An engineer's comparison of the four solid-state cooling approaches shipping in 2026: piezo MEMS, ionic, micro-blower and DBD plasma. Strengths and limits.
Three products you can buy today have no fan inside and still move air across a hot die: the Zotac ZBOX pico PI430AJ, the Iodyne Pro Mini and the Sonim MegaConnect, all listed on Frore Systems' own "where to buy" page [7]. Two years ago that list was empty. The fan is being displaced, and not by one technology alone.
Two walls pushed this. The first is geometric: an axial fan needs a rotor, a bearing and a hub, and below roughly 5 mm of z-height there is nothing left to spin. The second is reliability: in a sealed, dusty or continuously loaded box, the bearing and the intake fail first, and no amount of fan refinement fixes that.
Solid state cooling companies in 2026 are pursuing four physically distinct mechanisms: piezoelectric MEMS membranes, electrohydrodynamic ionic wind, piezoelectric micro-blowers and surface dielectric barrier discharge plasma. They are not interchangeable. Each wins a different constraint; the job is working out which one binds your design. We cover what solid-state cooling means separately.
Both drive a thin membrane at ultrasonic frequency, rectifying the oscillation into net flow. Frore describes AirJet as "tiny membranes that vibrate at ultrasonic frequency" [2]. xMEMS builds the same idea in silicon: Tom Tarter, Principal MTS Thermal Engineer at xMEMS Labs, describes "thin silicon membranes engineered into cantilevered vanes" with a "thin-film layer of lead zirconate titanate (PZT)" [10].
Frore Systems AirJet. The AirJet Mini G2 removes 7.5 W of heat per chip, is 2.65 mm thick, weighs 7 g and runs below 21 dBA [4]. The AirJet Mini datasheet gives 27.5 x 41.5 x 2.8 mm, 11 g, 5.25 W total heat dissipation (net 4.25 W) at 85 °C die and 25 °C ambient, and 1 W maximum power [2]. Back pressure is the number that matters most: 1750 Pa across the line [1][3][4][6], enough to force air through a dense fin stack, where thin-device designs usually choke. The IP68 variant, AirJet Mini Sport, is rated for "full performance recovery after submersion in over 1.5 metres of water for 30 minutes" and has inlet dust filters [5]. The AirJet PAK family scales the same cell to 8 W, 24 W and 34 W (PAK 1C, 3C and 5C), as thin as 6.5 mm [6]. In an Intel Wildcat Lake laptop reference design, AirJet Mini modules delivered 15 W sustained in an 11.3 mm chassis at 28 dBA [8].
What it is best at: high static pressure in a thin package with a genuine ingress story. Nothing else here combines 1750 Pa with an IP68 rating.
Where it is limited: the wear mechanism is fatigue of a moving membrane and its piezo film, and Frore publishes no MTBF figure on the pages reviewed here. Airflow in CFM is also unpublished, so it cannot be compared against a fan curve.
xMEMS µCooling. The same physics at a smaller scale. xMEMS' own published figures put the XMC-2400 at roughly 25 cc/s (0.06 CFM) with maximum static pressure exceeding 1300 Pa, in a package of 7.42 x 9.48 x 1.13 mm, drawing on the order of 200 mW for full flow including the ASIC, at 13.7 dBA measured at 10 cm and a drive voltage below 20 Vpp [10]. The larger XMC-4800 reaches 48 cc/s with static pressure above 1100 Pa in a 9.93 x 14.35 x 1.13 mm package [10]. The µCooling page describes bi-directional flow adjustable up to 39 cc/s [9].
What it is best at: smallness and power budget. A 1.13 mm package drawing 200 mW is the only part here that fits smart glasses or an M.2 drive without a chassis redesign.
Where it is limited: the heat removed per part is small, and xMEMS publishes no watts-of-heat figure, only airflow and pressure. That is arguably the more honest specification, but the system modelling falls to you.
Ventiva's Ionic Cooling Engine moves ionised air by electrohydrodynamic flow with no moving parts [11]. It is the closest competitor to YPlasma in physics and the furthest in geometry. Mark MacDonald, a thermal technologist at Ventiva, describes a module approximately four millimetres tall running at around five kilovolts and very low current, with total system power for a standard 62 mm module typically under one watt, airflow of three to five metres per second, and a pressure head of typically 10 to 20 Pa [17]. Ventiva's site states a height as small as 3 mm, above 3 CFM, and below 15 dBA [11].
Commercially Ventiva has moved fast. ICE9 for up to 25 W TDP was stated available in December 2024, with a 40 W version in productisation for 2027 [13]. In May 2025 the company announced an ICE9 system delivering up to 100 W TDP [12]. At CES 2026 it showed a Compal laptop reference design using three 62 mm modules for a 28 W CPU and 44.3 W total platform power in a sub-16 mm chassis, reclaiming around 7,200 mm² of board space [14][15]. Manufacturing is contracted to Crystal Precision in Malaysia, with stated capacity up to four million units per month [16].
What it is best at: volumetric flow and board area. Ventiva is the only company here credibly claiming to replace a laptop's whole fan-and-heat-pipe subsystem rather than supplement it.
Where it is limited: pressure head. At 10 to 20 Pa [17], an ICE module cannot force air through a restrictive fin stack the way a 1750 Pa AirJet can; it needs an open duct. It also needs a kilovolt-class supply and produces ozone, which Ventiva addresses with manganese dioxide catalytic coatings and reports as sub-100 ppb exhaust concentrations compliant with regulatory limits [17].
The oldest family here. Murata's MZB1001T02 microblower is a piezoelectric diaphragm pump 20 mm square and 1.85 mm thick, delivering at least 0.70 L/min and at least 1.42 kPa static pressure at 15 Vpp, with a resonant drive frequency of 24.0 to 27.0 kHz [18][19]. Synthetic jet coolers work similarly, pulsing a cavity to entrain surrounding air.
What it is best at: maturity and availability. 1.42 kPa from a 1.85 mm device that has shipped in volume for years, through ordinary distributors, is competitive with anything newer.
Where it is limited: flow rate. 0.70 L/min is about 11.7 cc/s, well under the xMEMS parts, and the part is specified as an air pump, so there is no published thermal rating. Murata publishes no power or acoustic figure on the pages reviewed here, and in practice these serve gas sampling more than silicon cooling. We compare plasma actuators and piezoelectric fans separately.
A dielectric barrier discharge actuator puts two electrodes either side of a thin dielectric, one exposed and one encapsulated. A high-voltage AC waveform ionises a shallow layer of air at the exposed electrode edge and drags those ions along the surface, which in turn drag neutral air. The result is a wall jet of a few metres per second. The discharge is non-thermal, so the gas stays near ambient and adds no meaningful heat to what it cools. The Y-Flow actuator is roughly 200 micrometres thick.
The measured case is narrow and well documented. Go, Garimella, Fisher and Mongia superimposed an ionic wind on an existing 0.3 m/s bulk flow and obtained roughly a factor-of-two increase in the average heat transfer coefficient and about 20 K of additional cooling, drawing under 100 mW, specifically 67 mW at 15 µA [20][22]. Electrical-to-fluid conversion efficiency for corona-based ionic wind is in the region of 1 to 2 per cent [22].
What it is best at: occupying space nothing else fits in. At 200 micrometres it is a film you laminate onto the hot surface, not a component you find room for. That suits sealed enclosures, edge AI modules running continuously and the stagnant air around optical transceiver cages, where the constraint is the absence of any airflow at all.
Where it loses, plainly: static pressure is negligible, so a surface DBD actuator will not push air through a fin stack; against Frore's 1750 Pa it is not in the same contest. Bulk flow is metres per second, not tens, so a high-volumetric-flow problem still wants a fan, and efficiency as a bulk air mover is poor. It needs a kilovolt-class AC supply, generates some ozone, and the dielectric ages under continuous electrical stress.
| Frore AirJet Mini G2 | xMEMS XMC-2400 | Ventiva ICE9 | Murata MZB1001T02 | Surface DBD (Y-Flow) | |
|---|---|---|---|---|---|
| Mechanism | Ultrasonic piezo membrane | PZT silicon vanes | Corona EHD ionic wind | Piezo diaphragm pump | DBD surface ionic wind |
| Thickness | 2.65 mm [4] | 1.13 mm package [10] | 3 mm and above [11] | 1.85 mm [18] | ~0.2 mm actuator |
| Air moved | Not published | 25 cc/s [10] | Above 3 CFM [11] | 0.70 L/min at 15 Vpp [18] | Wall jet, a few m/s |
| Static pressure | 1750 Pa [4] | Above 1300 Pa [10] | 10 to 20 Pa [17] | 1.42 kPa at 15 Vpp [18] | Negligible |
| Power | Not published (1 W for AirJet Mini [2]) | ~200 mW [10] | Under 1 W per module [17] | Not published | Not published (lab: under 100 mW [20]) |
| Acoustics | Below 21 dBA [4] | 13.7 dBA at 10 cm [10] | Below 15 dBA [11] | Not published | No mechanical noise source |
| Heat removed | 7.5 W per chip [4] | Not published | Up to 100 W system TDP [12] | Not published | Not published |
| Drive voltage | Not published | Below 20 Vpp [10] | ~5 kV [17] | 10 to 20 Vp-p [18] | Kilovolt-class AC |
Claiming that surface plasma wins on every axis would be convenient. It does not, and saying so would be worth nothing to a thermal engineer.
Frore Systems has shipping consumer products, an IP68 rating and by far the highest published back pressure. Ventiva has OEM reference designs with Compal, a stated 100 W system and a contracted mass-production line. xMEMS publishes a fuller set of airflow, pressure, power and acoustic figures than most of this field discloses. All three are further along commercially than YPlasma, which is early stage with product in development rather than in volume.
What surface DBD has that none of them have is thickness measured in micrometres and nothing that moves, vibrates or wears mechanically. That is not a general-purpose advantage. It is decisive only where there is no room for a component and no vent to put it behind. If that describes your problem, talk to us. If it does not, one of the other three is probably the better answer, and we would rather say so.
What is the best AirJet alternative in 2026? It depends on the binding constraint. For a thinner and lower-power part, xMEMS µCooling at 1.13 mm and roughly 200 mW [10]. For replacing a whole laptop thermal subsystem, Ventiva's Ionic Cooling Engine [14]. For a sealed enclosure with no space for any component, a surface DBD plasma actuator. No alternative matches AirJet's published 1750 Pa of back pressure [4].
Is Ventiva's Ionic Cooling Engine the same thing as a DBD plasma actuator? They share a physical family, electrohydrodynamic ionic wind, but not a geometry. Ventiva builds a discrete module described as approximately four millimetres tall that moves air through a duct at three to five metres per second with a 10 to 20 Pa pressure head [17]. A surface DBD actuator is a roughly 200 micrometre film that creates a wall jet along the surface it is bonded to, with essentially no static pressure.
How thin can solid-state cooling actually get? The thinnest packaged component with published figures is the xMEMS XMC-2400 at 1.13 mm [10]. Frore's AirJet Mini Slim and Mini G2 are 2.65 mm [3][4]. A surface DBD actuator is around 200 micrometres, but it is a laminate rather than a drop-in component and it moves far less air.
Does ionic cooling produce ozone, and is that a problem? Any discharge in air produces some ozone. Ventiva reports sub-100 ppb exhaust concentrations using manganese dioxide catalytic coatings, which it describes as compliant with regulatory limits [17]. For DBD the quantity depends on waveform, geometry and duty cycle, and is manageable by design, but it is a legitimate question that deserves measurement rather than dismissal.
Which of these is furthest along commercially? Frore Systems, by the test of products you can buy: the Zotac ZBOX pico PI430AJ, the Iodyne Pro Mini and the Sonim MegaConnect [7]. Ventiva has the largest stated manufacturing capacity at up to four million units a month [16] and OEM reference designs [14]. YPlasma is earlier than both.