> ## Content Index
> Fetch the complete content index at: https://adikumar.co/llms.txt
> Use this file to discover other available public pages before exploring further.

# Vicor direct-to-pin power for AI accelerators
- URL: https://adikumar.co/tech-spotlight-01-vicor-direct-to-pin-power/
- Published: 2026-09-08T14:30:00.000Z
- Updated: 2026-10-07T05:22:11.000Z
- Description: Vicor (NASDAQ: VICR) has been selling factorized-power architecture for forty years. The last three years brought VPD and direct-to-pin modules into the mainstream as AI accelerator power density broke the incumbent multi-stage delivery model. Sub-$4B market cap, ~$400M revenue, meaningfully hi
- Author: Adi Kumar
- Tags: Tech Spotlights, AI Infrastructure, Vicor, Power semiconductors, GPU power, #Essay

Exec brief · Vicor (NASDAQ: VICR)

**Rubin-generation current density (\~1500A per rail at \~1.6V core, per public presentations and industry estimates) may push conventional multi-phase buck past its economic comfort zone.** That physical constraint is what creates the underwriting opportunity in Vicor's factorised power architecture (FPA) plus vertical power delivery (VPD). The opportunity is architectural. It does not rely on any single Nvidia commitment or on a management growth story.

**At \~$200 Vicor falls into the BUY zone of a five-band price framework.** Market cap \~$8.8B on 44M shares. FY26 revenue guide $600M+; long-term target $2.5B at 70% GM (management aspiration, not underwritten forecast). The June 2026 print at $367 represented framings B (mispriced pivot) and E (IP licensor re-rating) getting priced in; the pullback to sub-$200 has re-opened the risk-adjusted entry.

At \~$200, the target-price distribution is sufficiently asymmetric to justify a position, provided the Rubin design win and licensing tripwires remain intact. Weighted expected 2-year target lands at $370\. Weighted 2Y IRR from $200 is +36%. The scenario range is wide, from Bear $71 to Blue-sky $962, reflecting genuine bifurcation on Rubin.

**The historical arc shows that design-win outcomes are rarely binary wins. They are graduated BOM allocations across multiple conversion stages, with different economic weight per stage.** Everything else in this piece supports or falsifies that question.

**Companion:** [Interactive Monte Carlo target model](https://adikumar.co/vicor-model/). Four scenarios, editable probability weights and driver distributions, 10K draws, live target-price and IRR distributions.

TL;DR

Vicor's direct-to-pin power architecture wins on PDN resistance and thermal density against MPS and Infineon and loses on manufacturing scale. The Rubin design-win decision, expected Q1 2027, drives the primary catalyst. At approximately $200 today, weighted expected 2Y IRR runs to +36% inside the BUY zone.

The counter-thesis rests on hyperscaler dual-sourcing preference. Nvidia has historically alternated between Vicor and MPS or monolithic silicon across generation cycles, and any Rubin sole-source outcome runs the tail risk of being designed out on the next architectural refresh. Base case assumes multi-generation stickiness once qualified. Bear case treats each generation as a fresh design-in fight.

**Zones:** BUY at \~$200\. TRIM at $275+. AVOID above $325\. **Vetoes:** Rubin design win lost to MPS or Infineon; adverse outcome on Foxconn appeal; hyperscaler dual-sourcing pattern reasserts on the next architectural refresh.

**Disclosure.** The author may hold a position in Vicor (VICR) or related securities. This analysis reflects the author's views on publicly available information and does not constitute investment advice or a recommendation to buy or sell any security.

### 01The value creation engine

Vicor Corporation was founded in 1981 by Patrizio Vinciarelli in Andover, Massachusetts. Vinciarelli remains Chairman and CEO 44 years on and holds approximately 19% of shares outstanding through direct ownership. Two-founder public companies with intact founder-CEOs in their late 70s are unusual, and the governance implications recur through this analysis.

The business designs high-density power delivery modules and licenses the underlying architectural IP. Revenue comes from two lines. **Advanced Products** targets AI accelerator, telecommunications, electric vehicle and hyperscale computing customers. **Brick Products** serves aerospace and defence, industrial equipment, rail and instrumentation. Advanced Products is the growth engine. Brick Products is the mature base.

FY2025 revenue

$407.7M

+13.5% YoY. $2.61 diluted EPS

Q2 2026 revenue

$143.4M

above $142M revised guide (was $126M)

Advanced Products (Q2)

$94.2M

+26% YoY. +45% sequential. 66% of Q2 revenue

FY2026 guide (raised)

\>$600M

raised from \~$570M in Q2 2026

Revenue trajectory: reported to guide to long-term target $500M $1B $1.5B $2B $2.5B $407.7M FY2025 actual \~$570M FY2026 initial guide \>$600M FY2026 revised (Q2 lift) $2.5B Long-term management aspiration roughly 4x FY2026 guide at 70% gross margin Source: Vicor company disclosures, Q2 2026 earnings release. Long-term target has no publicly disclosed horizon date. 

Reaching $2.5B at 70% gross margin implies roughly 6x the current run rate at a margin profile typically seen in software businesses. This is management's long-term aspiration. The valuation model below does not depend on it.

The value creation engine is the Factorised Power Architecture patent portfolio combined with the manufacturing capability to ship modules based on it. Both channels contribute: module product revenue provides current cash flow, and IP licensing revenue provides the higher-margin re-rating optionality that the market only recently began to price.

### 02The technology moat

The design choice that defines Vicor is the separation of regulation from voltage transformation. Conventional power delivery for high-current loads uses multi-phase buck regulators: a single stage handles voltage step-down, regulation and averaging simultaneously, replicated across 8-16 phases running in interleaved PWM. Monolithic Power Systems, Renesas, Infineon and pre-Empower Analog Devices all took this approach for AI accelerator designs.

Vicor's Factorised Power Architecture (FPA) splits the two functions across two module types. The Pre-Regulator Module (PRM) takes a 48V bus input and produces a regulated intermediate voltage. The Voltage Transformation Module (VTM) then applies a fixed-ratio step-down of 32:1 or 48:1 using Vicor's Sine Amplitude Converter (SAC) topology. The SAC operates as a resonant converter at MHz-range switching frequencies with soft-switching characteristics that suppress switching losses. The physical result is a power delivery network resistance of approximately 5-7 microohms, roughly one-third to one-quarter of what multi-phase buck typically achieves.

Power delivery architectures for AI accelerators, compared MULTI-PHASE BUCK (baseline) 48V bus input ... ... ... 12 to 24 parallel POL phases 1V at die high current VICOR FPA (Factorised Power Architecture) 48V bus input PRM Pre-Regulator Module regulated VTM Voltage Transformation 32:1 or 48:1 SAC (resonant) 1V at die high current Metrics compared in the table below. 

| Metric                         | Multi-phase buck (baseline)              | Vicor FPA                                            |
| ------------------------------ | ---------------------------------------- | ---------------------------------------------------- |
| PDN resistance                 | **15 to 25 μΩ**                          | **5 to 7 μΩ**. Roughly 1/3 to 1/4 of buck.           |
| Switching frequency            | 500 kHz to 2 MHz                         | MHz range (higher operating point)                   |
| Switching characteristics      | Hard-switched                            | Resonant, soft-switched (SAC topology)               |
| Stage architecture             | Single-stage (48V direct to 1V)          | Split (PRM regulates, VTM steps down at fixed ratio) |
| Phase or module count per rail | 12 to 24 parallel POL phases             | One PRM. One VTM.                                    |
| Efficiency implication         | Higher switching losses at MHz operation | Suppressed switching losses via SAC resonance        |

PDN resistance is the first-order metric. At high-current rails, every microohm shows up as thermal loss in the compute die.

The second architectural choice is Vertical Power Delivery. Because the VTM operates in a compact form factor, it can be mounted directly beneath the compute die on the substrate underside (dropping the lateral placement of prior generations). This choice frees the entire die perimeter for signal and interconnect I/O, and drops the 30-50% perimeter reservation typically consumed by lateral power modules. At the \~1500A-plus, \~1.6V operating point often associated with Rubin in public presentations and industry estimates, the perimeter constraint alone could materially change the topology comparison. At those current densities, conventional multi-phase buck becomes increasingly difficult to justify: PDN resistance, thermal density and die-perimeter consumption stop being secondary considerations and become architectural constraints.

Freeing the die perimeter: lateral vs vertical power delivery (cross-section) LATERAL (baseline) cold plate Compute die GPU or ASIC substrate POL modules occupy space lateral to the die Perimeter used by power modules 30-50% VERTICAL (Vicor VPD) cold plate Compute die GPU or ASIC (full perimeter clear) substrate VTM underside VTM mounted below the substrate, directly under die Perimeter used by power modules 0% 

| Dimension                           | Lateral (baseline)                              | Vertical (Vicor VPD)                                 |
| ----------------------------------- | ----------------------------------------------- | ---------------------------------------------------- |
| Power module location               | Multi-phase buck modules at the die perimeter   | VTM on substrate underside, directly beneath the die |
| Die-perimeter reservation for power | **30 to 50%** consumed by lateral power modules | **0%**. Full perimeter available for signal I/O      |
| Signal I/O headroom                 | Constrained by power reservation                | Unconstrained by power placement                     |
| Cold-plate real estate              | Power modules and cold-plate coexist laterally  | Bottom-side VTM competes with chip package pin field |
| Architectural inflection            | Sufficient at standard current densities        | Enabled at \~1500A. \~1.6V (Rubin operating point)   |

At the Rubin operating point, die-perimeter reservation becomes an architectural constraint on topology choice.

Where Vicor trades relative to its peers EV/Sales vs. gross margin · bubble area = market cap ($4B-$120B) · FY26 estimates 30% 40% 50% 60% 70% 0x 5x 10x 15x 20x EV / SALES (FY26) GROSS MARGIN AI / architecture premium Traditional analog / power median EV/S median GM Vicor $8.8B · 15x EV/S · +47% growth MPS $40B · 18x EV/S · +35% growth ADI $120B · 11x EV/S · +15% growth Silicon Labs $4B · 5x EV/S · +20% growth Infineon $50B · 3x EV/S · +8% growth Renesas $30B · 3x EV/S · +10% growth The outlier Only company under $10B in the premium zone. MARKET CAP SCALE $5B $30B $120B Source: consensus FY26 sell-side estimates, market caps as of Aug 2026 

Multi-phase buck versus Vicor's Factorised Power Architecture. Different comfort zones, different manufacturing scale, different IP moat structure.

The trade-off is straightforward. Multi-phase buck wins on cost, silicon-fab-native manufacturing scale, and simplicity at power levels below roughly 1000W per rail. FPA's advantages in efficiency, PDN resistance and thermal density become increasingly valuable above that threshold. The commercial question is which side of the threshold the AI accelerator design is on for the next design cycle. Through H100 and GB200, that answer favoured MPS. Rubin is where the electrical and packaging constraints may make architectures like FPA plus VPD economically compelling instead of only technically advantageous.

**Packaging and substrate co-design is the second-order constraint.** The topology comparison above is purely electrical. The packaging implications are equally consequential and split into distinct commercial frictions that a topology diagram does not capture:

1. **Yield liability.** Placing a VTM directly beneath the compute die on the substrate underside introduces yield exposure that has no analog in lateral multi-phase buck. Solder-bump or micro-via failure under the interposer nullifies an entire compute assembly. The commercial question of who takes the yield loss (Vicor, the OSAT, the ODM, or the hyperscaler) remains unsettled and is a real friction point in the Rubin qualification cycle.
2. **OSAT maturity.** The ODM and OSAT community has extensive experience yielding lateral buck at CoWoS-class complexity but limited experience yielding bottom-side VTMs at commercial volume. TSMC and ASE need to co-develop the qualification path with Vicor, treating the design as a full re-qualification.
3. **Thermal crowding.** At the \~1800W per-GPU budget indicated by Rubin-generation public commentary, top-side cooling is already saturated by the compute die. A bottom-side VTM competes for cold-plate real estate with the chip package pin field. Multi-phase buck lands laterally, out of the cold-plate zone, which is mechanically less elegant but operationally understood.
4. **Procurement inertia.** Multi-phase buck may be electrically inferior above \~1000W per rail, but its assembly chain is fully de-risked at every hyperscaler procurement organisation. FPA advantages have to be paid for in packaging-partner qualification and yield-loss allocation, neither of which appears on a topology diagram.

The transmission mechanism

How the investment case chains from physics to multiple:

**Higher accelerator current density** → PDN resistance and thermal density become binding constraints → **architectural shift** from multi-phase buck to factorised power + VPD → **Vicor qualification** on Nvidia and hyperscaler custom silicon programs → **design win at commercial volume** → module product revenue growth + **licensing revenue leverage** from ITC-driven adoption → **gross margin expansion** toward the 70% target → **multiple re-rating** from specialty semi (5-10x) toward IP-plus-hardware (15-25x).

Each step holds a probability. None is a certainty. The Rubin design win is the earliest observable link in the chain that would move the market's implied probability on every downstream step.

The Rubin-to-target chain Each link is a probability-weighted event, not a certainty. The chain runs from left to right; break any link and downstream states re-anchor. RUBIN OUTCOME No design-in → design-in + licensing (five graduated states) BOM CONTENT CAPTURE Zero → selected stages → material multi-SKU MODULE REVENUE Bear \~$0.7B → Bull \~$2.2B FY28 LICENSING UPSIDE One licensee today → $150-250M/yr upside case GROSS MARGIN \+ MULTIPLE GM 55-70% EV/S 4-15x (scenario-dependent) TARGET PRICE Weighted expected $370 range $71-$962 Reading right-to-left is also useful: any target price implies a specific set of assumptions about every prior link in the chain. 

The Rubin-to-target chain. Each link holds its own probability, and breaking any one link re-anchors every downstream state.

### 03Vicor versus MPS · head-to-head solution comparison

MPS is Vicor's principal incumbent architectural rival for AI power delivery, while Infineon and ADI/Empower increasingly contest the same high-current design space.

| Dimension                         | Monolithic Power (MPS)                                               | Vicor                                                                 |
| --------------------------------- | -------------------------------------------------------------------- | --------------------------------------------------------------------- |
| **Architecture**                  | Multi-phase buck (8-16 phases interleaved)                           | Factorised Power (PRM regulation + VTM transformation)                |
| **Regulation topology**           | Voltage averaging via PWM interleaving                               | Current multiplication via Sine Amplitude Converter                   |
| **Stage architecture**            | Single-stage integration (regulation, step-down, averaging combined) | Split-stage separation (PRM regulates, VTM steps down at fixed ratio) |
| **PDN resistance**                | \~15-25 microohms                                                    | \~5-7 microohms                                                       |
| **Thermal comfort zone**          | Up to roughly 1000W per rail                                         | Extends past 2000W per rail                                           |
| **Placement relative to die**     | Lateral (beside compute die)                                         | Lateral or vertical (VTM under die)                                   |
| **Die perimeter consumption**     | 30-50% consumed by power modules                                     | 0% with vertical placement                                            |
| **BOM cost per GPU**              | Lower (silicon-fab-native, high volume)                              | Higher (specialty modules, lower volume)                              |
| **Manufacturing scale**           | Hundreds of millions of units annually                               | Thousands to low millions annually at higher ASP                      |
| **IP moat**                       | Circuit design, integration know-how, foundry access                 | FPA + SAC + VPD patent portfolio (licensable)                         |
| **Licensing revenue**             | None disclosed                                                       | Growing (Q2 2026 guidance raise partly driven by new licensee)        |
| **Current AI supplier status**    | Dominant on H100 and GB200 in production                             | Displaced from H100/GB200; RFQ candidate for GB300 + Rubin            |
| **Market cap (approx, Sep 2026)** | \~$40B                                                               | \~$8.8B                                                               |
| **EV/Sales (FY26)**               | \~15-20x                                                             | \~15x                                                                 |

Both companies trade at comparable EV/Sales multiples on FY26 revenue, but the underlying growth composition is different. MPS is being valued on continued AI supplier dominance and share gains across every hyperscaler; Vicor is being valued on the Rubin re-entry plus licensing wave. The dispersion of outcomes for Vicor is materially wider than for MPS. That is why the Vicor Monte Carlo below has a much longer tail on both sides than an MPS model would show.

**Incumbent competitive responses are actively reducing the topology-migration urgency.** MPS is advancing hybrid multi-phase configurations and dual-sided board designs that reduce lateral PDN resistance without full migration to split-stage FPA. If those configurations satisfy the Rubin operating range with a 2-3% efficiency penalty versus Vicor, hyperscalers may accept that cost in exchange for MPS supply-chain scale and hyperscaler-tested qualification depth. Infineon brings the more asymmetric threat on the supply side: Kulim and Villach fabs give Infineon roughly 5x Vicor's wafer volume alongside existing hyperscaler qualification relationships. If Infineon under-die VPD proves good-enough technically, the procurement equation may favour Infineon regardless of Vicor's PDN-resistance lead. The Vicor bull case therefore has to explain both why the incumbent workarounds fall short AND why the hyperscaler procurement organisation reweights toward electrical performance, with supply-chain security dropping in weight.

**Infineon and ADI/Empower are the two competitors that deserve as much analytical weight as MPS.** Infineon has developed its own Vertical Power Delivery topology that mounts the VRM directly under the compute die, targeting the same physical constraint that Vicor exploits with VPD. Infineon's advantage is manufacturing scale (roughly 5x Vicor's wafer volume through its Kulim and Villach fabs) plus incumbent qualification relationships with every hyperscaler. If Nvidia decides that Rubin's power-delivery regime requires a VPD topology but wants to preserve MPS-class supply-chain scale and cost, Infineon becomes the natural bridge candidate. ADI/Empower (post the $1.5B July 2026 acquisition) inherits FinFast integrated voltage regulator IP plus ADI's aerospace/defence and industrial customer relationships. FinFast is architecturally distinct from FPA but occupies similar power-density territory, and ADI's scale plus its integrated silicon capacitor (ECAP) technology give it a real qualification pathway.

Nvidia's Rubin supplier decision weighs several attributes together. On PDN resistance and thermal budget, Vicor leads. On manufacturing scale and hyperscaler qualification depth, MPS and Infineon both lead. On strategic-buyer credibility as an incumbent power vendor, ADI leads post-Empower. Cost, manufacturability, qualification history, reliability, supply-chain resilience and packaging capability all enter the same purchasing decision alongside the electrical performance argument. The bull case for Vicor on Rubin therefore has to explain both superior physics and competitive traction across the attributes MPS and Infineon currently lead on.

### 04The Nvidia design-in arc

The arc reads as a U. Vicor loses the design that started the AI era, watches the incumbent multi-phase buck approach come under increasing thermal and packaging pressureasing thermal and packaging pressure at Blackwell, and returns to the RFQ as socket-mounted designs and next-generation current densities emerge.

The Nvidia arc: GPU power trajectory and Vicor's design-win position by generation Per-GPU power in watts across five Nvidia generations. The multi-phase-buck comfort zone ends around 700-1000W; above that, thermal density and PDN resistance become topology-defining constraints. 400W 800W 1200W 1600W 2000W Multi-phase buck comfort zone ends Thermal density + PDN resistance bind 2020 2021 2022 2023 2024 2025 2026 2027 A100 400W H100 700W GB200 Blackwell 1200W GB300 Blackwell Ultra 1400W Rubin 1800W A100 Vicor bricks in ecosystem H100 DISPLACED by MPS GB200 Blackwell Out at scale, MPS primary with thermal wobbles GB300 Blackwell Ultra Contested: socket-mounted design re-opens supplier RFQs Rubin FRAMED as the only solution for 1.6V core at 1500A+ Arc reads as a U: Vicor loses the design that started the AI era, watches MPS struggle at the density inflection, and comes back into the RFQ as socket-mounted designs return. Between 2020 and 2027, Nvidia's power delivery architecture moved across four distinct procurement regimes, each recorded here to reveal what drove supplier turnover across generations. 

**H100 (2023). The displacement.** MPS displaced Vicor as primary VRM supplier on Nvidia H100\. The reason was structural. At 700W per SXM module, multi-phase buck sat inside its thermal comfort zone. MPS combined lower cost, lateral integration simplicity and hyperscaler-tested supply-chain scale. Vicor's FPA delivered better efficiency at a price premium the H100 design did not need to pay. VICR stock derated from roughly $100 to roughly $40 across 2023-2024 as the market internalised that the AI era's flagship design had gone to a competitor.

**GB200 Blackwell (2024-2025). The wobble.** Blackwell pushed per-GPU power to 1200W and the Grace-Blackwell superchip to 2700W. MPS remained the primary VRM supplier. At the higher Blackwell power budget, industry reporting (Edgewater Research and others in 2025) highlighted thermal and Nvidia-allocation concerns, while Renesas and Infineon gained share as Nvidia diversified its supplier base. Vicor was largely out at commercial scale on GB200, in evaluation only.

**GB300 Blackwell Ultra (2025-2026). The reset.** Rumours from GTC 2026 and supply-chain leaks indicate Nvidia is moving Blackwell Ultra to a socket-mounted configuration, the first Nvidia GPU using a mechanical socket. AMD moved to socket-mounted packaging with MI300A in 2023; Nvidia follows. Socket-mounted designs re-open the power delivery RFQ. Mechanical constraint changes and field-serviceability requirements favour vendors with vertical-integration IP and reference-design tooling. Vicor is back in evaluation at GB300, alongside ADI's post-Empower FinFast platform.

**Rubin (2027). The framing test.** Public presentations and industry estimates around Rubin point to roughly 1.6V core operation at current levels above 1500A per rail. If those industry estimates prove directionally correct, conventional multi-phase buck becomes progressively harder to defend at that operating point: PDN resistance dominates, thermal density dominates, and the perimeter footprint constraint becomes binding on system-level design. Factorised power architecture with vertical power delivery is one of the small number of solutions positioned to deliver. Rubin shifts the power-delivery regime in a direction where Vicor's historical cost disadvantage may cease to dominate the purchasing decision. That is a materially different claim from architectural inevitability. The underlying current-density constraint is credible on the physics. The stronger claim Vicor makes publicly, per GTC 2026 and APEC presentations, is that it is essentially the only supplier capable of addressing the current density at the 1.6V Rubin core voltage. That is a vendor assertion that Infineon and ADI/Empower have to disprove or validate through qualification. Infineon has a competing VPD approach placing the VRM directly under the GPU. ADI/Empower has FinFast targeting similar current levels. Vicor has company at Rubin, though the credible field narrows substantially from the H100 and GB200 era.

Nvidia GPU power delivery: supplier position matrix by generation Supplier position on each Nvidia GPU RFQ. Cell colour marks position. Vicor loses the H100 design, misses GB200 at scale, re-enters at GB300, is positioned as the lead-RFQ candidate for Rubin. H100 (2023) GB200 (2024-25) GB300 (2025-26) Rubin (2027) Vicor OUT OUT EVALUATION LEAD RFQ MPS PRIMARY PRIMARY (thermal wobbles) CONTESTED AT RISK on density Renesas SECONDARY GAINED SHARE IN THE MIX UNCLEAR Infineon SECONDARY GAINED SHARE CONTESTED VPD-topology RFQ ADI (post-Empower) n/a, pre-acquisition n/a, pre-acquisition EVALUATION via Empower FinFast LEAD RFQ via FinFast + ECAP PRIMARY / LEAD RFQ PRIMARY WITH STRESS / GAINED SHARE CONTESTED / AT RISK EVALUATION OUT SECONDARY n/a, not yet in market 

Supplier position across the four Nvidia GPU generations. Vicor loses H100, misses GB200 at scale, re-enters at GB300, is positioned as a lead-RFQ candidate for Rubin.

**Second-source friction cuts against Vicor even if the technical qualification lands.** Hyperscalers and Nvidia refuse mission-critical power delivery from a single source; the H100 displacement of Vicor by MPS was partly a supply-chain security decision as much as an electrical one. Under Vicor's ITC-driven licensing framework, competitors that agree to license FPA can theoretically become qualified second-source suppliers. In practice, moving from licensing agreement to qualified merchant-built FPA module at production yield takes twelve to eighteen months. If that lag runs past the Rubin ramp qualification window, Vicor could win the design position technically but slide commercial volume into late 2027 or 2028 while Nvidia routes primary volume to whatever second-source alternative exists.

**Rubin outcome matrix.** The design-win outcome spans five graduated levels, not binary. Vicor content capture on Rubin can land at any of five graduated levels, each with a distinct technical and economic implication. The right-hand column maps each outcome to the framings discussed below.

| Outcome                      | Technical implication                                                               | Vicor economics                                                                                           | Framing mapping                           |
| ---------------------------- | ----------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------- | ----------------------------------------- |
| **No design-in**             | Conventional POL from MPS or Infineon remains adequate for Rubin power delivery     | Thesis breaks; the value-trap framing dominates                                                           | Broken AI, Value Trap                     |
| **Qualification only**       | Architecture validated on reference designs without production volume               | Optionality preserved; small revenue lift; the architecture-ahead-of-its-time framing partially confirmed | Architecture Ahead partial                |
| **Partial content**          | Vicor captures selected conversion stages or specific accelerator variants          | Moderate upside; base scenario compatible                                                                 | Mispriced Pivot, Architecture Ahead       |
| **Broad platform design-in** | FPA plus VPD becomes material Rubin BOM across multiple accelerator SKUs            | Core bull case; both the architecture-ahead-of-its-time and the IP-licensor-re-rating framings strengthen | Architecture Ahead, IP Licensor Re-rating |
| **Design-in plus licensing** | Hardware content + IP-royalty economics from competitor licenses under ITC exposure | Blue-sky case; multiple re-rating available                                                               | IP Licensor Re-rating, Blue-sky           |

The critical distinction is between the middle three outcomes. Public commentary on "Vicor at Rubin" typically collapses "qualification only", "partial content", and "broad platform design-in" into a single "design win". The economic implications differ by an order of magnitude. Watching disclosure evidence in Q3 2026 through Q2 2027 for which of these outcomes lands is the most valuable analytical exercise a reader can do.

### 05Commercial pipeline

Vicor's confirmed AI-adjacent customer base is thinner than the current market cap implies. Nvidia is the historic anchor customer, with exposure now much reduced post-H100 displacement. Hyperscaler custom silicon programs are referenced repeatedly in earnings commentary but not consistently named. Marvell has appeared in trade-press coverage as a Vicor engagement, though the disclosure trail is weaker than the Ferric-Marvell announcement discussed elsewhere on this site. The public commercial pipeline outside AI includes aerospace and defence primes (Lockheed, Raytheon historically), industrial equipment makers, rail infrastructure, and telecom OEMs.

Backlog stood at $176.9M at FY25 year-end. Q2 2026 inventories increased as the company prepared for higher volume against that backlog, per the 10-Q. The 45% sequential Advanced Products growth in Q2 2026 is the clearest signal yet that the growth engine is accelerating, although the public disclosure does not establish how much of that acceleration is attributable to AI programs specifically; the 10-Q language points to higher royalty revenue and improved product demand, with no named AI customer disclosed. Investors underwriting the current price implicitly accept that a substantial portion of the growth comes from as-yet-unnamed hyperscaler custom silicon programs plus early Rubin design-in activity, both of which will eventually be disclosed but are currently opaque.

The Brick Products base at approximately $150-170M annual revenue provides valuation support that most AI-story valuations lack. Even in a bear scenario where Advanced Products growth stalls, the industrial and defence revenue base remains cash-generative and profitable.

**The central commercial tension.** The stock is increasingly priced like an AI infrastructure winner, while the publicly attributable AI customer base remains relatively narrow and thinly disclosed. The Q2 2026 45% sequential Advanced Products print landed as reported, but investors have to accept management's framing about where the growth is coming from without granular customer-level attribution. That opacity is the sharpest bear argument on the table. The bear case only requires the currently-unnamed customer base to turn out less durable or less AI-specific than the multiple implies, with Rubin held constant.

### 06Segment split and financial performance

**The two-line business:** Advanced Products (AI + hyperscaler + EV + telecom) versus Brick Products (aerospace + defence + industrial + rail).

| Metric                   | FY2024  | FY2025           | Q2 2026           | FY2026 guide      | Long-term target |
| ------------------------ | ------- | ---------------- | ----------------- | ----------------- | ---------------- |
| Total revenue            | $359M   | $407.7M (+13.5%) | $143.4M           | $600M+            | $2.5B            |
| Advanced Products (est.) | \~$210M | \~$265M (+26%)   | $94.2M (+45% seq) | \~$430M           | \~$2.0B          |
| Brick Products (est.)    | \~$149M | \~$143M (-1.6%)  | \~$49M            | \~$170M           | \~$500M          |
| Gross margin             | \~48%   | \~53%            | \~57%             | \~60%             | 70%              |
| EPS (diluted)            | \~$0.90 | $2.61            | n/a               | \~$4.00 (implied) | n/a              |

**Advanced Products is the operating leverage engine.** Growth accelerated from 26% year-over-year in FY25 to 45% sequential in Q2 2026\. Brick Products has been in mild decline (-1.6% FY25) but remains cash-generative. The revenue mix shift from 58% Brick / 42% Advanced (FY24) to a Q2 2026 run-rate of roughly 34% Brick / 66% Advanced is the reason gross margin has climbed from 48% to a Q2 2026 blended level near 57%. Advanced Products delivers structurally higher gross margin, and the AI design wins deliver higher still because they include royalty revenue.

**Insider signals.** Vinciarelli holds 8.4M shares (approximately 19% of the 44M outstanding). Vinciarelli adopted a Rule 10b5-1 selling plan on February 26, 2026 and has subsequently sold shares under that plan at prices in the roughly $260-$354 range during the June quarter and into early July 2026 per Form 4 filings. Aggregate proceeds to Vinciarelli from Vicor share sales over the twelve months to June 2026 (across the current 10b5-1 plan plus any preceding sales) were approximately $500M (calculated from Form 4 filings; readers can audit at the [SEC EDGAR Form 4 filing page for Vicor Corp (CIK 0000751978)](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000751978&type=4&dateb=&owner=include&count=40&ref=adikumar.co)). Large founder selling under a pre-arranged plan is procedurally neutral. It is worth monitoring for changes in pace or plan terms; the base level is a null signal.

Comp set · EV/Sales versus Gross margin Bubble area scales with market capitalisation. FY26 estimates. Vicor and MPS trade at multiples the traditional analog stack does not. 30% 40% 50% 60% 70% 0x 5x 10x 15x 20x EV/Sales (FY26) Gross margin AI / architecture premium Traditional analog / power Vicor $8.8B · +47% MPS $40B · +35% ADI $120B · +15% Silicon Labs $4B · +20% Infineon $50B · +8% Renesas $30B · +10% Bubble area = market cap $5B $30B $120B 

### 07Public-market comp set

Vicor screens against a set of specialty semi companies with power delivery exposure. The multiples table below anchors the valuation discussion below.

| Company                            | Market cap | FY26 revenue est. | EV/Sales | Gross margin  | Growth    |
| ---------------------------------- | ---------- | ----------------- | -------- | ------------- | --------- |
| **Vicor (VICR)**                   | \~$8.8B    | \~$600M           | \~15x    | \~60% (Q2 26) | +47% FY26 |
| Monolithic Power (MPWR)            | \~$40B     | \~$2.2B           | \~18x    | \~55%         | +35% FY26 |
| Analog Devices (ADI, post-Empower) | \~$120B    | \~$11B            | \~11x    | \~65%         | +15% FY26 |
| Silicon Labs (SLAB)                | \~$4B      | \~$800M           | \~5x     | \~55%         | +20% FY26 |
| Infineon Technologies (IFX)        | \~$50B     | \~$16B            | \~3x     | \~40%         | +8% FY26  |
| Renesas Electronics (RNECY)        | \~$30B     | \~$10B            | \~3x     | \~55%         | +10% FY26 |

Vicor and MPS both trade at premium multiples versus the mainstream power-semi peer group because both are seen as AI-share-taking specialists. Vicor's \~15x is slightly below MPS's \~18x despite Vicor's faster near-term growth rate, reflecting the market's remaining scepticism about the design-win trajectory. If Rubin design-in confirms, the multiple gap should close.

The relevant sensitivity is that specialty-semi multiples compress quickly when growth disappoints. Silicon Labs' multiple has compressed materially over the past three years as AI-adjacent revenue conversion ran slower than earlier enthusiasm implied. That trajectory is the reference case for what typically happens to VICR-type multiples if Rubin does not land.

### 08The competing framings

The Vicor investment case appears in the market as a set of competing framings, each a coherent read on the same set of facts. We address each explicitly because different practitioners will find different framings persuasive, and the investment call needs to be robust to which framing the reader ends up assigning highest weight to.

**Broken AI story.** Vicor lost the H100 design to MPS, then failed to win back the GB200 design at scale. The AI story is over; what remains is a specialty industrial power business with an ageing founder-CEO. *Current assessment:* historically valid through 2023-2024, superseded by the 2026 re-rating. The framing is materially weakened if Advanced Products sustains growth above 25% year-over-year AND disclosed customer or product evidence confirms that the acceleration comes from AI or hyperscaler programs, separated from royalty timing, mix, and industrial/telecom demand. Revenue growth by itself is insufficient; the growth must trace to durable AI-adjacent programs. The framing reactivates if later prints show growth deceleration or growth that fails the attribution test.

**Mispriced pivot.** The market has been valuing Vicor as a module supplier when the real business is IP licensing plus a differentiated architecture positioned for the next-generation power delivery cycle. The February 2025 ITC win, the new licensee signed in Q2 2026, and the $2.5B long-term revenue target all support the re-pricing thesis. *Current assessment:* was true when market cap was $4B in early 2025\. Now that market cap has moved to approximately $8.8B (having reportedly peaked at approximately $17B in June 2026), the mispricing has been substantially resolved. Additional mispricing exists only if the Foxconn appeal cuts favourably plus additional major OEM licensees sign in Q3-Q4 2026.

**Architecture ahead of its time.** Vicor loses individual design battles because multi-phase buck is cost-optimal at each generation's power budget, but gains ground as current density forces the topology to shift. Rubin is the inflection point where FPA plus VPD may become economically compelling, moving from technically advantageous into commercial pull. *Current assessment:* this is the current market consensus that justifies the $8.8B market cap. It requires Rubin design-win confirmation at commercial scale to hold. It is the framing on which most of the bull case rests.

**Value trap.** Good technology, entrenched competitor (MPS has 10x manufacturing scale plus hyperscaler-first relationships), founder-driven company with a 19% control block that limits any acquirer's willingness to pay a premium, IP that is harder to monetise at scale than management claims. Vinciarelli's 10b5-1 selling program adopted in early 2026 is an active governance factor to monitor for changes in pace or plan terms. *Current assessment:* active risk. This framing gains weight if the Rubin design win is disclosed as MPS/Infineon primary or if the founder selling accelerates. It is the bear-scenario anchor.

**IP licensor re-rating.** The larger value creation comes from licensing FPA, with module sales to competitors under ITC exposure treated as a residual path to competitors under ITC exposure. The February 2025 ITC ruling created a compulsion-to-license among Delta, Cyntec and other named respondents. The Q2 2026 guidance raise, per company disclosure, from approximately $126M to $142M was partly driven by one new licensee signing. The public evidence establishes one new all-inclusive license agreement plus higher royalty revenue in Q2 2026\. It does not establish a contracted path to two or three additional major licensees. The $150-250M annual licensing revenue outcome is properly labelled an upside scenario. It does not represent a base-case forecast or a disclosed pipeline. If two or three additional major OEMs sign licenses in the next 12-18 months, licensing revenue could plausibly reach $150-250M/year at approximately 90% gross margin, and that revenue-mix shift alone would justify material multiple expansion. That is the licensing upside case. *Current assessment:* partially priced. The market has recognised the licensing tailwind, but the fuller potential requires either a favourable Foxconn appeal outcome or additional OEMs facing parallel ITC actions.

### 09Price-band verdict framework

The valuation framework we apply to Vicor is a price-band verdict grid. For each price band from $150 to $400, we compute implied EV/Sales under three FY28 revenue scenarios and assign a verdict on whether that price is asymmetric buy, fair or overpay.

Vicor (VICR) price-band verdict framework at Sep 2026 44M shares outstanding, FY26 revenue $600M+, long-term target $2.5B at 70% GM. Verdict per band derives from implied FY28 EV/Sales at $1.5B revenue base case. Today: trading under $200\. Lands in the BUY zone. $150 - $175 GREAT BUY FY28 EV/S: 4.4x - 5.1x Mkt cap $6.6B - $7.7B $175 - $225 BUY (today: sub-$200) FY28 EV/S: 5.1x - 6.6x Mkt cap $7.7B - $9.9B $225 - $275 HOLD FY28 EV/S: 6.6x - 8.1x Mkt cap $9.9B - $12.1B $275 - $325 TRIM FY28 EV/S: 8.1x - 9.5x Mkt cap $12.1B - $14.3B $325 - $400 AVOID · SHORT IF WEAKENS FY28 EV/S: 9.5x - 11.7x Mkt cap $14.3B - $17.6B $150 $175 $200 $225 $250 $275 $300 $325 $350 $375 $400 Share price (USD, illustrative bands · today \~$200) ▼ TODAY GREAT BUY $150 - $175 Prices in bear-case Rubin loss. Downside supported by brick + non-AI base. Rubin optionality is free. BUY (today: sub-$200) $175 - $225 Prices in mixed Rubin + stable licensing. Room for re-rating on Rubin scale-up. Positive EV. HOLD $225 - $275 Fully prices Rubin design win \+ licensing $50-100M/year. Balanced risk-reward. TRIM $275 - $325 Requires Rubin AT SCALE \+ licensing wave >$100M/year. Margin of safety erodes. AVOID · SHORT IF WEAKENS $325 - $400 SaaS-multiple pricing on semi biz. Vinciarelli 10b5-1 selling active here. Requires perfect execution. 

Five price bands with colour-coded verdicts. Today at \~$200 lands inside the BUY zone.

**Today at approximately $200 lands inside the BUY zone.** The implied FY28 EV/Sales range under a base-case $1.5B FY28 revenue is 5.1-6.6x, a multiple that becomes defensible if the Rubin design win and licensing cases convert. Room for re-rating exists if Rubin design win confirms at scale.

The framework flips to HOLD in the $225-275 band, TRIM in $275-325 (where Vinciarelli's 10b5-1 program has been active), and AVOID or SHORT-candidate above $325\. The framework flips to GREAT BUY below $175, where downside is anchored by the brick and industrial revenue base providing valuation support around a \~$5B market-cap reference case, with Rubin optionality effectively free.

### 10The scenario-weighted target model

The price-band verdict framework is the static view. The dynamic view is a Monte Carlo target model that draws from four scenarios with probability weights and derives a distribution of target prices plus a probability-weighted expected value.

Vicor scenario targets versus current price Target price per scenario as bar height. Bar width proportional to probability weight. Reference lines show current price (red) and probability-weighted expected target (deep vermillion). $0 $200 $400 $600 $800 $1000 $71 BEAR $262 BASE $607 BULL $962 BLUE-SKY 25% prob FY28 rev $0.7B · EV/S 4x 2Y IRR -41% BASE 40% prob FY28 rev $1.4B · EV/S 8x 2Y IRR +14% BULL 25% prob FY28 rev $2.2B · EV/S 12x 2Y IRR +74% BLUE-SKY 10% prob FY28 rev $2.8B · EV/S 15x 2Y IRR +119% ▲ Today $200 ◆ Weighted expected $370 TARGET PRICE (USD) Bar width proportional to scenario probability weight. IRR shown is CAGR to reach the scenario target from $200 over two years. 

Scenario targets versus current price. Bar width proportional to probability weight; bar height is the scenario target.

Vicor scenario tree: probability-weighted target price Each scenario holds a probability, an implied FY28 revenue and multiple, and a target share price. Weighted expected target uses these probabilities as our current confidence read. Today: \~$200\. Probability-weighted 2Y expected target: $370\. Weighted expected 2Y IRR: +36%. TODAY $200 Sep 2026 25% BEAR FY28 rev $0.7B · EV/Sales 4x Target $71 2Y IRR -41% Rubin lost; licensing plateau; growth stalls 40% BASE FY28 rev $1.4B · EV/Sales 8x Target $262 2Y IRR +14% Rubin partial win; licensing $80M/yr; steady growth 25% BULL FY28 rev $2.2B · EV/Sales 12x Target $607 2Y IRR +74% Rubin design win; licensing wave; margin expands 10% BLUE-SKY FY28 rev $2.8B · EV/Sales 15x Target $962 2Y IRR +119% Rubin sole-source; strategic bid; blue-sky IP WEIGHTED EXPECTED TARGET $370 +85% vs today · 2Y IRR +36% Weighted range (bear to blue-sky): $71 - $962\. Base + Bull cover 65% of probability mass. CONFIDENCE READ These probability weights are our current read. They will shift with each earnings disclosure. Rubin design-win confirmation alone would move bull + blue-sky above 50%. SENSITIVITY Bear prob 15% → weighted target $442 Bear prob 25% → weighted target $370 (this) Bear prob 35% → weighted target $299 

**Rubin design-win outcome as the primary latent driver.** The four scenarios below are financial expressions of the underlying strategic question about Rubin content capture, mapped from the outcome matrix above. Each scenario holds its own probability weight; the interactive model at /vicor-model/ lets you re-weight directly. A future v2 will decompose the Rubin outcome states as first-class Monte Carlo inputs, so the financial scenarios derive causally from Rubin content capture, with parallel constructs held out.

**Scenario table and probability weights (weighted expected target $370):**

| Scenario     | Prob | FY28 Rev | EV/Sales | Target | 2Y IRR from $200 | Dominant framing                                                  |
| ------------ | ---- | -------- | -------- | ------ | ---------------- | ----------------------------------------------------------------- |
| **Bear**     | 25%  | $0.7B    | 4x       | $71    | \-41%            | Broken AI or Value Trap                                           |
| **Base**     | 40%  | $1.4B    | 8x       | $262   | +14%             | Mispriced Pivot with partial Architecture Ahead of its Time       |
| **Bull**     | 25%  | $2.2B    | 12x      | $607   | +74%             | Architecture Ahead of its Time with partial IP Licensor Re-rating |
| **Blue-sky** | 10%  | $2.8B    | 15x      | $962   | +119%            | IP Licensor Re-rating with strategic sole-source                  |

**Gross margin mechanics deserve explicit treatment.** The $2.5B revenue target at 70% blended gross margin embeds a specific revenue mix assumption. Module product revenue delivers approximately 50-55% gross margin at scale (consistent with specialty semi peer economics). IP licensing and royalty revenue delivers 90%+ gross margin. To reach 70% blended, licensing needs to become approximately 30-40% of the revenue mix at approximately $2.5B run rate. Licensing is currently a small single-digit share. If module hardware continues to dominate FY28 revenue (75%+ mix), blended gross margin lands closer to 55-60%, and the specialty-semi multiple compresses toward 6-8x EV/Sales, well below the 12-15x embedded in the Bull and Blue-sky scenarios. The scenarios in the model therefore embed both the revenue path AND a specific mix-shift toward licensing that the reader should treat as a joint probability with the underlying revenue path.

**Tail-scenario asymmetry.** The Bull and Blue-sky scenarios together hold 35% probability mass but contribute approximately 67% of the probability-weighted expected value ($248 of $370). That concentration is the actual investment tension. The underwriting outcome is dominated by a set of scenarios that must land through joint clearing of three linked conditions (all partially causally connected: a Rubin BOM win raises licensing credibility, which raises adoption probability, which raises the licensing mix and gross margin): Rubin BOM capture at commercial scale, licensing wave conversion to multiple additional major OEMs, AND gross-margin mix shift toward the 30-40% licensing share. Any one of the three failing collapses the target toward the Base ($262) or below. The interactive model at /vicor-model/ lets you re-weight scenarios directly or compress the terminal multiples to visualise this sensitivity.

**Probability-weighted expected value: $370\. Weighted 2Y IRR from $200: +36%.**

The $370 figure is the mean of a deliberately wide outcome distribution designed to make the asymmetry of the risk-adjusted setup visible. It does not represent a recommended exit price and does not imply that $370 is where the stock should trade.

The 90% scenario probability mass spans from $71 (bear scenario target) up to $607 (bull scenario target), meaning the Bear + Base + Bull scenarios together cover 90% of probability weight. This is a scenario span, distinct from a conventional Monte Carlo P5-P95 interval calculated across all 10,000 draws. The interactive companion at /vicor-model/ produces the actual model P5-P95 range, which is wider than the scenario span because it captures within-scenario variance. Base and Bull scenarios together cover 65% of the probability mass. Blue-sky is treated as a low-probability tail contributing approximately $96 to the weighted target on its own weight.

Interactive companion model

[→ Open the Vicor probabilistic target model](https://adikumar.co/vicor-model/)

Adjust probability weights and P10/P50/P90 driver distributions per scenario. 10,000 draws per run. Target-price and 2-year IRR distributions update live. Correlated inputs (Gaussian copula, ρ=0.6 within scenario). Includes sensitivity tornado to identify which input drives the most variance in the weighted target.

**Sensitivity to bear-probability re-weighting.** At Bear = 15% the weighted target rises to $442\. At Bear = 35% it falls to $299\. Even under the more bearish sensitivity the weighted target clears today's price by a wide margin.

### 11Vetoes and tripwires

The thesis vetoes below would materially damage the underwriting economics regardless of anything else. One risk override accompanies them: a governance and execution factor that does not automatically kill the thesis but introduces enough uncertainty to force re-underwriting. Three tripwires would narrow the bull case without necessarily killing the standalone story.

Thesis vetoes

- **Vicor loses Rubin design win entirely to MPS or Infineon.** If Q1 2027 or later disclosures confirm MPS or Infineon as primary Rubin VRM supplier and Vicor as evaluation-only or fully out, framings C, E and Blue-sky all collapse. The residual thesis is the value-trap framing at approximately $70-100/share.
- **Adverse outcome on Foxconn appeal and related licensing enforcement.** Vicor is appealing the ITC's finding that FII USA and Ingrasys have a license to patent '761 via boilerplate purchase-order clauses. The two-sided outcome tree matters. Appeal outcome favourable to Vicor: Foxconn's license claim is overturned, the exclusion order applies more broadly, and enforcement leverage against Delta, Cyntec, Quanta, and other respondents strengthens. The IP-licensor-re-rating framing strengthens materially and licensing revenue converges toward the $150-250M upside scenario. Appeal outcome unfavourable to Vicor: the ITC finding stands, boilerplate purchase-order licenses become a widely-used defensive template, and enforcement leverage against other respondents weakens. The IP-licensor-re-rating framing weakens and licensing revenue plateaus near current levels. The Foxconn veto fires in the unfavourable case.
- **Hyperscaler dual-sourcing pattern reasserts on the next architectural refresh.** Nvidia and hyperscaler ASIC customers have historically resisted sole-sourcing on critical power silicon. Vicor has been designed out on prior generation transitions when the incumbent decision favoured MPS or monolithic silicon. Even if Rubin lands Vicor as primary, the follow-on architecture (Rubin Ultra, Feynman, or downstream custom ASIC) could pivot to a second source. Base case assumes multi-generation stickiness once qualified. Bear case treats each generation as a fresh design-in fight and prices the Rubin design win as a single-generation event.

The Federal Circuit outcome brings systemic implications beyond the specific Foxconn dispute. If the appellate court upholds the doctrine that standard purchase-order terms confer an implied license to downstream ODMs, the ITC-driven royalty monetisation framework faces systemic leakage across every merchant power-semi supplier's licensing model. Cyntec, Delta, Quanta, and any future respondent could invoke the same boilerplate-clause defence. Vicor licensing revenue projections depend on this precedent not becoming universally applicable. This is a legal-precedent risk larger than the specific case, and the appellate outcome is the licensing catalyst most likely to reprice the framework in Q1 2027.

Risk override

**Risk override · Vinciarelli succession event.** Vinciarelli is 78 and has been Chairman/CEO for 44 years with no publicly named successor. A sudden health event or forced retirement would trigger a strategic disruption with unpredictable effects on Rubin design-in negotiations, licensing negotiations, and any strategic-acquirer conversation. The 19% control block adds complexity to any transition.

Tripwires

- **Gross margin drops below 55% sustained for two consecutive quarters.** The 60%+ implied FY26 gross margin depends on continued mix shift toward Advanced Products plus licensing revenue. If either lever fails, the margin story compresses. The long-term $2.5B at 70% GM target becomes unreachable.
- **Advanced Products year-over-year growth drops below 15% for two consecutive quarters.** The +45% sequential Q2 2026 print is exceptional. Sustained growth of 25%+ year-over-year is what the current market cap implies. If growth decelerates below 15%, the AI-hockey-stick narrative dies and the valuation compresses to industrial-power multiples.
- **Vinciarelli 10b5-1 selling accelerates beyond the current pace or is amended to increase share disposition.** The February 2026 plan is a soft neutral signal at current pace. An amendment to accelerate would be a firmer signal from the person with the most information about the company's actual trajectory.

### 12The investment call

The call is trigger-based. At approximately $200 today, the price lands inside the BUY zone with a weighted expected 2Y IRR of +36%. That is the base recommendation.

- **Upgrade to STRONG LONG if:** Rubin design-win confirmed publicly by Nvidia or a major hyperscaler ASIC customer, AND recurring licensing revenue crosses $50M/year in two consecutive quarters. Under this trigger the framework moves Bull + Blue-sky combined above 50% probability, and the weighted expected target moves toward $500.
- **Downgrade to HOLD if:** Nvidia GB300 or Rubin design win disclosed as MPS or Infineon primary, OR Vinciarelli 10b5-1 selling accelerates beyond current pace. Under this trigger the framework moves Bear above 40% and the weighted expected target compresses toward $250.
- **TRIM at $275 and above.** The upper end of the HOLD band and the start of the TRIM zone in the price-band framework. Vinciarelli was actively selling in this zone during Q1-Q2 2026.
- **AVOID above $325; short candidate only if Rubin or licensing evidence subsequently weakens.** Above $325 requires successful execution across Rubin design win, licensing wave, margin expansion and industrial recovery simultaneously. SaaS-multiple pricing on a specialty semi business without the recurring-revenue characteristics that justify SaaS multiples. Overpayment thesis at this level; short thesis requires additional catalyst.
- **Downgrade to SIT-OUT if:** Both thesis vetoes fire together, or one thesis veto fires alongside a risk-override event. Reassess from a full re-underwriting standpoint (with existing weights held out).

Target prices age poorly. Use the interactive model at [/vicor-model/](https://adikumar.co/vicor-model/) with your own probability weights and driver distributions to generate a target range that reflects your current confidence read on each of the vetoes and tripwires above.

### 13Watch list · Q4 2026 through Q4 2027

The following disclosures will materially shift the framework over the next 12-15 months. Each includes an approximate calendar window and the framing it affects.

Watchlist timeline · Q4 2026 through Q4 2027 Timeline covers disclosure events from Oct 2026 through Q4 2027, plus one ongoing monitor. Each event lifts or lowers specific framings; arrows show bidirectional impact. ONGOING · MONITORED CONTINUOUSLY Vinciarelli 10b5-1 pace changes · Form 4 filings: signals if selling accelerates or amends Value Trap (D) ↑↓ Oct 2026 Q3 2026 earnings call Licensing revenue break-out, Rubin design-in commentary Architecture (C) IP Licensor (E) ↑↓ ↑↓ Q4 2026 Hyperscaler ASIC confirmation Any named customer: Marvell, Broadcom, AWS, Google TPU, Meta MTIA Architecture (C) ↑↓ Q1 2027 Foxconn ITC appeal outcome Bidirectional: favourable strengthens E; unfavourable fires Foxconn veto Mispriced Pivot (B) IP Licensor (E) ↑↓ ↑↓ March 2027 GTC 2027 Rubin production status, supplier disclosures Architecture (C) · Blue-sky ↑↓ Q2 2027 Rubin design-in confirmation window Which of the 5 Rubin outcome states lands: no design-in / qualification only / partial / broad / design-in + Architecture (C) · Blue-sky ↑↓ Q3-Q4 2027 Additional OEM licensees under ITC exposure Second and third major licensees would move E from partial-priced to fully priced IP Licensor (E) ↑↓ Framing key: Broken AI Story (A) · Mispriced Pivot (B) · Architecture Ahead of Time (C) · Value Trap (D) · IP Licensor Re-rating (E) · Blue-sky = tail scenario. Arrows ↑↓ mark bidirectional impact. 

Watchlist events on a horizontal timeline. Bidirectional arrows indicate whether the event strengthens or weakens the named framings.

Set alerts on Vicor's IR page for 10-Q + 10-K + 8-K filings, and on the ITC docket for any appeal-related orders. The Rubin story specifically will unfold across GTC 2027 (March) and the subsequent hyperscaler earnings cycle. The licensing story will unfold via quarterly guidance revisions and formal 8-K disclosures.

Related essays in the corpus

**[Tech Spotlight 05: Wolfspeed's Second Act](https://adikumar.co/tech-spotlight-05-wolfspeed-sic-retrospective/).** Sibling issue in the Tech Spotlight series. Wolfspeed's post-Chapter 11 MV upstream thesis in the AI DC power stack, three equity scenarios, and the customer set spanning Amperesand, DG Matrix, Heron Power, and the SST programmes at Eaton, Hitachi, ABB, Schneider, Siemens, and Vertiv.

**[Empower Semiconductor: What ADI bought for $1.5B](https://adikumar.co/empower-adi-15b-acquisition/).** The comparable M&A precedent for the same power-semi segment. Establishes the acquisition price range for an IVR-adjacent design-win at scale.

**[Underwriting Ferric after the Empower exit](https://adikumar.co/ferric-underwriting-standalone-ivr-2026/).** The competing architectural approach to on-package power for AI accelerators.

**[Grid-to-Chip P5: Rack to Gate](https://adikumar.co/grid-to-chip-p5-rack-to-gate/).** Where SiC, GaN, POL and the last power-conversion cascade land in the Grid-to-Chip framework. Situates Vicor's direct-to-pin in the broader power topology.

**[The AI Semiconductor Waterfall: $34M of silicon per MW of compute](https://adikumar.co/ai-semiconductor-waterfall-per-mw/).** Where power semiconductors land in the total silicon spend per megawatt of AI compute capacity.

**[Wide-Bandgap Semis in 2026](https://adikumar.co/wbg-semis-2026-nvts-wolfspeed-innoscience/).** NVTS, Wolfspeed, Innoscience and the contest for share in the upstream power segment feeding the same AI infrastructure buildout.

### 14Glossary of terms used

- **FPA (Factorised Power Architecture).** Vicor's design approach separating voltage regulation (PRM) from voltage transformation (VTM) into two modules. Contrasts with multi-phase buck where a single stage handles both.
- **SAC (Sine Amplitude Converter).** Resonant converter topology inside the VTM. Operates at MHz-range switching frequencies with soft-switching to suppress switching losses.
- **PRM (Pre-Regulator Module).** Vicor module that takes a 48V bus input and produces a regulated intermediate voltage. Handles the regulation function.
- **VTM (Voltage Transformation Module).** Vicor module that applies a fixed-ratio step-down (typically 32:1 or 48:1) using the SAC topology. Handles the transformation function.
- **VPD (Vertical Power Delivery).** Placement approach that mounts the VTM directly beneath the compute die (dropping the lateral placement), freeing the die perimeter for signal I/O.
- **DPVRM (Direct-to-Pin Voltage Regulator Module).** Vicor product family delivering regulated voltage directly to the compute die pins with minimal PDN path length.
- **PDN (Power Delivery Network).** The full electrical path from voltage regulator output to the compute die core. PDN resistance is a critical parameter at high currents.
- **VRM (Voltage Regulator Module).** Generic term for a module that regulates a supply voltage to a specific value for a load. Includes both multi-phase buck and factorised power approaches.
- **Multi-phase buck.** Conventional power delivery topology using 8-16 buck converters running in interleaved PWM to reduce ripple and increase current capability. Dominant approach for AI accelerator designs through H100 and GB200.
- **ITC (International Trade Commission).** US federal agency with authority to issue exclusion orders barring importation of infringing products. Vicor won a Final Determination in February 2025 on patents 9,516,761 and 9,166,481 against several respondents including Delta and Foxconn subsidiaries.