Vicor's $197 Million Quarter and the $43 Million Question

Vicor's fiscal third quarter was, by any mechanical measure, a monster. Product and royalty revenue of $197 million rose 37.5% sequentially and 78.5% year over year; consolidated gross margin hit 68.3%, up more than a thousand basis points from the year-ago quarter; GAAP diluted EPS came in at $1.33 on 47.6 million shares. Royalty revenue alone was $66.6 million, up 119.1% sequentially. Book-to-bill cleared 1.1 and, by Philip Davies's account, the company booked $100 million in the first two weeks of the fourth quarter alone.

And yet the single most important number on the call was a guidance figure roughly a third smaller than the quarter just delivered.

The $43 million tell

Jim Schmidt's Q4 framing: 10% product revenue growth, and “at least $43 million in royalty revenue assuming no new license agreements are closed this quarter” — 2026-10-20. Against a $66.6 million actual, that is a ~35% sequential decline in the highest-margin line in the business — the line that produced essentially all of the gross-margin expansion.

Management anticipated the objection and attacked the framing rather than the arithmetic. “It is better for, I think, the audience and the analyst to think about the royalty stream as recurring. Now. And not a nonrecurring activity” — 2026-10-20, Davies said. Patrizio Vinciarelli pushed harder: “that which was labeled nonrecurring has been recurring pretty periodically” — 2026-10-20. The licensing practice is now anchored to royalty rate stages that escalate roughly 5% per month after an ITC investigation is instituted — stage 2, stage 3, and beyond — which is a genuinely clever mechanism: it costs late adopters more to wait than to sign.

But "recurring" is doing a lot of work here, and it is a change of register, not just of vocabulary. In July, Schmidt walked analysts through the mechanics of the prior license in quarters, not run-rates: $15 million recognized in Q2, then “$5 million in Q3 and $10 million per quarter for the following four quarters” — Jim Schmidt, Chief Financial Officer · 2026-07-21. That is the shape of a contract, not an annuity. Today's VPD license is a different animal — a per-unit inclusive license struck at stage 2 rates — but the transition quarter is still, definitionally, a catch-up quarter, and the guide says so.

The margin corollary matters too: royalty is not clean. Legal expense jumped $11.1 million to $16.1 million, and Schmidt flagged a run-rate of roughly 20% of royalty revenue. Vinciarelli has been explicit that litigation partners take a share, though “subject to caps and so on and so forth” — Patrizio Vinciarelli, Chief Executive Officer · 2026-04-21. So the incremental royalty dollar lands at something closer to 75–80 cents than 100.

From optionality to commitment

The genuinely new fact this quarter sits on the balance sheet, not the income statement. After two quarters of hedging, Vicor bought both sites: Fab 2 in Merrimack and land for Fab 3. The company claims a near-5x capacity expansion, calls that conservative, and plans initial equipment deployment in Fab 2 in the second quarter of 2027. Compare to July, when the answer was “We have several options at this point. We made some offers. None of them was taken up yet... We'll probably be making decisions in the next few weeks.” — Patrizio Vinciarelli, Chief Executive Officer · 2026-07-21 They decided.

What is striking is how the capacity conversation has changed in character. Fab 1 was long described as a $1 billion facility, then $1.5 billion with incremental expansion. Now the constraint language is about site geometry — “unlike Fab 1, we are not going to be limited by” — 2026-10-20 plain elbow room “within either 1 of these sites” — 2026-10-20. Vinciarelli also revealed the backlog is far larger than the prepared remarks implied, noting that “as of yesterday, the backlog is actually closer to $400 million” — 2026-10-20 — versus the $46 million figure in Schmidt's script — and said he expects to fund both fabs without raising equity. February's estimate for one new facility was “of the order of $250 million, $300 million” — Patrizio Vinciarelli, Chief Executive Officer · 2026-02-20; now two sites are bought and the number is deferred to the next call.

Both sites have significant upside, that could support even more than a 5x expansion.

The tape disagrees, loudly

Here is where the story turns from good to interesting. Revenue ran at $143 million in the June quarter, and gross margin stood at 58.0% — the step-function to 68.3% is licensing mix, full stop. Meanwhile price-to-revenue of 25.7x says the market has already capitalized a licensing future it has not yet seen repeated.

The tape has stopped cooperating. VICR peaked at $379.78 on June 30 and trades 46.2% below it, down 25.9% over 90 days — including a violent 9-day +58% surge in September (the guidance raise) followed immediately by a 29% two-week give-back. The broader theme is sagging too: AI data center is a powerful 360-day advancer but a 90-day and 30-day decliner, with 47 names down versus 6 up in the last month. co packaged optics, High Bandwidth Memory and PCIe Gen 6 have all rolled over across the same window. Vicor's blowout is landing into a decaying hardware tape.

There is one clean thematic confluence worth flagging: 800V DC rack distribution is a rising global keyword, and Vinciarelli spent real call time arguing the industry is solving the wrong problem — “The older focus about 800-volt is really missing the bullet with respect to the AI challenge” — 2026-10-20. That is a point-of-load argument against the entire rack-architecture narrative the market is currently paying for. It is either the most valuable contrarian insight on the call or the most expensive one.

What is actually being paid for

Two quiet shifts are worth naming. First, lead customer fell off the company's keyword map this quarter after dominating discussion for over a year; the language migrated to licensees, hyperscalers, and the ITC. Concentration risk is being reframed as enforcement leverage. Second, capital is going to work: capex was $11 million last quarter, with $48.6 million still committed on construction in progress and two fabs left to equip. R&D intensity stays modest at roughly 15% of revenue, which is the point: the IP is a sunk cost now harvesting.

The honest read is that Vicor has built two businesses with opposite risk profiles and is asking investors to underwrite the smoother one while reporting the lumpier one. The $43 million baseline is the number to hold them to: if Q4 royalty lands meaningfully above it without a new deal signing, the "recurring" claim earns its keep. If it lands at $43 million, the market will treat the $66.6 million quarter as the peak of a cycle rather than the floor of a run-rate — and at 90x trailing earnings, it will not be forgiving about the difference.