Vishay Precision Group, Inc. (VPG)

A sleepy precision-sensor company that the market has suddenly decided is a humanoid-robotics pure-play — the story is real, the chart is not investable here.

Investment Research · Photoncap-style deep dive · v1 of "VPG" · May 14, 2026

What Vishay Precision Group physically does

Vishay Precision Group makes the components that let a machine know how hard it is pushing, how much it is carrying, and where its joints are. The core technology is the foil strain gauge — a microscopically thin metal pattern bonded to a flexure that changes electrical resistance by a few parts per million when the flexure deforms under load. VPG has made these for decades through its Micro-Measurements brand, and it converts them into finished sensing products: load cells (the Tedea-Huntleigh, Sensortronics, Revere and Celtron brands that go into industrial scales and weigh-bridges), precision resistors (the Foil Resistors business, spun out of the original Vishay heritage), and increasingly force/torque and tactile sensing modules aimed at automation. The company reports in three segments — Sensors, Weighing Solutions and Measurement Systems — and the strain-gauge die sits underneath essentially all of it.

Why this matters for robotics: a humanoid robot has two perception problems. It needs to see (machine vision, covered elsewhere in this batch), and it needs to feel. Feeling is the harder, less-solved problem. To pick up an egg without crushing it, to insert a connector, to walk over uneven ground without falling, a robot needs force/torque sensing at the wrist and ankle and tactile sensing at the fingertips. A six-axis force/torque sensor at the wrist tells the controller the full vector of contact force; a tactile array on the fingertip gives a pressure map across the contact patch. VPG's strain-gauge expertise maps directly onto both — the company has publicly described its push into "fingertip tactile sensing" and a broader "strategic shift to robotics" (Sidoti conference, early 2026).

The honest framing is that VPG is not yet a robotics company. It is a ~$307M-revenue precision-sensing company with a small, fast-growing robotics order book bolted onto its Sensors segment. The strain-gauge IP is genuinely differentiated — accuracy, drift, hysteresis and temperature stability are hard-won and VPG is among the best in the world at it — but the question for the trade is whether a few million dollars of humanoid orders justifies a stock that has nearly doubled off its 50-day average.

Product roadmap

VPG's robotics push runs through the Sensors segment rather than a single named product line, which makes the roadmap less crisp than a semiconductor-equipment name — but the company has been specific in its 2026 disclosures. The foundation is the established force-sensor and load-cell portfolio (the VPG Force Sensors brands — Tedea-Huntleigh, Sensortronics, Revere, Celtron) plus the Micro-Measurements strain-gauge catalogue, both of which predate the robotics story by decades and continue to serve industrial weighing, avionics and test-and-measurement.

The robotics-specific work is newer. Through 2025 VPG began shipping custom force/torque and tactile sensing modules to humanoid developers, and on the Q1 FY2026 call (May 12, 2026) management quantified the pipeline concretely: VPG has shipped to its first two humanoid customers, taken an initial prototype order from a third, and is in early discussions with a fourth — described as a startup in defense and industrial applications. The second customer placed an order of roughly $1.5M in late 2025, primarily for tactile sensing. Q4 FY2025 humanoid-related bookings were about $0.8M. The company says it has already shipped sensors covering "a few dozen robots," each robot using roughly 20–30 sensors, and expects a follow-on order for hundreds of units in Q2 FY2026, with the potential to ramp to "hundreds of robots per week" by the end of 2026 — that last figure is management aspiration, not booked backlog, and should be read as such.

What VPG does not make is the robot, the actuator, the vision stack or the controller. It is a component supplier, and even within sensing it is a tactile/force specialist — it does not make the magnetic joint-position sensors that Allegro and others supply. The financial model management has put forward assumes a 50% annual revenue growth rate for the humanoid business off a 2025 baseline, which gets the robotics line to "low tens of millions" — confirmed as a management framework, not consensus.

The financial print

VPG closed FY2025 (year ended December 31, 2025) with revenue of $307.2 million, essentially flat at +0.2% year-over-year, and an operating margin of just 4.5%, down from 5.5% in 2024 — results released February 11, 2026. Full-year net earnings were $5.3 million, or $0.40 per diluted share, down from $0.74 in 2024; adjusted EBITDA was $28.2 million at a 9.2% margin; gross margin slipped to 38.9% from 41.0%. This is the uncomfortable part of the story: the core business spent 2025 going sideways at thin and falling margins, with Q4 FY2025 operating margin compressing to 2.3% on unfavorable mix, inventory reductions and higher fixed and logistics costs.

The Q1 FY2026 print on May 12, 2026 is what changed the narrative. Revenue was $84.4 million, up 17.6% year-over-year and roughly 8.3% above the $77.9M consensus, with Sensors segment revenue of $33.3 million up 23.1%. The headline number, though, was bookings: $102.1 million, up 25.5% sequentially, the third-highest quarterly bookings in company history, lifting book-to-bill to 1.21 and backlog to about $125 million. Sensors booked at a 1.36 book-to-bill. Adjusted EPS was still only $0.07 — the orders are running well ahead of the P&L, which is the bull case and the valuation problem in one sentence. Forward consensus for FY2026 revenue sits around $326–329 million per compiled sell-side estimates, up from ~$320M ninety days ago. At $102.94 the stock trades on a forward P/E of 65.8 with a market cap of $1.37B — a multiple that only makes sense if you underwrite the humanoid ramp aggressively. The 1-year return into mid-May 2026 is well over +150%, with the stock up roughly 57% on the Q1 print alone. The next binary is Q2 FY2026 earnings, expected around August 11, 2026.

Customer mix today

VPG's revenue splits by segment rather than by named customer, and the company does not disclose customer concentration the way a single-product supplier would. On a reporting basis, Sensors is roughly 40% of revenue, Weighing Solutions roughly 35%, and Measurement Systems roughly 25% — these are approximate proportions derived from segment disclosure, not audited customer percentages. The end-market spread inside that is wide: semiconductor equipment, data centers, avionics, military and space, industrial automation, precision weighing and test-and-measurement. Q1 FY2026's bookings strength was attributed specifically to semiconductor equipment, data centers, avionics, military/space and select industrial markets — robotics was a contributor but not the largest one.

The structural shift the bulls are paying for is the humanoid line going from essentially zero in 2024 to a low-single-digit-million run-rate in 2025 to management's "low tens of millions" framework for 2026. Put in proportion: even at the optimistic end, humanoid robotics is plausibly 5–10% of total revenue in 2026 — real, growing fast, but not yet the company. The 2024-versus-2026 change that matters is less the customer mix than the orders mix: a flat-revenue, thin-margin industrial company is suddenly booking at 1.21x with a humanoid pipeline of four named developers. Whether that converts to durable revenue at acceptable margin is the entire question, and Q1 FY2026 did not yet answer it on the P&L.

What's actually happening at the humanoid developers

VPG has not named its humanoid customers, so this section is necessarily built from the company's own characterizations on the Q1 and Q4 calls rather than from customer-side confirmation. The picture management painted: customer one and customer two are both past prototype and into follow-on ordering — customer two's ~$1.5M late-2025 order was "primarily for tactile sensing," which tells you VPG is winning the fingertip-sensing socket, not just the wrist force/torque socket. Customer three placed an initial prototype order. Customer four is in early discussions and is described as a defense/industrial robotics startup, which is a slightly different profile than a consumer-humanoid OEM.

The mechanism that matters is sensor count per robot. Management's "20–30 sensors per robot" figure is the leverage: if the humanoid market moves from prototype fleets of dozens to early-production fleets of hundreds, VPG's unit volume scales 10–20x without winning a single new customer. The Q2 FY2026 "follow-on order for hundreds of units" that management flagged is the near-term proof point — if it lands, the 50%-CAGR framework gets more credible; if it slips, the multiple is indefensible. The thing to watch with genuine skepticism is that every humanoid OEM is simultaneously evaluating whether to build tactile sensing in-house, and several (Tesla Optimus most prominently) have shown a strong vertical-integration instinct. VPG winning prototype sockets is not the same as VPG holding production sockets at scale.

The competitive threat

VPG's competitive set splits by where you look. In its legacy weighing and force-measurement business, MarketBeat's competitor screen lists Keysight, Teledyne, Allient, Diversified Technical Systems and Teradyne among ~17 named competitors — a fragmented, mature field where VPG competes on accuracy and brand. In the robotics-specific force/torque and tactile niche, the more relevant competitors are the dedicated robot-sensing specialists: ATI Industrial Automation (owned by Novanta), Bota Systems, OnRobot, and a wave of Chinese tactile-sensing startups, plus the humanoid OEMs' own internal efforts.

There is no active IP litigation to flag. The competitive risk here is not a courtroom; it is two slower-moving forces. First, commoditization — fingertip tactile sensing is an area of intense academic and startup activity, and a foil-strain-gauge approach is not the only architecture (capacitive, optical/vision-based and MEMS tactile arrays all have credible proponents). VPG's edge is manufacturing maturity and sensor-grade accuracy, but a "good enough and cheap" Chinese array could undercut it at the volumes that matter. Second, the OEMs themselves: a humanoid developer that decides tactile sensing is core IP can design VPG out. The bull rebuttal is that VPG's strain-gauge process is genuinely hard to replicate and that OEMs racing to production would rather buy than build — but that is an argument, not a moat you can measure yet.

The terminal risk

The terminal risk for VPG is that the humanoid tactile-sensing socket it is winning today is a prototype-phase socket, not a production-phase one. The structural transition that would obsolete the current VPG approach is twofold. The first path is vertical integration: if the humanoid winners — and this is a winner-take-most market — bring tactile and force sensing in-house the way Tesla has signaled across much of its actuator and sensor stack, VPG's design wins evaporate at exactly the moment volume arrives. The second path is architectural: foil strain gauges are a mature, ~80-year-old technology, and the next generation of tactile sensing may be vision-based (a camera looking at a deformable gel fingertip, the GelSight/MIT lineage) or MEMS-based. If the industry standardizes on a non-strain-gauge architecture for fingertips, VPG's incumbency in industrial load cells does not transfer.

VPG does have a credible near-term roadmap — it is shipping, iterating with four named developers, and its accuracy advantage is real. But the multiple the market is now paying (66x forward earnings) implicitly assumes VPG both scales with the humanoid market and holds its sockets through the production transition. Neither is proven, and the terminal risk is precisely that the company is a prototype-era supplier being priced as a production-era one. This is the constraint on how much you can pay — and at current levels you are paying a lot.

Bull / Gap / Optionality (Photoncap framing)

Bull

1. The order book has genuinely inflected. Q1 FY2026 bookings of $102.1M, up 25.5% sequentially, were the third-highest in company history, with book-to-bill at 1.21 and backlog at ~$125M — and Sensors specifically booked at 1.36 (company release, May 12, 2026). This is not a robotics-only story; it is broad strength across semi equipment, data centers and avionics. Even setting humanoid aside, the core business has visibly turned, which de-risks the floor.

2. The humanoid pipeline is real, named and widening. VPG has shipped to two humanoid customers, has a prototype order from a third, and is in discussions with a fourth (Q1 FY2026 call). Customer two's ~$1.5M order was "primarily for tactile sensing" — VPG is winning the hardest, highest-value sensing socket, not the commodity one. Four developers in the pipeline at this early stage is a credible spread of bets.

3. Sensor-count-per-robot is the operating leverage. At management's stated 20–30 sensors per robot, VPG's unit volume scales 10–20x as the humanoid market moves from prototype dozens to production hundreds — with no new customer wins required. Management's 50%-CAGR humanoid framework off a 2025 baseline (Q1 FY2026 call) gets the line to "low tens of millions"; the Q2 "hundreds of units" follow-on order is the near-term test.

4. Operating leverage on the legacy business is dormant but real. FY2025 operating margin was a thin 4.5%, depressed by mix and inventory actions — but VPG ran 41% gross margin as recently as 2024. A bookings-driven revenue recovery against a fixed cost base should expand margins meaningfully through 2026, which is why FY2026 consensus revenue of ~$326–329M matters more than the trailing print.

5. Cross-theme relevance to Photonics (Theme 3). VPG's precision-resistor and strain-gauge content also feeds optical and photonics instrumentation, giving it a second secular tailwind beyond robotics — a modest diversification of the growth story that the pure-play framing understates.

Gap

1. The valuation has fully detached from the P&L. Q1 FY2026 adjusted EPS was $0.07 — unchanged year-over-year — while the stock trades at 65.8x forward earnings and jumped ~57% on the print. The orders are running far ahead of earnings, and the market is paying a production-scale multiple for a company still earning prototype-scale profits. Any wobble in the ramp story re-rates this hard.

2. The chart is one of the most overbought in the entire robotics universe. RSI 90.5 and +94.6% above the 50-day moving average is a statistically extreme reading — the stock has nearly doubled relative to its own recent trend. Even in a secular theme, mean reversion from a 90+ RSI is the base case; the question is not if but how deep.

3. The core business margin structure is weak. FY2025 operating margin of 4.5% and Q4 FY2025 of 2.3% show a company with little cushion. Humanoid revenue at "low tens of millions" is not yet enough to move consolidated margin, and if industrial/semi bookings normalize before humanoid scales, VPG is back to being a thin-margin industrial at a rich multiple.

4. Customer concentration and disclosure opacity. VPG does not name its humanoid customers or disclose their order cadence in detail, so the bull case rests heavily on management characterization. With a winner-take-most humanoid market, losing even one of the two shipping customers — to vertical integration or to a competitor — would be material and the market would learn about it late.

Optionality

The trade

VPG is the rare case where the thesis and the timing point in opposite directions, and price discipline has to win. The two-layer framework: Layer one — is the structural thesis real? Yes, qualifiedly. VPG has a genuine, widening humanoid tactile-sensing order book, the sensor-count-per-robot leverage is real, and the core business has visibly inflected on bookings. This is an early-cycle robotics-sensing name worth owning at some price. Layer two — is the entry timing acceptable? No. RSI 90.5 and +94.6% versus the 50-day average is among the most extended readings in the entire robotics universe, on a stock that just gapped ~57% on a print where EPS did not move. Paying 66x forward earnings into a 90+ RSI is buying the narrative at the moment of maximum narrative. The discipline is to put VPG on the sheet as a Bucket C name — significantly extended, chase risk explicit — and wait. The technically-honest entry zone is current ±5%, i.e. roughly $97.79–$108.09, but that should be read as where you would be a forced buyer only after a meaningful RSI reset, not where you initiate today; a more realistic accumulation zone is the post-gap consolidation back toward the low-to-mid $80s if the tape gives it. Size small — 0.25–0.50% of risk capital — precisely because the entry is poor and the position is a toe-hold on a structural theme, not a high-conviction initiation. Stop at roughly $78, below the pre-gap structural base; a break there says the bookings inflection was a one-quarter event. The named catalyst is Q2 FY2026 earnings around August 11, 2026, where the "hundreds of units" humanoid follow-on order either confirms the ramp or doesn't. If you want a cleaner expression of the robotics-sensing thesis without VPG's valuation and chart extension, Allegro (ALGM) — also in this batch — offers humanoid joint-sensing exposure at 31x forward earnings on a far less stretched chart. Conviction: 5 / 10.

Sources referenced inline throughout. Reference v1 of this template format: _Watchlist/hanmi-photoncap-style.md.

\newpage

ARBE — Arbe Robotics Ltd. · SKIP / WAIT (Tier-3) · Conv 3/10 · Bucket C

ticker: ARBE name: Arbe Robotics Ltd. theme: Robotics bucket: C conviction: 3 entryzonelo: 1.12 entryzonehi: 1.24 currentprice: 1.18 pricedate: 2026-05-14 positionsizepct: 0.4 stoploss: 0.78 thesisoneline: Pre-revenue 4D imaging-radar chip company with a real technology lead, a 2026 production target it has repeatedly missed before, and a runway measured in quarters. catalystnext: Q1 26 earnings catalystdate: 2026-05-28 rsi: 75.6 vs50ma: 55.0 forwardpe: 0.0 themecycleposition: early customermixsummary: Effectively pre-revenue — 2025 revenue ~$1.0M; engaged with ~15 OEMs / 11 at bid stage, one top-10 automaker chipset selection, no production contracts. terminalriskoneline: Tier-1 silicon incumbents (TI, NXP, Infineon) and Mobileye close the imaging-radar resolution gap with the balance sheets to outlast a cash-strapped micro-cap. bulldriverscount: 5 gapriskscount: 4 optionalitycount: 5 lastearningsdate: 2026-02-26 nextearnings_date: 2026-05-28