across those same vendors.
Axiado is a private start-up and publishes no shipment or customer figures; three years after the announcement no OCP module in the marketplace listing found here names its device, and Lattice's server revenue grew 85% in 2025 and faster in 2026.
Management says AMI's firmware is the first brought up on a new CPU or GPU and that its licensees include hyperscalers, OEMs and ODMs, so an open stack has not displaced the licence at the platform-enablement stage; that is the seller's account and no licensee has confirmed it.
The rivals: AMD, and 3 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| Not listed | Caliptra, the open-source silicon root of trust specified by AMD, Google, Microsoft and Nvidia under the CHIPS AlliancePriced against it: Not listed | Lattice FPGAs | On sale | Not listed |
| Private | Axiado's SCM3000 DC-SCM 2.0 modules on its single-chip TCUPriced against it: Private | Lattice FPGAs | On sale | Private |
| AMD | Altera's MAX 10 and Agilex 3 and AMD's Spartan and Artix low-end FPGAsPriced against it: Different product | Lattice FPGAs | On sale | Different product |
| Not listed | OpenBMC, the Linux Foundation's open-source baseboard-management firmware backed by Facebook, Google, Intel, IBM and MicrosoftPriced against it: Not listed | AMI platform firmware | On sale | Not listed |
How big it could get
How big this could get
Management cites up to 20 million servers in 2026 with about three of its chips in each: roughly 60 million sockets a year. The dollar size cannot be worked out, because no chip price is published.
Server spending was $453.5B in 2025 and is forecast at $647.0B for 2026. Lattice's computing and communications sales were $292.7M, about 0.06% of it.
The evidence under this - 4 claims
- IDC measured server units up only 3.3% in Q1 2026 while Lattice's Compute and Communications revenue rose 86%, so most of the growth is attach and price rather than unit growth, which is the harder part to verify from outside.
- The server forecast moved from 16.5 million to 20 million between the May and August calls, and the serviceable-market figures are the company's own.
IDC's own page names the risk: the forecast assumes hyperscaler capex continues at announced levels and that memory supply stays constrained through the first half of 2027; a capex cut by any of the four largest buyers would take the rate down before it showed in server units.
Who has signed
Management says 2027 is "pretty much booked". But its own annual report says orders can be cancelled up to 60 days before delivery without real penalty, and that backlog has not historically predicted sales. No customer and no volume is named.
The evidence under this - 2 claims
- The fiscal 2025 10-K states that backlog has historically not been a predictor of future sales because purchase orders can generally be revised or cancelled up to 60 days before delivery without significant penalty and part of revenue is turns business.
- No backlog figure is filed anywhere, and the customer capacity agreements are described on a call and have not been filed as exhibits.
The results
The numbers
Sales grew 33% over the twelve months to Jul 2026, faster than the 2.7% in the year to Jan 2026.
Operating margin is 7.4%, down from 29% in the year to Dec 2023.
Running costs take 62¢ of each sales dollar, up from 43¢ in the year to Jan 2022.
Each dollar put into the business earns 5.3% a year after tax, down from 18% in the year to Jan 2022.
It turned 31% of sales into free cash, $199M over the last twelve months.
Yearly sales growth has ranged from −31% to +28% over the last 5 years, with sales falling in 1 of them.
The share count fell 1.3% in 4.5 years, and pay in shares costs 16% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in January.
| Jan 2022 | Dec 2022 | Dec 2023 | Dec 2024 | Jan 2026 | 12 mo to Jul 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $515M | $660M | $737M | $509M | $523M | $651M |
| Sales growth | +26% | +28% | +12% | −31% | +2.7% | +33% |
| Profit | ||||||
| Gross margin | 62% | 68% | 70% | 67% | 68% | 69% |
| Operating margin | 20% | 28% | 29% | 6.8% | 2.1% | 7.4% |
| Net margin | 19% | 27% | 35% | 12% | 0.6% | 5.6% |
| What growth costs, per sales dollar | ||||||
| Research and development | 21% | 21% | 22% | 31% | 36% | 34% |
| Selling, general and admin | 20% | 18% | 19% | 23% | 29% | 27% |
| Pay in shares | 9.0% | 8.4% | 9.5% | 10% | 22% | 16% |
| Capital spending | 1.9% | 3.5% | 2.7% | 4.1% | 8.1% | 6.8% |
| Cash | ||||||
| Cash from operations | $168M | $239M | $270M | $141M | $175M | $243M |
| Free cash flow | $158M | $215M | $249M | $120M | $133M | $199M |
| Balance sheet | ||||||
| Cash and short-term investments | $132M | $146M | $128M | $136M | $134M | $173M |
| Debt | $158M | $129M | $0 | - | - | - |
| Net cash | −$26.4M | $17.0M | $128M | - | - | - |
| Returns and shares | ||||||
| Return on capital | 18% | 31% | 32% | 4.9% | - | 5.3% |
| Shares (diluted) | 142M | 141M | 140M | 138M | 138M | 140M |
What growth costs
When sales grow, Lattice gets more profitable, because its research and selling costs grow more slowly than sales. The reverse also holds: in the 2024 slump costs kept rising and profit nearly vanished. Stock pay is a much bigger cost now than in the last upswing.
The evidence under this - 2 claims
Non-GAAP operating expenses are guided to $83M to $90M in Q3 2026 against $67.1M in Q2 (C08, C41), part of it AMI, so the cost base is stepping up again as the deal lands.
The last report (4 Aug)
Quarter ended July 2026, filed figures.
Sales grew 62.2% to $201.1M in the June 2026 quarter, above the top of the company's own range, with computing and communications chips up 83%. Management expects about 65% growth in the September quarter.
The market sold anyway: 7.0% the next day and 19.6% more over six weeks, because the outlook trimmed the gross margin and flagged tight assembly capacity and rising costs. On the official accounts it earned 14 cents a share; leaving out stock pay, 53.
What it did to the thesisIt supported the plan: sales beat the company's own range and the next quarter should grow faster still. It also showed the risks, a thinner margin, tight assembly capacity and orders customers can cancel. The price pays for the plan already, so the view stays Neutral.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.53 against $0.44 expected; $0.14 as filed, the gap being what the company adds back.
The evidence under this - 5 claims
- The same call attributes the gross-margin step-down to a low-margin hardware pass-through business inside AMI and to industry-wide assembly constraints and expedite fees that will be partly absorbed.
- The company's own filing says purchase orders can be cancelled up to 60 days before delivery, so a guide built on bookings is not a guide built on firm orders.
- Both falls happened after large rallies (from $61 to $73 before November, from $95 to $155 before August), and the 252-session decomposition shows the multiple still 26.6% higher than a year ago, so the sells read as positioning at least as much as as a verdict on the print.
- Every listed FPGA and networking-silicon name is not checked here for the same August pattern, so a sector-wide de-rating cannot be ruled out as the cause.
What management did with the money
It borrowed $925M to pay for AMI and issued about 5.2 million shares to AMI's seller. Before the deal it had $173.3M of cash and no debt; it plans to bring debt below two years of cash profit by the end of 2027.
Insiders in 2026Q2: 17 purchases and 35 sales, from their Form 4 filings.
The evidence under this - 4 claims
The plan needs the AMI EBITDA and the FPGA growth to arrive; on Q2 2026 alone, annualised adjusted EBITDA of about $346M against $925M of debt is 2.7 times before any AMI contribution.
The price, and where we disagree
In short. Full. The price pays for Lattice's plan: orders booked for 2027 turning into sales, and AMI, the firmware business it bought, earning what management promises. Little is left if growth stalls.
The shares are 29.3% below their high while filed growth is accelerating. This market has twice sold Lattice after good results, on what they implied for margins and supply rather than for sales.
The next test is on 2 November: the first quarter with AMI inside, and the first with the chips and AMI reported separately.
The evidence under this - 3 claims
- Both falls happened after large rallies (from $61 to $73 before November, from $95 to $155 before August), and the 252-session decomposition shows the multiple still 26.6% higher than a year ago, so the sells read as positioning at least as much as as a verdict on the print.
- Every listed FPGA and networking-silicon name is not checked here for the same August pattern, so a sector-wide de-rating cannot be ruled out as the cause.
How we read the multiple
On the official accounts, which count stock pay to staff as a cost, Lattice earned 2 cents a share in 2025, so on that measure the price per dollar of profit is very high.
On the adjusted profit analysts forecast for the year to January 2028, which leaves stock pay out, buyers pay about 35 times. The whole company costs about 16 times this quarter's expected sales, taken over a full year.
Against its own history. Far dearer than its own normal, most of all on profit, which is still recovering from the slump; on sales the gap is narrower.
The evidence under this - 5 claims
- If the panel's year-two figure of $4.24 (6 analysts) for the year ending January 2029 is right, the multiple falls to about 26 times two years out, which is where a 40% grower with a 40% adjusted operating margin is not obviously dear.
- The 126-session decomposition shows the multiple fell 3.8% while revenue per share rose 21.5%, so the de-rating has already begun and the price is paying less for each dollar of sales than in March.
- Stock pay is settled in shares rather than cash, and the company has offset the dilution with buybacks: diluted shares rose only 1.8% over the year while it spent $115M repurchasing stock, so the cost shows up in cash spent on buybacks rather than in the share count.
- Every listed semiconductor peer reports on the same non-GAAP basis and the panel's estimates are struck on it, so the multiple is comparable across the group even if the profit definition is generous.
- The fiscal 2025 10-K states that backlog has historically not been a predictor of future sales because purchase orders can generally be revised or cancelled up to 60 days before delivery without significant penalty and part of revenue is turns business.
- No backlog figure is filed anywhere, and the customer capacity agreements are described on a call and have not been filed as exhibits.
- No AMI financial statements have been filed: the 8-K/A with audited financials and pro formas was not on EDGAR at 19 September 2026, so every AMI figure is management's description of a private company it bought from a private-equity seller.
- The Q3 guide itself carries the first evidence against the margin story: combined non-GAAP gross margin of 69.5% is below the FPGA business's 70% because of the AMI hardware line.
Priced against its rivals
None of the listed rivals above sells the same thing to the same buyers as a large part of what it does. A price comparison needs at least two, so the price is judged against its own history instead.
Also checked and set aside: Microchip Technology Incorporated (a substitute, not the same product).
Private or foreign rivals that cannot be priced: Altera Corporation, Efinix, Inc., Gowin Semiconductor Corp., Insyde Software Corp., Phoenix Technologies Ltd.
The bet: what it pays if it works
The bet is on AMI alone: the code that boots a server and lets it be run remotely. Its sales and margins are still management's figures, in no filing yet.
The chip business, held at its current pace and valued at the price buyers pay today, covers under two-thirds of the price. Even if AMI grows as management promises, the whole company would be worth less than today: an expensive bet.
If AMI earns nothing, shareholders keep the chip business and the loan that bought AMI.
| AMI platform firmware earnings in two to three years | Million dollars of AMI revenue a year needed | At the stated multiple |
|---|---|---|
| $76.00M | about 200 at 100% kept | 0.8× |
| $95.00M (the big win) | about 250 at 100% kept | 0.8× |
| $133.00M | about 350 at 100% kept | 0.9× |
To pay 2× (a Fair bet) it would need about $595.71M of AMI platform firmware earnings (about 1.6K million dollars of AMI revenue a year); to pay 5× (a Cheap bet), about $1.90B of AMI platform firmware earnings (about 5.0K million dollars of AMI revenue a year), at the same multiple.
Each row is AMI platform firmware alone, plus the rest of the company (Lattice FPGAs) counted at about $9.55B.
The steps it has to take
- Done AMI bought and financed: $1B in cash and 5.2 million shares
- In progress The low-margin hardware pass-through line exited due 2026-12
- Not started AMI's revenue and gross margin reported separately in a Lattice filing due 2026-11
- Not started AMI revenue about $250M in 2027, about 25% growth due 2028-02
- Not started Leverage below two times EBITDA due 2027-12
The detail behind the label
The evidence under this - 7 claims
- No AMI financial statements have been filed: the 8-K/A with audited financials and pro formas was not on EDGAR at 19 September 2026, so every AMI figure is management's description of a private company it bought from a private-equity seller.
- The Q3 guide itself carries the first evidence against the margin story: combined non-GAAP gross margin of 69.5% is below the FPGA business's 70% because of the AMI hardware line.
- None of AMI's figures has yet appeared in a Lattice filing: the 10-Q for the quarter that includes it is the first, and management says it will only then report AMI's revenue and gross margin separately (C50).
- The partial quarter still carries a low-margin hardware pass-through line being exited by the end of 2026 (C13), so the first reported margin will not be the go-forward one.
- Management says 2027 is 'pretty much booked' and the panel expects far more than a plateau (C09, C34), so holding the FPGA business flat understates it if the backlog becomes revenue.
- Q2's margin excludes stock pay of about a fifth of revenue (C05), so on GAAP the same business earns far less.
- If the panel's year-two figure of $4.24 (6 analysts) for the year ending January 2029 is right, the multiple falls to about 26 times two years out, which is where a 40% grower with a 40% adjusted operating margin is not obviously dear.
- The 126-session decomposition shows the multiple fell 3.8% while revenue per share rose 21.5%, so the de-rating has already begun and the price is paying less for each dollar of sales than in March.
Where we disagree with the market
An AI-server winner whose backlog covers 2027, so the fall since August is a chance to buy.
The growth is real and filed, but this is a fair price for the plan, and the plan carries $925M of debt and cancellable orders.
Analysts have been over 92% positive for a year. We read what the call does not stress: where the chips ship, how easily orders cancel, what stock pay costs, and customers writing their own replacements.
The case against us: if chip sales keep compounding above 40% into 2027 with the margin holding near 70%, and AMI prints above $200M, the market is right.
What analysts say, and the evidence
The evidence under this - 6 claims
- Caliptra's own scope statement is an SoC-internal root of trust for measurement; it does not sequence power, bridge buses or verify a BMC image from outside the SoC, which is the job the discrete control FPGA does.
- Lattice's Q1 2026 call frames its FPGAs as companions to CPUs, GPUs and BMCs rather than rivals, and AMI as silicon-neutral firmware validated across those same vendors.
Management says AMI's firmware is the first brought up on a new CPU or GPU and that its licensees include hyperscalers, OEMs and ODMs, so an open stack has not displaced the licence at the platform-enablement stage; that is the seller's account and no licensee has confirmed it.
What could go wrong, and how we would know
The risks
- Orders that can vanish. Backlog can be cancelled up to 60 days before delivery, and the company says it has not predicted sales.
- Customers designing it out. Caliptra and single-chip management modules move the security and control jobs off a separate chip.
- Debt from the AMI deal. Interest on the $925M loan could take about 17% to 20% of recent yearly profit before stock pay.
- China and two distributors. 67% of June-quarter sales shipped to Greater China, and two distributors carry most of the revenue.
- A cyclical past. Sales fell 30.9% in 2024, and the company accepts its markets can fall that far again.
The evidence under this - 8 claims
- Caliptra's own scope statement is an SoC-internal root of trust for measurement; it does not sequence power, bridge buses or verify a BMC image from outside the SoC, which is the job the discrete control FPGA does.
- Lattice's Q1 2026 call frames its FPGAs as companions to CPUs, GPUs and BMCs rather than rivals, and AMI as silicon-neutral firmware validated across those same vendors.
Axiado is a private start-up and publishes no shipment or customer figures; three years after the announcement no OCP module in the marketplace listing found here names its device, and Lattice's server revenue grew 85% in 2025 and faster in 2026.
Management says the deal is accretive to non-GAAP EPS from Q4 2026 after interest, and the loan amortises and can be prepaid without penalty, so the drag shrinks with each quarter of the described free cash flow.
What we still do not know
- What AMI actually earns. Its revenue and margins are management's figures, not yet in audited accounts.
- How many servers carry a Lattice chip. Nobody outside the company counts the sockets.
- Who the end customers are. Sales go through distributors to addresses that may be assembly plants.
The evidence under this - 4 claims
- No AMI financial statements have been filed: the 8-K/A with audited financials and pro formas was not on EDGAR at 19 September 2026, so every AMI figure is management's description of a private company it bought from a private-equity seller.
- The Q3 guide itself carries the first evidence against the margin story: combined non-GAAP gross margin of 69.5% is below the FPGA business's 70% because of the AMI hardware line.
- IDC measured server units up only 3.3% in Q1 2026 while Lattice's Compute and Communications revenue rose 86%, so most of the growth is attach and price rather than unit growth, which is the harder part to verify from outside.
- The server forecast moved from 16.5 million to 20 million between the May and August calls, and the serviceable-market figures are the company's own.
What would change our view
- Neutral→AttractiveThe backlog turns into filed revenue and AMI is audited
It needs all of:
- ○Chip sales grow more than 25% from a year earlier for four quarters running, turning "2027 is booked" into reported numbers.checked at the next results, due by 2 Nov
- ○AMI reports over $200M of 2027 revenue, at a gross margin above 70% before stock pay, inside the audited accounts.checked at the next results, due by 2 Nov
- Neutral→AttractiveThe deal deleverages on its own cash
It needs all of:
- ○Debt net of cash falls below two years of cash profit by the end of 2027, without selling new shares.checked at the next results, due by 2 Nov
- ○The operating margin before stock pay stays at 30% or more while supply costs are absorbed, from 38.3% in June.checked at the next results, due by 2 Nov
We would also be wrong if
- A cloud company or open server design that is shipping drops the separate control chip for a function built into another chip.
What is coming
- October 2026 First quarterly release of the AMI seller's lock-up: Whether THL sells its first quarter of the roughly 5.2 million shares as they free; the release dates were stated only on a call.
Every dated event ahead
- October 2026 First quarterly release of the AMI seller's lock-up: Whether THL sells its first quarter of the roughly 5.2 million shares as they free; the release dates were stated only on a call.
- 2 November 2026 Q3 2026 results: FPGA revenue against the $210M to $230M guide, gross margin against 69.5%, AMI's first two months reported separately, and whether '2027 is pretty much booked' survives.
- February 2027 Q4 2026 results and the FY2026 10-K: The first full quarter with AMI, the 2027 outlook against the $1.2B exit run rate, and debt repaid against the below-two-times target.
Since we wrote this
- 5 OctThesis reviewThe long case now sits under the verdictAfter the five-name pilot, the owner asked for the long case on every thesis.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 19 SepThesis writtenNeutral, medium convictionFiled growth of 62.2% and a fresh firmware deal, at a price that already pays for 2027 on a backlog stated only on a call.
Also checked 5 times with nothing new to act on.
Sources 27
Sources - 27
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-13Item 1: product families, platforms and nodes, foundries and OSATs, the distributor model and the two 30%-plus distributors, the 20-year longevity statement, backlog and the 60-day cancellation term, the competition wording; the risk factors on export controls and Taiwan; the income statement, cash flow (stock-based compensation $115.6M, buybacks $100.0M), and the customer and geography tables.backsC01C02C05C07C10C16C18C29C39C42C48C49C52C57C58
- S02primarycompany filing2026-08-04The debt note with the 2026 Credit Agreement terms and the $925M draw; the subsequent-events note on the AMI closing; channel and geography tables (95% distributors, 67% Greater China); the receivables concentration; the balance sheet and cash flow; the repurchase note and remaining authorisation; the end-market note; the 'no material changes' risk-factor statement.backsC03C04C05C06C07C11C12C15C16C32C33C42C50C57C58
- S03primarycompany filing2026-08-04GAAP and non-GAAP tables, the reconciliation (stock-based compensation by line), the Q3 2026 guide including AMI, the AMI financial highlights, the end-market dollar split and the supplemental balance-sheet metrics.backsC03C04C05C08C13C15C16C31C41
- S04primarytranscript2026-08-04Lattice Semiconductor Q2 2026 earnings call transcriptPrepared remarks: backlog into 2027, AMI's revenue mix and margins, the hardware pass-through, the financing structure and the below-2-times leverage target. Questions from eight analysts, a separate origin: supply constraints, cost pass-through, the margin step-down, '2027 pretty much booked', the 20 million server figure, the Rule of 105, the $3B 2030 aspiration and segment reporting from Q3.
This is evidenced by accelerated bookings and our backlog extending well into 2027. The latest Digitimes report shows the server TAM is now up to 20 million units forecast for 2026. The visibility is increasing daily. I mean it's really unprecedented. We've got visibility all the way to the end of 2027. 2027 is pretty much booked. As we exit 2026, we expect the business to be operating at a revenue run rate of more than $200 million with gross margins above 75% and EBITDA margins above 40%.
backsC08C09C13C14C28C30C41C43C44C50C59 - S05primarycompany filing2026-05-04Q1 2026 results, the Q2 2026 guide of $175M to $195M and $0.42 to $0.46, and the first disclosure of the new end-market groups.backsC31C49
- S06primarycompany filing2026-05-04Deal terms: $1.65B, $1.0B cash and about $650M of stock, 5.2 to 6.1 million shares with $57.3M of awards to AMI employees, AMI over $200M of 2026 revenue, THL as seller, Wells Fargo and Morgan Stanley committed financing.backsC11
- S07primarytranscript2026-05-04Lattice Semiconductor Q1 2026 earnings call transcriptThe serviceable market from $6B to $12B, server at 38% of Q1 revenue, AI at about 25% of 2026 revenue, 16.5 million servers and three FPGAs per server, channel inventory from six months to two, the THL lock-up of 12 months at 25% a quarter, and the analyst's statement that Altera is now a standalone company.
We expect our total serviceable available market to double from about $6 billion currently to about $12 billion jointly with AMI over the next three to four years. From a modeling point of view, we still have a forecast of a 16.5 million server market in 2026. We are roughly saying about three FPGAs per server overall. We gave you today a breakdown of that server business in Q1-about 38% of our revenue. They have a lockup that extends for 12 months from close-25% per quarter.
backsC09C14C28C38C43 - S08primarycompany filing2026-02-10Fiscal 2025 results, non-GAAP EPS of $1.05, the Q1 2026 guide of $158M to $172M, server revenue up about 85%, new-product revenue up about 70%, and the end-market dollar table including Communications and Computing at $292.7M.backsC31C32C33C39C49C53
- S09primarycompany filing2026-06-30The $950M delayed-draw term loan and $200M revolver, SOFR plus 1.00% to 1.75% by leverage ratio, the ticking fee, the 9 November 2026 commitment expiry, and the June 2031 maturities.backsC12
- S10primarycompany filing2026-07-27The closing date, AMI as a business unit under its own chief executive reporting to Lattice's, and the silicon-neutrality commitment.backsC11
- S11primarycompany filing2026-02-04A rival's naming of Lattice among its primary FPGA competitors; the ASIC and ASSP framing; the Spartan, Artix, Kintex and Virtex families; Embedded revenue of $3.5B and operating income of $1.2B.backsC17
- S12primaryindustry data2026-09-19Caliptra 2.0 specification - CHIPS Alliance CHIPS Alliance (AMD, Google, Microsoft, Nvidia)The customers' own root-of-trust design, its SoC-internal scope and its explicit exclusions.
The objective of Caliptra is to define core RoT capabilities that must be implemented in the System on Chip (SoC) or ASIC of any device in a cloud platform. The silicon iRoT scope includes all datacenter-focused server class SoC / ASIC (datacenter focused) devices (SSD - DC, NIC, CPU, GPU - DC). Contributions to this Specification are made under the terms and conditions set forth in Modified Open Web Foundation Agreement 0.9 (OWFa 0.9) by: AMD, Google, Microsoft, Nvidia. Explicitly out of scope is how silicon integration into backend work is performed
backsC19 - S13primaryindustry data2026-09-19Linux Foundation - OpenBMC Project community comes together at The Linux Foundation The Linux FoundationWho built and supports the open BMC firmware stack that competes with AMI's.
IBM is providing their OpenBMC code base to The Linux Foundation, and this project will be supported by several organizations, including Facebook, Google, Intel, and Microsoft.
backsC20 - S14primaryindustry data2026-09-19Open Compute Project - HWMgmt-Module-DCSCM-LTPI reference implementation, README Open Compute ProjectThe standards body's reference design for the module interface, and the rival device it targets; the January 2026 revision date.
This repository contains: Verilog Source Code of standalone LTPI IP (SCM + HPM); Intel Quartus Example Project File for Intel MAX 10 CPLD; Complete set of LTPI Unit Tests Implementation Package. LTPI IP that is compliant with the DC-SCM 2.x LTPI Specification. | 1.20 | 13 January, 2026 | Fix corner case bug in data channel managment for target and controller device | Implementation compliant with 1.2 LTPI Specification.
backsC21 - S15primarycompany release2023-10-10A competitor's own description of a single-chip module that removes the discrete programmable-logic device.
SAN JOSE, Calif., October 10, 2023 - 2023 OCP Global Summit - Axiado Corporation, an AI-enhanced hardware cybersecurity company, today announced the release of its SCM3000 family of modules compliant to the Open Compute Project's (OCP) data center secure control module (DC-SCM) Rev2.0 Ver 1.0 base specification. The DC-SCM Rev2.0 modules integrate Axiado's innovative TCU with AI/ML capabilities, which includes critical functions such as baseboard management controller (BMC), platform root of trust (PRoT), trusted platform module (TPM) and complex programmable logic device (CPLD), creating in effect a single-chip-based DC-SCM solution.
backsC22 - S16primarycompany release2026-09-16The response product: density, SERDES, the algorithm list, configuration time and availability.
HILLSBORO, Ore. - Sep. 16, 2026 - Lattice Semiconductor (NASDAQ: LSCC), the low power programmable and platform firmware leader, today announced Lattice Mach-N2 FPGA family. Built on the Lattice Nexus 2 small FPGA platform, Mach-N2 combines integrated flash, a hardware Root of Trust, and CNSA 2.0-compliant post-quantum cryptography (PQC) with crypto agility for trusted, always-on system control and security across next-generation systems. The new Lattice Mach-N2 devices are available for order today and samples have already shipped to industry-leading Compute and Communications customers. Up to 2X greater logic density for advanced system control functions. CNSA 2.0-compliant PQC: ML-DSA, LMS, and XMSS for authentication; ML-KEM for key exchange. Fastest-in-class instant-on configuration: a 220k system logic cell device fully configures in under 30 milliseconds. High-speed connectivity PCIe 4.0 and 10G Ethernet. Up to 16 Gbps SERDES.
backsC23 - S17primarycompany filing2026-07-30A hyperscaler's filed capital spending: $31.08B in the quarter and $130B to $145B anticipated for 2026.backsC25
- S18primaryindustry data2026-07-28IDC - Servers Market Insights (Worldwide Quarterly Server Tracker summary page) International Data CorporationThe measured rate and the forecast: 2025 $453,531M, 2026 $646,998M, 2027 $930,562M in the page's table. A commercial tracker with a stated method, not a statistical agency; its Q2 2026 press release returns 403.
The worldwide server market delivered 30.7% spending growth in the first quarter of 2026 while unit growth was modest at 3.3% year over year. Leading vendors confirm order pipelines are strong and that supply, not demand, represents the primary ceiling for growth in the current environment. Thus, the market is expected to continue in a very positive trajectory of 25.1% compound average growth through 2030. Total: 2025 $453,531 78.2%; 2026 $646,998 42.7%; 2027 $930,562 43.8% (Value US$M, Annual Growth %). Updated: July 28, 2026.
backsC26C27 - S19primarygovernment data2024-08-13NIST - FIPS 203, Module-Lattice-Based Key-Encapsulation Mechanism Standard National Institute of Standards and TechnologyThe post-quantum standard the security parts implement, from the body that wrote it.
This standard specifies a key-encapsulation mechanism called ML-KEM. The security of ML-KEM is related to the computational difficulty of the Module Learning with Errors problem. At present, ML-KEM is believed to be secure, even against adversaries who possess a quantum computer.
backsC24 - S20secondaryvendor data2026-09-19Research pack, frozen at content/LSCC/research-pack.jsonPrice, returns and the 252-session decomposition; the trailing multiples; the analyst snapshot, dated series and estimates; FINRA short interest; the insider window and base rate; the earnings history with filed GAAP EPS beside the adjusted figure; headcount from the 10-K; the filed-against-vendor cross-check.backsC29C31C32C33C35C36C37
- S21primarycompany filing2026-09-19Which open-market sales were under plans and which were discretionary; the chief executive's four November 2025 purchases; the three director sales.backsC37
- S22primarycompany filing2026-05-21The FPGA product description inside an embedded-control portfolio, and the absence of any named FPGA competitor.backsC47
- S23tertiarynews2026-08-05The objection coverage stated after each of the two falls: the margin guide, supply constraints and cost pass-through in August 2026; profit-taking after a run-up in November 2025. Also the two broker target increases. Used only for what people said, never for a figure.backsC35
- S24secondaryvendor data2026-09-18FMP end-of-day price history, LSCC, September 2025 to September 2026The closes around each print and the deal closing, the June high and the November low; observed trades, split-adjusted by the vendor.backsC32
- S25primarycompany filing2026-07-27The plan amended at closing to cover inducement grants to AMI employees; read to confirm it is not the credit agreement, which is not filed as an exhibit to that 8-K. No figure taken from it.read, but no claim rests on it
- S26primarycompany filing2026-03-18beneficial ownership at the 2 March 2026 record date, the executive officers and their start dates and records, the CD&A: the 2025 Corporate Incentive Plan's weights and payout, the TSR and revenue-growth PRSU designbacksC54C55C56
- S27primarycompany filing2024-02-16the FY2021 to FY2023 income statement lines as a share of revenue (R&D, selling, general and administrative), read through the frozen numbers card (content/LSCC/financials.json)backsC57C58