3×
STOCKDRIFTSRESEARCH
Three tickets to the
800-volt build-out.
In April we flagged three power-semi small caps — a binary breaker bet, a fab-for-free, and a turnaround priced as dead. All three ran: +100% to +159% at the peaks. Ten weeks later the sector is purging, and this briefing scores the call — and says what still makes sense at July prices.
→ to advance
STOCKDRIFTSRESEARCH
Executive summary
One thesis, three risk buckets: the 800 V transition is structurally short of new power silicon — and these names haven't moved yet.
01 — IPWR · $45M AT NOTE
Binary, asymmetric
800 V DC creates brand-new demand for solid-state circuit breakers — and B-TRAN is the only true bidirectional device aimed at it. Pre-revenue; $200M pipeline; CEO bought the 52-week low. Riskiest, highest upside.
JUL 2: $81M MC · +79% · peaked $9.30 · catalyst still ahead
02 — WOLF · $1.26B AT NOTE
Fab for free
Market cap ≈ cash. You got the world's only 200 mm SiC fab, 2,300+ patents and 300 mm leadership for $0 — while AI data-center revenue grew +50% QoQ and analysts still said "Reduce".
JUL 2: $40 · +28% · peaked $80.82 · alpha consumed
03 — MX · $155M AT NOTE
0.4× book pivot
Priced as a dying display company. Reality: a pure-play power-semi turnaround — 55 new products in 2025 vs 4 in 2024, buybacks not dilution, and the CEO, CFO, a director and an 8.5% activist all buying.
JUL 2: $4.12 · round trip · peaked $9.86 · cash floor intact
The playbook hasn't changed: voltage is the cheat code — crank it up early, keep it high, step down only inches from silicon.
Source: Stockdrifts Research — "Power Semi Small Caps" (Apr 2026). Market caps as of publication
blog.stockdrifts.io · 02 / 12
STOCKDRIFTSRESEARCH
Why now · the rotation
Optics ran first. Retail knows that playbook — and the same crowd is now coming for power semis.
$VICR since the Stockdrifts 800V note
Indexed, Oct 21 2025 = 1.0× · still ~5× today even after Jul 2's −20% sector purge
The physics hasn't changed
12–15 kW per pod of 8
A B200 / MI300X draws 0.9–1.4 kW each; clusters are multi-megawatt. At 12 V you'd need kilo-ampere bus-bars thicker than a wrist — losses alone would exceed the chips' TDP.
NVIDIA's own whitepaper
~45% less copper
in an 800 V plant. AC is giving way to DC microgrids — a from-scratch power stack, which means from-scratch silicon demand.
"Now that retail knows what happened to optics stocks, they are coming swinging for this sector too."
Source: Stockdrifts Research (Apr 2026); NVIDIA 800 V whitepaper copper claim; VICR move since Oct 21 2025 note
blog.stockdrifts.io · 03 / 12
STOCKDRIFTSRESEARCH
Pick 1 · IPWR · $45M MC · riskiest, highest upside
800 V DC needs a part that barely exists yet — solid-state circuit breakers — and the CEO just bought the 52-week low.
Insider conviction
$250K
CEO personal buy at $2.75 — exactly the 52-week low. Skin in the game at the bottom tick.
Sales pipeline
$200M
being chased by a $45M company. Lead-customer qualification Q4 2026, revenue ramp 2027.
IP moat
52+47
U.S. + foreign patents issued on B-TRAN, 76 more pending (as of 12/31/2025).
Runway
FY26+
Recently raised; cash to at least end of year. Capital-light — foundry partners on mainstream silicon, no exotic substrates.
- NEWThis is structural NEW demand, not share-stealing. Every 800 V DC bus needs bidirectional, high-current protection that today's breaker and FET stacks weren't built for.
- SYSDesign-win path suggests system-level attach. The Stellantis work points at sockets with long product lives once designed in — the good kind of lock-in.
The honest risk
Pre-revenue. Illiquid. Adoption could slip — and if B-TRAN isn't qualified by year-end, cash runway becomes a concern. A binary bet; the asymmetry is the point.
JUL 2: $5.04 · $81M MC — the coin is still in the air; Q4 qualification hasn't happened yet.
Source: Stockdrifts Research (Apr 2026) — IPWR investor deck; insider filing; patent counts as of 12/31/2025
blog.stockdrifts.io · 04 / 12
STOCKDRIFTSRESEARCH
The mechanism · why B-TRAN is different
One double-sided B-TRAN replaces four devices — a lane Navitas can't enter without new IP and a new process.
- ≠A fundamentally different device. A true bidirectional bipolar transistor — not a unidirectional FET or IGBT wired back-to-back.
- 4→1Fewer die, much lower conduction loss ⇒ smaller heat sinks and smaller bus-bars at module level.
- NVTSNavitas plays a different game. GaN FETs win in high-frequency conversion; in protector / contactor / SSCB roles their conduction losses and extra devices are typically worse.
- IPTo attack IPWR head-on, NVTS would need new IP and new process development — its whole portfolio is FET-based.
Source: Stockdrifts Research (Apr 2026) — B-TRAN: one double-sided device replaces two switches + two diodes in a bidirectional leg
blog.stockdrifts.io · 05 / 12
STOCKDRIFTSRESEARCH
Pick 2 · WOLF · $1.26B MC
The April alpha: Wolfspeed traded at cash — the world's only 200 mm SiC fab was priced at zero.
The market math — what $1.26B buys you
World's only 200 mm SiC wafer fabincluded — $0
2,300+ patentsincluded — $0
300 mm SiC technology leadershipincluded — $0
American fabs — supply-chain strategic assetincluded — $0
Renesas already paid up
$55–65 / share
implied by Renesas's $2B+ investment — nearly 2× the current price.
AI data-center revenue
+50% QoQ
SiC is becoming the bottleneck material of the AI power stack — a structural wafer capacity constraint.
And it played out — fast
$31 → $80.82
Analysts sat at "Reduce · $14.33" (48% below) when we published. The May 5 print forced the upgrades; the stock ran +159% in five weeks. Now $40 — EV ≈ $0.8B, no longer free.
Source: Stockdrifts Research (Apr 2026) — market cap vs cash at publication; Renesas investment; analyst consensus
blog.stockdrifts.io · 06 / 12
STOCKDRIFTSRESEARCH
Pick 2 · WOLF · what's still live at $40
The bankruptcy hangover is cured. What's left is the operating-leverage story — Mohawk Valley is still only a third full.
~30–35%
Mohawk Valley utilization
every +10% utilization drops straight to the bottom line
$6B → $2B
Debt after restructuring
emerged from Ch.11 — big progress
- SiCPhysics moat: SiC is the ONLY material that works for 800 V HVDC. There is no GaN or silicon substitute at this voltage class.
- 200Competition exists on 6-inch substrates — not on 200/300 mm. That's the honest risk, and also exactly where Wolfspeed is alone.
- 🇺🇸American fabs, 300 mm next. Supply-chain risk can make domestic SiC a strategic priority — an option the market prices at nothing.
LIVE: every +10% utilization drops straight to the bottom line
10 kV SiC MOSFET design wins
300 mm customer quals
Source: Stockdrifts Research (Apr 2026) — Mohawk Valley utilization ~30–35%; debt $6B → $2B post-restructuring
blog.stockdrifts.io · 07 / 12
STOCKDRIFTSRESEARCH
Pick 3 · MX Magnachip · $155M MC
Priced at 0.4× book as a dying display company — MX is actually a pure-play power-semi mid-pivot.
Balance-sheet floor
$103.8M cash
66% of the market cap — and they're buying back shares instead of diluting. Share count is shrinking.
Product engine restarted
55 vs 4
new products launched 2025 vs 2024 — a 14× acceleration after selling display to go all-in on power.
Margin inflection — now
9.3% → 14–16%
Q1 2026 gross-margin guidance, sequential. The inflection is starting while headline revenue still shows display runoff.
- AIThe invisible pipeline: 40 V/60 V MV MOSFETs for AI servers, IGBTs for solar/ESS, and a Hyundai Mobis automotive JV — all masked by declining display revenue.
- FETThe product is real: 0.7 mΩ Rdson · 397 A · 25% faster switching · PDFN56 — sync rectification in the final power stage, competing credibly with Infineon and Alpha & Omega.
The honest risk
Execution. Legacy pricing pressure. Power revenue must ramp faster than display runs off — against Infineon and ON Semi-sized competitors.
Source: Stockdrifts Research (Apr 2026) — MX Q1'26 guidance; product launch counts; MOSFET line specs
blog.stockdrifts.io · 08 / 12
STOCKDRIFTSRESEARCH
Pick 3 · MX · follow the buyers
Everyone with an information edge is buying MX: the CEO, the CFO, a director — and an activist.
CEO
30,000 sh
bought at $2.84
CFO
10,000 sh
bought at $2.87
Director
60,000 sh
bought in March
Byreforge LLC
13D · 8.5%
filed with activist intent — a loaded shareholder base.
The tell: volume, five days before the Q1 print
Daily volume, indexed to normal
"The volume surge today — +22.7% on 10–15× normal volume — suggests someone knows something."
Postscript, July: nobody knew anything. The Apr 28 print dumped MX −37% in two days — then a June squeeze to $9.86 round-tripped back to $4.12. The trading tell failed; the cash floor and insider case survived.
Insiders from $2.84–2.87 are still up ~45%. Chasers from the spike are not.
Source: Stockdrifts Research (Apr 2026) — insider Form 4s; Byreforge 13D; Apr 23 volume/price action
blog.stockdrifts.io · 09 / 12
STOCKDRIFTSRESEARCH
Scoreboard · Apr 24 → Jul 2
The call worked — all three ran +100% to +159% within five weeks. Then the sector started taking it back.
IPWR
+12% held
NOTE $4.50PEAK $9.30NOW $5.04
Doubled by late May on no news — the binary event is still ahead, so every dollar of this move was multiple expansion.
WOLF
+28% held
NOTE $31.23PEAK $80.82NOW $40.00
The May 5 print did exactly what we said — forced the upgrades — then the stock overshot the Renesas math and halved.
MX
−22% held
NOTE $5.25PEAK $9.86NOW $4.12
Two full round trips: the Q1 dump and a June squeeze. The only one you can re-enter below the note price.
Jul 2, one session: IPWR −10% · WOLF −11% · MX −10% · NVTS −13% · VICR −20% — the retail rotation we predicted arrived, overshot, and is now violently unwinding.
Source: Stockdrifts Research — daily closes Apr 24 → Jul 2, 2026; peaks intraday. "Held" = vs Apr 24 close
blog.stockdrifts.io · 10 / 12
STOCKDRIFTSRESEARCH
The verdict · at July prices
What still makes sense at July prices: one fresh entry, one hold, one smaller coin-flip.
MX · $4.12 · $150MTHE RE-ENTRY
Back below the noteCASH = 69% OF MC
The only one trading under its April terms after two full round trips. The cash floor ($103.8M) and buyback are intact; insiders from $2.84–2.87 never sold. The Q1 tell failed — so the new event is Q2 margin proof of the 9.3% → 14–16% guide.
Doesn't make sense if: display runoff keeps outrunning the power ramp
Doesn't make sense if: display runoff keeps outrunning the power ramp
WOLF · $40 · $2.08BHOLD — DON'T CHASE
Alpha consumedEV ≈ $0.8B NOW
The free-fab trade is over — that was April's gift. What remains is real but slower: utilization 30–35% → up, 300 mm quals, 10 kV design wins, and a price still under the Renesas-implied $55–65. A halving from $80.82 resets the froth, not the story.
Doesn't make sense if: you're buying for another 2.6× in five weeks
Doesn't make sense if: you're buying for another 2.6× in five weeks
IPWR · $5.04 · $81MSMALLER COIN-FLIP
Catalyst still aheadPRE-REVENUE
Q4 qualification hasn't happened — the asymmetry ($81M vs a $200M pipeline) survives. But the entry is ~80% above the CEO's $2.75, so the same coin-flip now pays worse odds. Size it like the lottery ticket it is, or wait for the qual headline.
Doesn't make sense if: a slip past year-end forces a dilutive raise
Doesn't make sense if: a slip past year-end forces a dilutive raise
NEXT EVENTS: MX Q2 margins · WOLF utilization print · IPWR Q4 qual
Author owns positions · not investment advice
Source: Stockdrifts Research — Apr 24 thesis re-scored on Jul 2 closes. "They are all in different risk categories, to be honest."
blog.stockdrifts.io · 11 / 12
The easy re-rate is over.
From here, it's execution.
Very risky stuff — most people should stay away.
For everyone else, three dates matter now: MX's margins, WOLF's utilization, IPWR's qualification.
For everyone else, three dates matter now: MX's margins, WOLF's utilization, IPWR's qualification.
Full research → blog.stockdrifts.io
Prequel: "High Voltage Is the New Oil"
Tell your friends — free for now
STOCKDRIFTS RESEARCH · 12 / 12