Vicor
Designs and US-manufactures (Andover, MA) proprietary modular power-conversion components (Factorized Power, Power-on-Package / vertical power delivery, 48V-to-point-of-load) sold to AI-compute, automotive/EV, aerospace/defense and industrial OEMs; increasingly monetizes a large patent portfolio via all-inclusive OEM licenses and per-unit royalties, enforced through the US ITC.
Earnings, margins, COGS & capex
Vicor inflected in 2025 from a multi-year margin trough into a high-growth, high-margin phase: FY2025 net income $118.6M / $2.61 diluted EPS (aided by a $45M patent settlement), gross margin up 610bps to 57.3%, and operating cash flow up to $139.5M from $50.8M (+174.5%). Q1 2026 accelerated (revenue +20.2% YoY, backlog +70% QoQ to $300.6M, book-to-bill >2), and management raised Q2 2026 guidance on May 26, 2026 to $142M from $126M on stronger product demand plus a newly signed all-inclusive OEM royalty license. The balance sheet is net-cash with minimal debt.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~43¢ is cost of goods and ~39¢ operating expense, leaving ~18¢ of operating profit (~26¢ net).
Revenue trend
Margins
up sharply YoY (from 47.2% year-ago quarter); royalty mix + utilization driven
inflected positive from near-breakeven 2024
up, but settlement-inflated
up sharply in absolute FCF YoY, boosted by settlement + working capital
COGS structure
COGS is dominated by in-house US wafer/module fabrication (Andover, MA) — proprietary MOSFET/magnetics-based power modules; heavily fixed-cost, so gross margin is highly operating-leverage-sensitive to volume and to the royalty mix (royalty revenue carries near-100% incremental margin). Margin recovery in 2025-26 reflects both higher factory utilization and a rising royalty share.
Capex
Modest, ~$20M/yr (FY2025); Q1 2026 capex was $12.4M — Andover (Fab One) capacity expansion largely complete, with management outlining a path to expand Fab One to ~$1.5B annual revenue and options for a second fab. Growth from AI/EV demand is capital-light, and the licensing model requires essentially no incremental capex.
Latest earnings
Beat — revenue $113.0M topped expectations and EPS was $0.44; stock +~145% YTD into the print; management raised Q2 guidance to $142M on May 26, 2026
Q2 2026 ~$142M (raised May 26 from $126M); FY2026 ~$570M, deliberately excluding potential new license deals until ITC litigation concludes (conservative licensing assumption)
- Q1 2026 revenue
- $113.0M (+20.2% YoY)
- Q1 2026 gross margin
- 55.2%
- Q1 2026 net income / EPS
- $20.7M / $0.44
- One-year backlog
- $300.6M (+70% QoQ, +75% YoY)
- Book-to-bill
- >2.0
- FY2025 diluted EPS
- $2.61
- Cash & investments
- ~$402.8M (2025-12-31), no material debt
Growth drivers
- 48V direct-to-GPU and vertical/lateral power delivery (Power-on-Package) for AI accelerators — the AI 'last-inch' power-delivery problem as GPU currents scale toward Blackwell/Rubin
- High-margin IP licensing + per-unit royalties, structurally expanding as ITC import bans push ODMs/OEMs to take licenses (new all-inclusive OEM license signed 2026 covering Factorized Power + Vertical Power Delivery; a second ITC investigation, 337-TA-1484, instituted Feb 2026)
- AI-data-center power demand ('AI power crisis') driving both module volume and licensing leverage
- Automotive/EV electrification (high-density DC-DC, 800V/48V architectures)
- Aerospace & defense (rad-tolerant, high-reliability modular power)
- Operating leverage on a largely fixed US manufacturing base — incremental volume drops through at high margin
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-03-02. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
Vicor is a small-cap picks-and-shovels winner of the AI power build-out with two compounding engines — differentiated high-density modules for GPU power delivery, and a newly credible, ITC-enforced licensing/royalty annuity that carries near-100% incremental margin — on a net-cash balance sheet, with backlog +70% QoQ and book-to-bill >2 signaling durable demand.
- AI accelerators are hitting a physics wall on power delivery at the die; Vicor's vertical/lateral power delivery and 48V-to-point-of-load directly address it as currents scale toward Blackwell/Rubin
- ITC exclusion orders (337-TA-1370) are already enforced by US Customs, and a second investigation (337-TA-1484) was instituted Feb 2026 — pressuring ODMs to take licenses and converting IP into a recurring royalty stream (a new all-inclusive OEM license drove the May 2026 Q2 raise)
- Financial inflection: gross margin +610bps to 57.3%, operating income to $81.8M, operating cash flow +174.5% YoY to $139.5M, and guidance raised (Q2 to $142M from $126M, FY2026 ~$570M)
- Backlog $300.6M (+70% QoQ, +75% YoY) and book-to-bill >2 point to visibility, not a one-quarter spike
- Capital-light: Andover (Fab One) capacity is largely built, so incremental AI/EV growth flows through at high margin with minimal capex
- FY2026 guidance is deliberately conservative — it excludes potential new license deals until litigation concludes, leaving upside optionality
Much of the story is priced in at ~94x trailing earnings / ~28x sales, while the two engines carry real fragility: Vicor already lost the flagship H100 GPU VRM socket to Monolithic Power, and the high-margin licensing annuity is litigation-contingent and lumpy — a flattering $45M settlement inflated FY2025, and an adverse ITC/court turn or a customer architecture shift would hit both the numbers and the multiple.
- Competitive displacement is not hypothetical — MPWR won the H100 VRM design (reportedly ~100% of H100) and is taking AI-power share with cheaper integrated multiphase solutions; TXN and ADI add scale-driven pricing pressure
- Licensing/royalty revenue depends on continued ITC/court wins and on infringers choosing to license rather than design around — inherently lumpy and reversible (a published short thesis frames AI-chip roadmaps as accelerating Vicor's competitive risk)
- FY2025 profitability was materially aided by a one-time $45M patent settlement; underlying product-only economics are less spectacular
- Small scale (~$450M revenue) versus multi-billion-dollar power incumbents means Vicor can be out-invested and out-priced in any socket it doesn't uniquely own
- Valuation embeds years of flawless licensing + AI share execution; any air-pocket in AI capex, a lost socket, or a delayed ITC ruling could compress a ~90x+ multiple hard
- Fixed-cost US manufacturing amplifies downside operating leverage if AI-module volumes disappoint
What it is worth
Market multiples cross-checked against FY2025 actuals and FY2026 guidance; qualitative sum-of-parts (hardware module business + emerging royalty annuity).
Competitive socket losses (MPWR/TXN) cap product growth and an adverse/delayed ITC outcome stalls the licensing annuity; underlying (ex-settlement) economics can't justify ~90x+ trailing earnings, and the multiple compresses materially even if revenue merely grows in-line.
AI product demand + a growing-but-lumpy royalty stream drive ~$550-600M FY2026 revenue with mid-50s% gross margin; the stock is priced for much of this, so returns track execution against already-high expectations rather than multiple expansion.
If AI-power sockets hold and the ITC-enforced licensing annuity scales into a recurring high-margin stream, revenue compounds toward and beyond the ~$570M FY2026 guide with 55%+ gross margins and rising operating leverage — supporting a premium multiple and further upside from today's level; licensing optionality is explicitly excluded from guidance.
At ~$12.9B market cap on ~$450M FY2025 revenue, VICR trades at ~94x trailing TTM P/E and ~28x trailing sales (~22x FY2026 guided ~$570M) — a growth/optionality multiple that capitalizes both AI-power share gains and a durable licensing annuity. (On FY2025 diluted EPS of $2.61 alone the multiple is ~112x; TTM EPS is higher after the Q1 2026 step-up.) Net cash (~$400M) modestly reduces EV. The multiple is highly sensitive to whether royalty revenue proves recurring (ITC-driven) versus lumpy (settlement-driven).
SWOT
Strengths
- Differentiated, patented power-density leadership (Factorized Power, Power-on-Package / vertical power delivery) suited to the hardest AI power-delivery problem
- Net-cash balance sheet (~$400M, no material debt) funding growth without dilution/leverage
- Rapidly rising high-margin royalty stream with near-100% incremental margin
- US-based in-house manufacturing — supply-chain control and defense-grade qualification
- Enforceable IP moat validated by ITC exclusion orders and import bans
Weaknesses
- Small absolute scale (~$450M revenue) versus diversified analog/power giants (TXN, ADI, MPWR, Infineon)
- Lost the high-volume H100 GPU VRM socket to Monolithic Power — a demonstrated design-win loss in the exact AI market it touts
- History of lumpy revenue, margin volatility, and multi-year execution/capacity missteps
- Royalty/licensing revenue is litigation-dependent and lumpy (a $45M settlement flattered FY2025)
- Fixed-cost manufacturing base makes margins sensitive to any volume air-pocket
Opportunities
- AI compute power demand scaling faster than incumbent multiphase VRM architectures can efficiently serve at the die
- Licensing flywheel — convert infringers into royalty-paying licensees across the ODM/OEM base (Blackwell/Rubin systems)
- Vertical power delivery becoming a standard requirement as GPU currents exceed ~1000A
- EV/800V automotive and defense diversification beyond compute
- Potential for royalty revenue to become a large, recurring, high-margin annuity
Threats
- Monolithic Power (MPWR), Texas Instruments (TXN), Analog Devices (ADI), Infineon, Renesas and Delta all targeting AI 48V power with scale and lower-cost integrated solutions
- TI's 300mm/scale economics pressuring module pricing and gross margin
- Customer/design-socket concentration around the NVIDIA compute ecosystem (accessed via ODMs) — a single architecture shift can displace Vicor
- Adverse or delayed ITC/court rulings would undercut the licensing thesis embedded in the valuation
- Rich valuation (~94x trailing TTM P/E, ~28x sales) leaves little room for execution error
Moats, dependencies & bottlenecks
Moats
vertical power delivery, converter topologies/control) enforced via ITC import bans (337-TA-1370; 337-TA-1484 instituted Feb 2026) Strong but contested patents expire and can be designed around; enforcement is active and litigation-dependent ITC exclusion orders give real teeth today (US Customs enforcement, ODMs pushed toward licenses), but the moat is legal/temporal, not permanent.
Power-density / architecture technology leadership for the hardest die-level AI power-delivery problem Strong in its niche real engineering lead, but well-capitalized incumbents (MPWR/TXN/ADI/Infineon) are targeting the same sockets Leads 48V direct-to-GPU / vertical power delivery; has also lost high-volume sockets (H100 VRM), so leadership is socket-by-socket, not absolute.
High for defense/aero, lower for cost-sensitive compute Supply-chain control and qualification barriers, but fixed-cost and sub-scale versus 300mm analog fabs.
Dependencies
accessed via ODMs) Both module volume and licensing leverage hinge on continued hyperscale AI accelerator build-out and on Vicor content in those systems.
Legal / regulatory The high-margin licensing/royalty thesis embedded in the valuation depends on continued exclusion orders (337-TA-1370) and the new 337-TA-1484 investigation, and on infringers taking licenses; FY2026 guide explicitly excludes new deals until litigation concludes.
Customer / competitive Losing a flagship socket (as with H100 VRM to MPWR) can remove a large volume tranche quickly.
Single-site US fab concentration; Fab One capacity is largely built, but ramp execution has historically been a swing factor.
Advantages
- Best-in-class power density for die-level / vertical power delivery as GPU currents scale
- Enforceable, ITC-validated IP that converts competitors into royalty payers
- Near-100%-incremental-margin royalty stream layered on the hardware business
- Net-cash balance sheet — grows without dilution or leverage
- Capital-light growth (capacity already built) → high FCF conversion in up-cycles
Weaknesses
- Sub-scale (~$450M revenue) versus TXN/ADI/MPWR/Infineon
- Demonstrated socket losses in flagship AI GPU power (H100 VRM)
- Lumpy, litigation-dependent royalty/settlement revenue (FY2025 flattered by $45M settlement)
- Fixed-cost manufacturing amplifies downside operating leverage
- Very high valuation leaves no margin for execution error
Bottlenecks
- Litigation timeline — the new ITC investigation (337-TA-1484, instituted Feb 11, 2026; public hearing set for Oct 14, 2026) gates when conservative-excluded license revenue can be recognized
- Winning and holding high-volume GPU power sockets against MPWR/TXN integrated solutions
- Fixed-cost single-site manufacturing that must stay well-utilized to protect gross margin
- Small engineering/commercial scale versus incumbents chasing the same AI-power dollars
Top signals & trends
Top signals
Forward-demand visibility, not a one-quarter spike.
Driven by both stronger product demand and a new all-inclusive OEM royalty license.
Mix shift to royalty + higher utilization.
Real-world monetization + expansion of the IP moat.
Bearish (valuation) · Prices in years of flawless execution; sentiment-crowded.
Competitive-displacement risk is documented, not theoretical.
Trends
Structural tailwind directly aligned with Vicor's differentiated architecture.
Well-funded incumbents targeting the same sockets with integrated, lower-cost solutions.
High-margin royalty annuity emerging as a second engine — but litigation-gated.
Diversifies beyond the AI-compute cycle.
Enables denser competitor solutions and pricing pressure, though Vicor also leverages advanced device tech.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Vicor manufactures its own MOSFET/magnetics-based power modules — vertically integrated, limiting external wafer-supplier dependence.
Substrate / packaging & passive-component vendors Sources advanced substrates, magnetics materials and packaging inputs for high-density modules (not individually disclosed).
Power modules and licenses tied to AI accelerator systems (Blackwell/Rubin generation) reach Vicor via ODM/OEM system builders.
End-demand for 48V/vertical power delivery in AI compute racks.
High-density DC-DC and 800V/48V architectures for electrification.
High-reliability, rad-tolerant modular power.
Most direct AI-power rival; won the H100 GPU VRM socket, displacing Vicor, with integrated lower-cost multiphase/48V solutions.
300mm/scale economics undercut power-management pricing and pressure Vicor's module margins.
Broad high-performance power portfolio; contests high-frequency/high-density converter niches.
WBG (SiC/GaN) leader with International Rectifier assets; strong in automotive/industrial and pushing into AI 48V.
Named among the AI power-delivery competitor set expanding into 48V/data-center power.
GaN-focused; pushes fast-charge/point-of-load and next-gen data-center power that can erode legacy volumes.
High-voltage/GaN power-conversion specialist competing in point-of-load transitions.
Taiwan-listed power-systems giant and an ITC respondent; scale competitor in data-center power delivery (context only, non-US listing).