Vicor filed its FY2025 10-K on March 2. The stock was at $201. Five days later it's $162, down 19%. The filing was bullish. The CEO started selling the same day.

This is a company that makes genuinely differentiated power delivery components for AI datacenters. It also has the most misleading income statement in the semiconductor sector and a founder who just filed to sell 3% of the float while telling you the best is ahead.

Let's decompose.

The Product Is Real

Gen 5 Vertical Power Delivery (VPD) delivers current vertically through the substrate into the processor at 5 A/mm², versus the industry standard lateral delivery at 1.5 A/mm². Next-gen GPU and XPU roadmaps require >3 A/mm². If you're building AI accelerators at scale, VICR's Gen 5 is — today — the only production-ready solution that meets spec.

This isn't hand-waving. MPWR's 10-K (filed Feb 27) doesn't mention VPD at all. On MPWR's Q2 2024 call, an analyst asked CEO Michael Hsing directly about vertical vs lateral power delivery. He deflected. If MPWR had a competing VPD product, they'd say so.

Fab 1 in Andover can generate "$slightly above $1 billion" at full utilization. It's currently running at roughly 35%. Book-bill exceeded 1.2x in Q4 and continued higher into Q1 2026. Backlog rose to $177M from $155M. Content per AI accelerator is estimated at $200-$400/unit.

These are real demand signals at a company with a real technical moat.

The Earnings Are Fake

Not fake-fake. Legally accurate. But wildly misleading.

Reported FY2025 net income: $118.6M. Diluted EPS: $2.61.

Now strip out two one-time items:

• $45M patent litigation settlement received in Q2
• $24M tax benefit from deferred tax asset valuation allowance release

Normalized net income: ≈$54M. Normalized EPS: ≈$1.19.

The reported P&L makes VICR look 2.2x more profitable than it actually is. At $162.67 today, the stock trades at ≈39x normalized forward earnings, not the ≈62x trailing that shows up on a screener. Neither number screams "cheap" for a company whose CEO explicitly refuses to provide quarterly guidance due to "uncertainty of timing."

Reported gross margin: 57.3%. Strip the settlement: ≈52.6%. Still an improvement from 51.2% prior year, but not the 600bps expansion the headline suggests.

The Monopoly Has a Clock On It

Here's where the bull case starts cracking.

ADI (Analog Devices, $40B market cap) — shipped its first vertical power product in Q4 2025. Across three consecutive earnings calls (Q1 2025, Q4 2025, Q1 2026), ADI CEO described VPD as at the "knee of the curve" with "exponential growth." ADI has existing hyperscaler relationships and manufacturing scale VICR cannot match.

TDK — launched stackable µPOL modules for vertical power delivery on February 5, 2026.

Renesas — partnered with Flex Power Modules on VPD development, reportedly gaining allocation on NVIDIA's Blackwell platform, and supporting NVIDIA's 800V DC architecture with GaN solutions.

That's three entrants in under six months. VICR's VP Sales said "I'm not hearing anybody in volume" on the Q3 call. By Q1 2026, ADI had shipped. The monopoly window isn't 12-18 months — it's more like 6-12, and that's optimistic.

VICR CEO Vinciarelli's response? "They acknowledge they're out of luck." That's a 75-year-old founder who built something brilliant and can't see the competitive response forming. History is littered with technology monopolists who confused first-mover advantage with permanent moat.

But the IP Business Is Separate — And It's Interesting

The most underappreciated part of the VICR story: the ITC enforcement strategy targets bus converter patents (NBMs), NOT VPD specifically. This means the IP licensing revenue is partially independent from the VPD product monopoly.

Royalty trajectory: $15.9M (2023) → $46.6M (2024) → $57.4M (2025). 90% CAGR.

ITC Investigation #1 (completed): Limited Exclusion Order issued against Delta, Cyntec, Foxconn, Quanta. Importation ban on infringing bus converters.

ITC Investigation #2 (instituted Feb 17, 2026): Expanding scope to Luxshare and MetaPWR — Chinese contract manufacturers. Broader net, more licensing leverage.

CEO's "$300M" target is cumulative contribution through 2026, not annual run rate. Still, the mechanism is real: hyperscalers deploying AI systems at scale are using VICR-patented power distribution topology. The ITC exclusion order threat forces licensing negotiations. The $45M settlement from a single licensee shows deal magnitude.

The analog risk is Apple vs Masimo: Apple redesigned the Watch to avoid the ITC exclusion order. Hyperscaler respondents (Delta, Foxconn, Luxshare) have the engineering resources to design around the patents. But that takes time and money, and in the interim, licensing revenue accrues.

If you're going to be bullish on VICR, this is the factor to be bullish on — not the VPD product monopoly.

The Insiders Know Something (Or They Don't Care)

CEO Vinciarelli adopted a 10b5-1 plan on March 2 — the same day the bullish 10-K was filed — to sell up to 1,000,000 shares through December 2027. That's 2.97% of the common float.

But it's not just the CEO. CFO Schmidt: 35,224 shares. Director Eichten: 40,587. VP Sales Davies: 42,289. Twenty-eight insider selling transactions in the past three months. Six-plus insiders sold $14.7M in a 10-day window at $195-$210 near the all-time high.

The CEO retains 9.83M shares (99.5% of his position). This isn't "heading for the exits." It's liquidity management by a 75-year-old founder at record prices. But the breadth and coordination tell you something: nobody on the inside is adding.

The market noticed. DailyPolitical attributed 9.7% of the 19.2% weekly decline specifically to insider selling. The options market is bearish — put/call volume ratio at 2.12, put implied volatility elevated 3.7% above calls (unusual inversion).

Factor Decomposition

Single-factor (SPX only): 80% idiosyncratic. Looks good on paper.

But VICR doesn't trade in a vacuum. ADI fell 11% the same week. MPWR fell 10.5%. Strip semiconductor sector beta and the AI infrastructure theme, and true company-specific variance drops to 55-65% — below the 75% threshold where we'd want to see returns driven by stock-specific insight rather than sector waves.

The +173% 1Y return is partially semiconductor sector beta dressed as stock-specific alpha. Don't confuse riding the AI wave with having a moat.

Within the idio component, four independent drivers:

Edge-weighted alpha calculates to ≈1.7% annualized. That's noise, not signal.

Scenarios

Probability-weighted EV: $156. Current price: $163. The stock is 4% rich to fair value.

The asymmetry is telling. Bull case requires everything to go right: VPD monopoly holds, fab ramp executes on schedule, IP enforcement scales, ADI stumbles. Bear case only needs one thing to go wrong: the unnamed lead customer delays, or ADI qualifies at a hyperscaler, or the fab ramp takes 18 months instead of 12.

The Fortress Balance Sheet Is Real But Irrelevant

$402M cash, zero debt, can self-fund a second fab ($250-300M). This is genuinely strong. But balance sheet strength doesn't create alpha when the stock already trades at 39x normalized earnings. The balance sheet is a survival factor, not a return factor. It means VICR won't go bankrupt. It doesn't mean the stock goes up.

Bottom Line

VICR has a real technology (Gen 5 VPD), a real IP enforcement strategy (ITC), and a real demand tailwind (AI datacenter power). The bull case exists and it's not stupid.

But: the earnings are 2.2x overstated by one-timers, three competitors entered VPD in six months, the CEO filed to sell 3% of the float on the same day as the bullish 10-K, management won't guide on the ramp they're supposedly confident about, and the stock trades at 39x normalized earnings.

The one factor where signal is clearest — insider selling — is bearish. The one factor that's most defensible — IP licensing — is the one analysts talk about least.

Probability-weighted EV is $156 against a $163 stock. Fairly priced at best, slightly rich at worst. At $163, the margin of safety doesn't exist. The execution confirmation point is Q1 product revenue >$100M (Apr 28 earnings). Below $120-130, the risk-reward shifts.

Three predictions recorded for calibration:

• Q1 2026 product revenue >$100M: 55%
• VPD monopoly eroded by year-end 2026: 75%
• FY2026 royalty revenue >$80M: 60%