Conferencia de resultados del T3 de 2026 de Navitas Semiconductor (NVTS): la infraestructura de IA impulsa el crecimiento

Puntos clave

• Los ingresos del tercer trimestre de 2026 alcanzaron los 14,0 millones de dólares, lo que supone un aumento del 33% intertrimestral y del 38% interanual. Se trata del primer trimestre de crecimiento interanual de Navitas Semiconductor desde 2024.
• Los ingresos procedentes de aplicaciones de alta potencia aumentaron más del 50% intertrimestral y registraron un crecimiento de tres dígitos interanual. La dirección prevé que los ingresos procedentes de dispositivos móviles y del sector de consumo de gama baja pasen a ser insignificantes para finales de año.
• El margen bruto no GAAP se amplió en 50 puntos básicos en términos intertrimestrales hasta el 40,0%, lo que marca el tercer trimestre consecutivo de mejora.
• Navitas registró una ratio book-to-bill de aproximadamente 2 en el tercer trimestre y una cartera de pedidos que se extiende hasta mediados de 2027. La dirección prevé que los ingresos procedentes de infraestructuras de IA más que se tripliquen en el año natural 2027.
• Las previsiones de ingresos para el cuarto trimestre se sitúan en 16,0 millones de dólares, más o menos 0,5 millones de dólares, lo que representa un crecimiento intertrimestral del 14% y un crecimiento interanual del 119% en el punto medio.
• La adquisición de Claros añade tecnología de regulador de voltaje integrado (IVR, por sus siglas en inglés) a la cartera de GaN y SiC de Navitas, lo que extiende su plataforma de potencia desde la red eléctrica hasta la XPU.

Resultados financieros principales

Rendimiento empresarial y operativo

La transición de Navitas para alejarse de los productos móviles y de consumo de gama más baja se encuentra cerca de su finalización. Los mercados de alta potencia explicaron el crecimiento del trimestre, mientras que la dirección prevé que casi la totalidad de los ingresos provenga de aplicaciones de alta potencia para finales de 2026.

Se espera que la infraestructura de IA —incluidos los centros de datos de IA y los sistemas de red eléctrica y energía que los respaldan— supere el 40% de los ingresos totales al cierre de 2026. La dirección atribuyó esta dinámica al aumento de las exigencias de densidad de potencia en fuentes de alimentación de CA a CC, conversión de CC a CC, unidades de alimentación ininterrumpida por batería y futuras arquitecturas de 800 voltios.

La empresa detalló cuatro puntos de inflexión en la infraestructura de IA. La sustitución del silicio por SiC de 1,2 kV en sistemas CA-CC de mayor potencia ya está contribuyendo a los ingresos. Los programas vinculados a arquitecturas de mayor voltaje para sidecar, CA-CC, CC-CC y respaldo por batería están entrando en la cartera de pedidos para 2027. Navitas también está enviando muestras de más de una docena de programas nativos de 800 voltios y ha registrado un pedido de un millón de unidades de GaN para un diseño de potencia de CC a CC de 800 voltios. Las oportunidades en SiC de ultraalto voltaje para transformadores de estado sólido se posicionan principalmente como un motor a partir de 2028.

La plataforma IVR de Claros combina control digital, transistores de potencia, condensadores, inductores integrados y empaquetado. La dirección afirmó que su arquitectura modular permite respaldar el suministro de energía vertical e integrado para diferentes procesadores y necesidades de potencia. Se prevé que un cliente principal inicie un incremento significativo de la producción a finales de 2028 y durante 2029.

Navitas también destacó su colaboración con NVIDIA en una placa de distribución de energía de 800 a 12 voltios, su licencia de tecnología GeneSiC y el trabajo de diseños de referencia con Microchip, una subvención del Gobierno de Estados Unidos para desarrollar semiconductores de SiC de 10 kV y los envíos iniciales de obleas GaNFast de quinta generación fabricadas en Estados Unidos a través de GlobalFoundries.

Previsiones de la dirección

El incremento previsto de los gastos operativos en el T4 refleja los costes generales vinculados a Claros y la inversión adicional en I+D y capacidad operativa. Tras este repunte inicial, la dirección espera que los gastos operativos aumenten a una fracción del crecimiento de los ingresos.

De cara a 2027, la dirección prevé que la infraestructura de IA impulse un crecimiento intertrimestral de los ingresos de dos dígitos y señaló que este crecimiento podría acelerarse a medida que se solapen las sucesivas transiciones de arquitectura. La hipótesis de planificación de la empresa asume que los ingresos no vinculados a infraestructuras de IA se mantendrán prácticamente estables, mientras que los ingresos de infraestructuras de IA tienen el potencial de más que triplicarse.

La dirección también prevé que continúe la expansión del margen bruto gracias a las economías de escala, una mayor proporción de infraestructuras de IA y la reducción de los costes de fabricación. Señaló que la incorporación de IVR podría situar con el tiempo el perfil de margen bruto consolidado por encima del 50%, aunque no ofreció un horizonte temporal.

Riesgos y aspectos a tener en cuenta

• Navitas sigue siendo deficitaria y se prevé que los gastos operativos no GAAP del T4 aumenten de forma sustancial tras la adquisición de Claros.
• El inventario aumentó en 12,9 millones de dólares en términos intertrimestrales, debido a que la empresa creó reservas de obleas de TSMC y se preparó para la demanda prevista de los centros de datos de IA.
• El pago en efectivo de aproximadamente 126 millones de dólares por Claros redujo el efectivo proforma a unos 405 millones de dólares tras el cierre del trimestre.
• Varios programas de crecimiento se encuentran todavía en fase de diseño, cualificación, toma de muestras o pruebas de sistemas, por lo que los plazos y la magnitud del escalado de la producción representan variables clave.
• La dirección describió la visibilidad fuera del ámbito de las infraestructuras de IA como más limitada y asumió unos ingresos ajenos a la IA prácticamente estables en sus previsiones para 2027.
• La empresa subrayó que los clientes de infraestructuras de IA exigen una elevada fiabilidad y una rápida ejecución, lo que genera exigencias operativas a medida que Navitas escala su actividad.

Puntos destacados del turno de preguntas y respuestas con analistas

La dirección indicó que la ratio book-to-bill del T3 fue de aproximadamente 2 y se había mantenido holgadamente por encima de 1 durante los dos trimestres anteriores. La cartera de pedidos se prolonga hasta mediados de 2027; el primer punto de inflexión en potencia para IA ya contribuye a los ingresos y el segundo cuenta con una representación creciente en la cartera.

En cuanto a la posición de mercado, la dirección estimó que Navitas podría avanzar hacia una cuota del 30% al 40% en los programas de fuentes de alimentación en los que participa, en función del número de proveedores homologados. También señaló que algunos programas del segundo punto de inflexión son de proveedor único, mientras que otros conllevan suministro compartido.

Respecto a Claros, la dirección destacó la densidad de integración y la modularidad como sus principales elementos diferenciadores. La plataforma está diseñada específicamente para el suministro de energía a XPU y se puede configurar en matrices para responder a diferentes requisitos de voltaje, corriente y distribución física.

Navitas definió su modelo de fabricación fabless integrada como un punto intermedio entre un IDM tradicional y una estructura puramente fabless. La empresa pretende mantener la propiedad de la tecnología crítica y de equipos seleccionados, al tiempo que se asocia con fundiciones para garantizar capacidad, mejorar la previsibilidad de costes y respaldar la escala.

La dirección reiteró que la rentabilidad depende de la escala de ingresos, de la continua expansión del margen bruto y de que los gastos operativos crezcan a un ritmo más lento que los ingresos. Su modelo ilustrativo apuntaba a un margen bruto consolidado superior al 50% una vez que escale el IVR y a un nivel de ingresos para alcanzar el umbral de rentabilidad en la franja baja de los 60 millones de dólares, sin concretar una fecha objetivo.

Transcripción completa de la conferencia de resultados

Transcripción completa de la conferencia de resultados

Comentarios de la dirección

Operator

Good afternoon, and welcome to Navitas Semiconductor's Third Quarter 2026 Financial Results Conference Call. [Operator Instructions] As a reminder, this conference call is being recorded today, Tuesday, November 3, 2026.

I would now like to turn the call over to Brett Perry of Shelton Group Investor Relations. Brett, please go ahead.

Brett Perry

Thank you, and good afternoon. Welcome to Navitas Semiconductor's Third Quarter 2026 Financial Results Conference Call. Joining us today are Navitas President and CEO, Chris Allexandre; and CFO, Tanya Stevens.

I'd like to remind listeners that management's prepared remarks contain forward-looking statements, which are subject to risks and uncertainties, and management may make additional forward-looking statements in response to your questions. Therefore, the company claims the protection of the safe harbor for forward-looking statements that's contained in the Private Securities Litigation Reform Act of 1995.

Additional results may differ from those discussed today, and therefore, we refer you to a more detailed discussion of the risks and uncertainties in the company's filings with the Securities and Exchange Commission, including the company's filings on Forms 10-K and 10-Q.

In addition, any projections as to the company's future performance represent management's estimates as of today, November 3, 2026. Navitas assumes no obligation to update these projections in the future as market conditions may or may not change, except to the extent required by applicable law.

Additionally, the company's press release and management statements during this conference call will include discussions of certain measures and financial information in GAAP and non-GAAP terms. Included in the company's press release are definitions and reconciliations of GAAP to non-GAAP items, which provide additional details.

For those of you unable to listen to the entire call at this time, a recording will be available via webcast for 90 days in the Investor Relations section of Navitas' website at www.navitasseemi.com.

And now it's my pleasure to turn the call over to Navitas President and CEO. Chris, please go ahead.

Chris Allexandre

Good afternoon, and thank you for joining us on today's third quarter 2026 earnings call. We appreciate your continued interest and support as we near completion of our strategic transformation and begin executing as Navitas 2.0.

Our third quarter results reflect another period of solid financial performance and accelerated momentum. Total revenue was on the high end of guidance at $14 million, up 33% sequentially and 38% year-over-year, marking the company's first quarter of year-over-year growth since 2024. This revenue growth was driven by high-power markets, which increased more than 50% sequentially and grew by triple digits year-over-year.

As anticipated, mobile and low-end consumer continued to decline as a percentage of revenue and remains on path to becoming insignificant in the fourth quarter. By year-end, high-power will essentially represent the entire company, and we will have substantially exceeded mobile and low-end consumer markets.

In fact, as we now delivered multiple quarters of double-digit sequential revenue growth and gross margin expansion, driven by high-power market and an acceleration in AI infrastructure, our core focus. Gross margin expanded sequentially and was above the midpoint of guidance, reflecting the continued benefits of favorable revenue mix and increasing scale.

Notably, third quarter non-GAAP gross margin of 40% was 130 basis points higher than the third quarter of 2025 when we began the Navitas 2.0 journey. Book-to-bill remains well above 1 and has been for the last several months, and we currently have solid backlog extending into mid-2027.

We also expect continued revenue growth and margin expansion in the fourth quarter, which will represent the fourth consecutive quarter of double-digit sequential revenue growth and margin increase with the company exceeding 2026 as a high per company. Briefly stepping back, roughly a year ago, we made a deliberate decision to pivot away from mobile and low-end consumer and focus our resources on the high-power market where GaN, high-voltage SiC and ultrahigh-voltage SiC provide greater differentiation and stronger long-term growth.

We labeled AI data center, energy and grid infrastructure, high-performance computing and industrial electrification high-power market as clear focus and AI infrastructure as North Star. The success of this pivot is now clearly visible in our revenue mix. At the beginning of 2026, Navitas 2.0 was only a strategy, and now it is ready.

By the fourth quarter, we expect mobile revenue to be effectively gone, as stated earlier, and the company revenue to be essentially all high power. More than offsetting our exit from lower-margin mobile and low-end consumer has been the tremendous growth of AI infrastructure, our North Star. AI data centers plus the grid and energy infrastructure required to power them. We anticipate AI infrastructure to represent more than 40% of our total revenue exiting this year, and we believe this business is poised to triple in the calendar 2027.

Over the last few months, it has become clear and broadly accepted that power is now the #1 constraint on AI compute. The question is no longer only how many GPUs or XPUs can be manufactured and deployed, but how many can realistically be powered. In other words, power efficiency is becoming capacity. When access to power is constrained, every watt saved in conversion and delivery can become another watt available for compute. It increases the scale of compute that can be deployed inside a given power envelope, not only the energy cost.

Together with our high and ultra-high voltage SiC, GaN and recently acquired Claros IVR technology stack serving both vertical and embedded power delivery architectures further positions us to break the AI infrastructure power wall by addressing the efficiency, power density and performance bottlenecks across the full power chains from grid to XPU.

With our strategic acquisition of Claros in early October, Navitas has now a complete platform that is capable of addressing the final stage of delivering efficient power all the way to the XPU. I want to emphasize that there is a growing number of processor sites, CPU, GPU, TPU, XPU and other custom processors as well as companies developing them, each requiring a specific power implementation for the last step of power conversion to core voltage.

This requires a modular approach, which is one of the key differentiators of Claros IVR technology stack as it allows several IVR to be combined very flexibly in a power array to deliver the required power level and density. Navitas was already unique by having both GaN and SiC. And now with Claros, we are even more so.

We are now an enabler of efficient power delivery from the source of energy, starting in a modernized grid with our ultra-high-voltage SiC, then high-density and high-efficiency centralized AC-DC and DC-DC conversion with our high-voltage SiC and GaN. And finally, the power delivery to the XPU with the Claros IVR technology stack.

Going one step deeper, IVR technology solves a real-world scaling problem. Processor power demands are increasing faster than lateral delivery can support with package power moving from roughly the mid-1 kilowatt range towards several kilowatts at sub-1 volt core and several thousand ramp.

By moving conversion closer to the XPU with vertical power delivery directly beneath the processor and embedded power delivery ultimately in the PCB package or substrate, the path is shortened from centimeters towards millimeters or even less. This reduced resistance, inductance, drop and heat.

As an integrated IVR platform, not another power FET, we are combining digital control, power transistors, high-density capacitors, embedded inductors and packaging in one architecture. The platform is also designed to scale with processor currently through a repeatable building block of the order of 40 amp through larger chips and a modular power array into multiple kilo amp system packaged and embedded.

In simple terms, the Cloud IVR technology stack with its ability to serve both future VPD and EPD architectures give us a complete power solution platform. Together with the addition of GFS, it doubles our 2030 SAM to over $8 billion and represent a multiyear content growth opportunity that layers on top of the existing inflection points we are pursuing with our high-voltage SiC and GaN.

The ultimate benefit is Navitas's ability to help customers convert more of every available megawatt into useful AI compute. While our transformation to a high-power company is effectively complete, our go-forward strategy remains grounded in multiple key pillars of transformation that we have discussed over the last 12 months, market focus, technology leadership, operational efficiency and financial discipline.

On market focus and technology leadership, we continue to advance a series of customer and technology proof points. Since our previous conference call, we announced our collaboration with NVIDIA on an 800-volt to 12-volt power delivery board for next-generation AI infrastructure. This builds on our previously announced 800-volt 6 and 800-volt 250-volt DC-DC conversion board, further highlighting the breadth of our system knowledge and capability.

We also licensed GeneSiC technology to Microchip, underscoring the strength of our trench assisted planner process foundation and supporting expanded SiC adoption, particularly in Asia, while also creating long-term potential for additional foundry source. We recently partnered with Microchip to deliver an 800-volt reference design for next-generation AI data center.

Then in late September, Navitas was selected by the U.S. government to develop next-generation 10 kV SiC power semiconductor, reinforces both technology superiority and domestic supply chain resilience. And we also announced the first shipments of our U.S. manufactured fifth-generation GaNFast technology as part of our long-term strategic partnership with GlobalFoundries.

Following shipments of the first wafers in September for the initial product family, including 650-volt GaNFast, we expect to have strategic customers sample very soon. On a related note, we previously talked about pursuing a Foundry plus model. Going forward, we've renamed and we refer to the strategic operating model as Integrated Fabless Manufacturing or ISM, a deliberate alternative to the traditional IDM approach.

The core difference and principles are that we choose what to own, where to invest and who to partner with, allowing us to retain the capabilities that differentiate product while leveraging partners for manufacturing at scale. Importantly, we remain ownership of critical technology IP, selected manufacturing equipment together with targeted co-investment, not purely supplier collaboration.

For purpose of today's call, I want to briefly take this opportunity to clearly define and lay the foundation of our IFM model as I expect it to play a strategic role in our go-forward strategy. Looking more near term, I will also note that we are actively taking advantage of our strengthened balance sheet to strategically build inventory buffers, including raw materials and secure future capacity to support unexpected strong and consistent mid- to long-term growth.

Turning to an update in our progress towards the four inflection points that I introduced in our last conference call, which I intend to keep using to map our growth to architecture change in AI infrastructure. Today, the transition to 800-volt architecture for next-generation AI data center is now certain. The debate is no longer whenever 800 volt is happening, but instead how it will unfold.

As previously outlined, the transition will have a series of inflection points that will drive increasing momentum and growing need for higher power solutions. Navitas's unique combination of high-power GaN, high-voltage SiC and our IVR capabilities will allow us to capitalize on each of these inflections. I want to emphasize that we are not dependent on one architectural transition and that we are not waiting for the final data center architecture to emerge.

These inflection points represent successive opportunities, each of which can increase content per system as the overall market continue to grow and expand. The first inflection, pre-800 volt, is already well underway. We are seeing an acceleration of silicon replacement for any ACDC above a few kilowatt. Products based on our SiC technology addressing this opportunity are in production and driving accelerated growth in the fourth quarter and the first half of 2027 across multiple hyperscalers and OEM ODM.

We are seeing an acceleration of demand, in particular for 1.2 kV SiC, where our previously mentioned Gen 5 is gaining strong traction. The second inflection, the SiC AR rack and the first rollout of plus or minus 400-volt and/or 800 volt is also accelerating. We discussed this in our last earnings call, and it was clear at the OCP Summit in October that this is coming and now definitely coming in 2027.

We are seeing broader design conversion and program across more customers and applications, including AC-DC, DC-DC and battery backup units or BBU, all with more dollar content. Large orders already in place and slated to ramp in the first half of next year. This will drive growth in 2027 on top of the first inflection point continuing across AI data centers that not yet to adopt high-voltage DC. The power level and density acceleration of AC-DC PSUs is really stunning and driving this content jump.

On the DC-DC side, the same is true. We are currently progressing through the final stage of designing and qualification with many customers and orders already in place, which will give us additional programs in support of continued growth over the coming year. What is very interesting about the second inflection point is the variety of applications in both, DC-DC, AC-DC, BBUs and others and the use of different technologies between SiC and GaN, which reinforce Navitas's strategic positioning.

The third inflection point or Nive 800 volt continues to progress. We all know by now, and it was reconfirmed again at OCP that this is the only path to achieving really high-density, high-power rack. I mean 500 kilowatt and plus and up to megawatt racks. We are advancing design and preparation for this high-density rack across multiple vendors depending on the XP. Everyone is looking at it. We continue to believe Native 800-volt solution will be in production before the end of 2027 and be material in 2028.

Today, we are sampling and progressing with customers on more than a dozen programs and are currently in final stage with certain leading OEM ODM for 800-volt power DC-DC conversion board. One existing customer targeting hyperscalers has already placed a large order for significant system test, which reinforce Navitas's strong market position.

The fourth inflection point the complete modernization of energy and grid infrastructure, while further out, it is also moving forward, driven by the speed and energy requirement of AI data centers. We are continuing to work to secure initial design wins with our ultra-high-voltage modules in solid-state transformers, which is really more of a 2028 and beyond growth driver.

In the interim, we're also seeing an acceleration of engagement for shorter-term revenue opportunities battery with battery energy storage system, or BSS and other power grid-related applications, which are driving high demand for ultra-high voltage SC. A portion of these opportunities are expected to continue to grow in calendar 2027, layering on top of the first and second inflection point with AI data center.

In closing, I'm very pleased with the progress. Navitas 2.0 is no longer a plan, but a reality. The results are demonstrated by the revenue mix, growth and margin expansion. By year-end, we will be a high-power company with mobile and low-end consumer essentially eliminated. The market rationale has sharpened. Power is the immediate constraints on increased compute.

Therefore, power efficiency is the solution to expanding capacity. The architecture of AI infrastructure is in the early stage of being rebuilt from the grid to the XP. The acquisition of Cloud completes Navitas' platform and our grid to XPU strategy. With SiC Gallium IVR, we are positioned not only to participate but to enable the multiyear transition with more than one monetization point. The timing of our complete transformation is ideal as we are at the intersection of 2 secular trends, the growth of AI compute and the rearchitecture required to it.

From the grid to the XPU, our end goal is to help customers put more of every available megawatt into useful AI compute. Specific to the fourth quarter, the midpoint of our guidance represents another consecutive quarter of double-digit sequential growth and more than 100% growth year-over-year, driven by high power and AI infrastructure and contributing to further gross margin expansion.

Our expectations for the fourth quarter also represent a return to growth for the full year, increasing approximately 7% in '26 with high-power growth more than offsetting our exit from the mobile business. We believe this is only the beginning as we remain committed on the path of delivering continued growth, incremental gross margin expansion and ultimately profitability.

With that, I'll pass the call to Tonya to review our third quarter financial results and outlook for the fourth quarter.

Tonya Stevens

Thank you, Chris. Before I begin, please note, unless otherwise indicated, I will focus my comments on non-GAAP results. A detailed reconciliation of all non-GAAP to GAAP financial measures can be found in our press release published earlier today.

Revenue in the third quarter of 2026 was at the high end of guidance, increasing 33% sequentially to $14.0 million, an acceleration from the 22% sequential growth we delivered last quarter. This represents an increase of approximately $3.5 million from the $10.5 million in the second quarter.

As Chris highlighted, the double-digit growth was driven by increased traction in high-power markets, which grew more than 50% sequentially and by triple digits year-over-year, reflecting our nearly complete shift in revenue composition away from our historical mobile and low-end consumer business.

We continue to expect revenue contribution from mobile to become insignificant as in low single digits by year-end. As a result of improved product mix and higher quarterly revenue, gross margin expanded by another 50 basis points sequentially to 40.0%. This represented our third consecutive quarter of sequential gross margin expansion.

Our accelerating shift in overall revenue mix toward higher-value, high-power markets and away from mobile and low-end consumer remains fundamental to our gross margin expansion strategy. We expect further gradual expansion of gross margin over the coming quarters as we continue to drive top line growth in high-power markets.

Operating expenses for the third quarter were $16.5 million compared to $15.5 million in the prior quarter and $15.4 million in the same quarter a year ago. Consistent with the guidance we provided last quarter, OpEx increased by approximately $1.0 million as we began making targeted incremental investments in the business, particularly in new R&D programs to support accelerated growth.

Having diligently maintained effectively flat OpEx through the first half of the year, we are increasingly focused on deploying resources and investments in a deterministic way to support the longer-term success and sustained growth of the transformed company. OpEx for the quarter increased roughly 6% sequentially, which was meaningfully lower than our top line growth rate of 33%.

This and future incremental OpEx, including expenses related to Claros IVR technology stack and product development will be allocated to scaling the business, including investments to accelerate new product development, strengthen our engineering and application support for key committed programs and reinforce operational readiness in advance of expected growth in ramping shipments.

Loss from operations for our third quarter was $10.9 million compared to a loss of $11.4 million in the prior quarter and $11.5 million in the third quarter of 2025. In Q3, weighted average basic and diluted shares outstanding were approximately $261.5 million, resulting in a Q3 loss per share of $0.03 compared to a $0.04 per share loss in the prior quarter and a loss of $0.05 per share in the year ago third quarter.

Turning to the balance sheet. Cash and cash equivalents at the end of the third quarter 2026 were $531.5 million compared to $557.4 million at the end of the second quarter. The decrease in cash and cash equivalents primarily reflects our ongoing build of TSMC buffer inventory. Also note, the quarter end cash balance was prior to the payment of the cash portion of the Claros transaction, which closed on October 5.

Our effective pro forma cash balance adjusted for approximately $126 million payment for Claros subsequent to the quarter end would have been approximately $405 million. As a reminder, the company continues to have no debt. Also of note, I want to briefly highlight that we did not utilize any of our ATM facility during the third quarter and are comfortable with our current cash position following the Claros acquisition.

Post Claros transaction, we continue to have ample capital resources to further accelerate and scale as a high-power company. This includes further anticipated strategic investments in support of advancing our integrated fabless manufacturing or IFM model initiatives, potential capacity expansion and supply reservation agreements with our U.S.-based Foundry partners as well as potential pursuit of selective strategic opportunities. With respect to inventory, we ended the third quarter with $32.4 million of inventory, including $11.2 million of inventory classified as long term on the balance sheet. This compared to $19.5 million in the prior quarter and reflects our previously planned build of appropriate buffers of TSMC wafers to ensure a smooth transition for our customers as well as other supply chain-related investments.

Similar to the previous quarter, the sequential $12.9 million increase in Q3 inventory primarily reflects our measured investment to support AI data center customers' future anticipated long-term growth. More broadly, channel and distributor inventory remains healthy.

Moving to guidance for the fourth quarter of 2026. We expect continued sequential growth with revenue increasing 14% to $16.0 million, plus or minus $0.5 million. At the midpoint, this represents 119% growth year-over-year and what will be our fourth consecutive quarter of sequential growth, driven by high-power markets and specifically AI infrastructure.

Also based on this guidance, we anticipate achieving revenue growth of 7% for the full year 2026 with our core high-power business continuing to deliver triple-digit growth even with our mobile business dropping to effectively zero after representing a large majority of total revenue last year.

Exiting the year, we expect high-power markets to represent almost all of our business with over 40% of total revenue coming from AI infrastructure, which includes both our AI data center business and the grid and energy infrastructure business needed to support AI data centers.

Lastly, I would also add that we closed out the third quarter with very strong book-to-bill ratio, well above one, and we currently have solid backlog that extends into mid-2027. Non-GAAP gross margin is expected to expand to 40.5%, plus or minus 100 basis points, which at the midpoint represents a 50 basis point increase, reflecting our fourth consecutive quarter of margin expansion and the continued favorable shift in the revenue mix toward high-power markets and marginal incremental benefit from improved scale.

As previously discussed, OpEx will step up in the fourth quarter as we take on the overhead operating expenses associated with Claros as well as continue to make incremental investments organically, primarily focused on R&D and operational readiness in advance of our anticipated growth over the coming year. Non-GAAP operating expenses for the fourth quarter, which again will include Claros, are anticipated to be $25.0 million, plus or minus $1 million.

That concludes our formal remarks. Operator, please open the call for questions.

Operator

[Operator Instructions] And your first question comes from the line of Quinn Bolton with Needham & Company.

Preguntas y respuestas

Shadi Mitwalli

This is Shadi Mitwalli on for Quinn. Congrats on the results. I think you guys mentioned that backlog is extending into mid-2027. And as these high-power markets start to inflect, it would be great to get a sense on how we should be thinking about growth heading into next year.

Chris Allexandre

Thank you, Shadi, for the question, and I appreciate you giving us a platform to talk about the outlook for next year, right? So as we talked about in the last two earnings, we've seen consistent increase in book-to-bill. As a matter of fact, the book-to-bill last quarter or Q3 was actually two and has been consistently way above one for the last two quarters, right?

So the way we see '27 with the inflection point unfolding on top of each other is the path for us to be a double-digit growth quarter-over-quarter fueled by the acceleration of AI infrastructure, which we mentioned in the earlier remarks that will triple from the baseline in '26. And that double-digit growth will continue to accelerate as each of the inflection point comes on top of each other.

So basically, the message today is we've done the first part of the transformation of the company, which is the financial pivot we talked about, consumer to high power, as you mentioned. And now we're entering the other phase of the company where the growth is just starting driven by the AI infrastructure will give us multiple years of double-digit quarter-over-quarter growth.

Shadi Mitwalli

Great. That was helpful. And then maybe just on the Claros acquisition. We see a number of IVR acquisitions we announced over the past few months. So I was just curious, can you guys talk about what you guys saw in Claros and maybe some of the advantage of their technology versus competing IVR technology?

Chris Allexandre

Thank you. Very good question. So the first thing before I answer specifically your question, what I would say is the fact that other MOSFET and silicon companies are acquiring IVR is a confirmation that what we talked about, which is the architecture has to change, okay?

The way lateral power delivery today moving to vertical power delivery using the same technology kind of put together in a vertical package is not going to sustain the density and the amount of current that those processors need, right? What we saw in Claros, and I'll put aside the culture, the very similar to -- and the quality of the engineering team is a real IVR technology stack designed for XPUs.

Instead of IVR that was designed for computing converted, okay, to AI data center like we've seen in some of the companies that just got acquired, right? A true 3.3-volt starting point, allowing to do a 3:1, everything integrated in the stack, the capacitors, the inductors, the controllers, the FinFET, basically allowing you to do real high density because you don't need external component, right?

And that's what we saw. And the modularity, which I mentioned in the webcast, is also important. What we see is, of course, everybody talks about the lead GPUs, but there's as many processors now that we have use case. And all those processes, call it CPU, XPU, NPU, LPU,PU customs, right? They all need super high density, high current type of IVR and power delivery. And I think the flexibility and the modularity that we saw in Claros is unique, and that's why we made that move.

Operator

Your next question comes from the line of Kevin Garrigan with Jefferies.

Kevin Garrigan

Chris, congrats on the great results and all the progress. Going off of Shadi's question, can you just give us a sense of how much of the 2027 growth is supported by awarded programs and what is already in your backlog today versus expected design conversions?

Chris Allexandre

I'll start and Tonya can add some comments, right? So, we have backlog way into '27 already. So we cannot say the whole year is done, right? But what I would tell you is we are very surprised, and I mean, positively by the acceleration of the Inflection 1, which we saw really in Q3 and Q4 this year, but Inflection 2, right, which is the sidecar where we see an accelerated conversion of the number of programs across AC-DC, DC-DC and BBUs.

So more programs converting faster into higher power, converting from design into design win into backlog. And last but not least, because we talked about inflection in Inflection 3 in the last call, now we start to see -- and we mentioned it in the remarks, right, we've actually got a pretty large order, 1 million units of GaN devices for an 800-volt DC-DC high-density brick for Inflection 3, right? So things are really coming into shape. I would tell you, Kevin, slightly faster than I even expected six months to three months ago.

Tonya Stevens

Yes. And I would just go on top of that. That's exactly what I see and how I think about it. The Inflection 1, like Chris talked about, it's in our revenue. We saw it in our revenue last quarter. We see it in our revenue this quarter accelerating. Inflection 2 in our backlog, that's helping us fill out 2027. And even Inflection 3, we're starting to see that move beyond just high-volume orders and sampling. He talked about the one million units. That's a significant amount of volume that we're looking at for Inflection 3. So that's exactly what we're seeing.

Kevin Garrigan

Yes. Got it. Okay. That's perfect. And then just on gross margins, congrats on the continued expansion. Can you just give us a sense about how we should think about the gross margin profile between silicon carbide, GaN and IVR as each business scales? So, we don't really kind of give detail on the split on gross margin between the different technology. What I would tell you is, if you look at the gross margin expansion in '26, right?

Chris Allexandre

So, let's go back. A year ago, to this day, we talked about transforming this company from consumer to high power. Basically, the gross margin has increased, if you look at the guidance for Q4 by two points. That was essentially moving from consumer to high power. The next phase of the expansion of gross margin is coming from scale as we continue to grow double digit, but also with the AI infrastructure, which is now 40% of the company in Q4 and poised to triple next year.

And that part of the business is actually much also on a higher margin, right? So what you're going to see is I just got the question before, how do we see growth, right? I mentioned double-digit quarter-over-quarter with that double-digit kind of accelerating over time.

I think you're going to see the same on the gross margin, consistent gross margin expansion quarter after quarter as we scale. And of course, as we continue to make fundamental transformation moving to 18-inch on the GaN, continue to work with our foundries on cost reduction and so forth. But I mean, look at this as a continuous gross margin expansion organically with the migration to high power and now to AI infrastructure.

Tonya Stevens

Yes. And I would just confirm that the way we think about it is not relative to the technology, but complexity. And we've given you sort of the insight before that the higher power, higher complexity, it is, the higher the ASP and margin tend to be. And that includes more difficult technology problems to solve, including the IVR technology stack. Those tend to contain higher margins. versus splitting that between GaN and SiC.

Chris Allexandre

Yes, Kevin, one thing I would add is, I mean, Tonya is spot on, right? Margin comes with value, value comes from the complexity you're solving the customer problems. What we see is look at Inflection 1 and Inflection 2, right, or even going to Inflection 3.

We are talking about multiple x of density of power on the AC/DC front, moving from silicon to silicon carbide, then moving to GaN when the add-on become native. Those complex problems, customers are basically retributing with good margin if we solve their problem with speed. And I think this is what kind of goes back to the core of what we are trying to do as a company.

Operator

Your next question comes from the line of Kevin Cassidy with Rosenblatt.

Kevin Cassidy

And I'll echo the congratulations on the steady progress. In the IVR market, that's a new market. How early on in the XPU or GPU design do you have to be to define the IVR? Do you have to work with them as they're designing the XPU?

Chris Allexandre

So it's a very good question, Kevin. Thanks for asking it. So first of all, I think what you need is what we call an IVR technology stack. That's what we got with Cloud, right? So, integration of the FinFETs, the controller, the inductors, the capacitors all together into the same technology.

Then you apply that to either vertical power delivery, which is basically a chip that you put on the backside of the PCB or in the PCB, right, in the form of an array, as we talked about, to deliver the power then to the customer or even going one step further where it's not a chip anymore, you embed the full system or a subpart of the system into the substrate and the XPU. In both ways, you have to work with the XPU vendors.

In both ways, you have to work with the OEM or power OEM that serve the last step of our delivery to the XPUs to basically fit the right architecture. And that's what I got the question from Sean earlier on, I think it was Kevin. What did we like about Claros, right? Well, I liked beyond the best density in my opinion, because there's no caps outside is this modularity. So every customer needs a different type of shape of the Power VRM and a different level of voltage and current, right? And that's kind of where you have to work with those.

The other thing I would say is keep in mind that many customers are becoming vertically integrated. So we used to have GPU companies. Now we have hyperscalers doing their own processors. We also have processor companies that become rack company, right? So the answer to your question is the technology stack is the same, but you have to work with the processor company, which are becoming rack company or even hyperscaler company to basically deliver the power delivery.

Kevin Cassidy

That's great. Yes, it's getting to be a complicated design cycle. And which brings me to an interesting strategy you brought up is integrated service manufacturing where you're going to buy the equipment. And I guess, is that going to be in the back end? Like, as you're describing the IVR could all be different shapes and packages for different designs. Is that where you're bringing in the equipment? Or will it be on the front end, too?

Chris Allexandre

So, let me just correct you for a minute, right? We talk about integrated fabless manufacturing for IFM, right? Primarily on the foundry side. So the best way to think about this is if you look at both SiC and GaN, -- we actually own the technology stack, same on the IVR, right? So, on the SiC side, we have GenSiC, which is a full stack high-voltage, high-voltage technology.

On the GaN side, we overlay on top of TSMC in the past, now GF, our own IP to differentiate in reliability and other figure of merits and same with the IVR when we own the full stack, right? So, what we want to do is to own the stack but partner with foundries where we invest capital to basically secure capacity and supply to enable cost predictability and competitiveness.

So, the best way to think about this is a bit of the similar kind of we announced with GF on the GaN side and other foundry we work with, can we put the best of both together. We don't want to be an IDM. We don't want to be a fabless, we're somewhere in the middle, and we believe this is the best capital-effective way to deliver the high growth while controlling our technology.

Operator

Your next question comes from the line of John Tantan with CJS Securities.

Jonathan Tanwanteng

Congrats on a nice quarter and outlook. I was wondering I think I heard pretty clearly that you expect to grow data center infrastructure revenue by 3x going into next year. What do you think the non-data center portion contributes? If you can tell us what's involved there and what the contribution might be?

Chris Allexandre

Yes. Jon, thanks for asking that question. As I talked about, we have backlog way into '27. As a matter of fact, the visibility we have in AI infrastructure is much longer and much extended compared to other markets, right? So what we've modeled for next year is anything non-AI infrastructure is more or less flat.

We see some growth in industrial. We see some flattish in the computing. Remember, we addressed the super high-end computing market, not the computing market as a whole, right? So we've assumed that flat because we don't have a full year or even two quarters type of visibility.

But on the AI infrastructure, as you pointed, because of the visibility we have, because of the accelerated momentum in Inflection 1, Infelction 2, because of number of design win converting, we are clearly seeing a potential for that business to more than triple next year, right?

So the best way to think about it is '26 was a year of high power basically compensating the exit of mobile. This is done. This even allowed us to get higher than when we started this journey from a revenue standpoint and getting 2026 with $16 million. Next year is really the growth is coming from AI infrastructure, and we assume the rest being flat.

Jonathan Tanwanteng

Okay. Great. And then more interestingly, as you head into '28, I think you've got multiple accelerations, right? You have the NVIDIA 800-volt products ramping. You have the SST products ramping, you have Claros ramping. Do you see further acceleration in '28? Or is it hard to comp against the growth you're going to see in '27 or maybe it's a smaller base that you can still accelerate from there?

Chris Allexandre

Of course, it's more difficult to predict as you get longer in the horizon, right? What I would tell you is we see based on the product we have, the technology we have, the engagement we have, the inflection points, double digit, and that double-digit quarter-over-quarter is going to accelerate as each of those inflections overlay with each other. So it's not inconceivable to see the quarter-over-quarter compound growth to start to accelerate late '27 into '28 into '29. As each of those technology, SiC, GaN and even IVR overlay on top of each other and each of those inflections come on top of each other.

Jonathan Tanwanteng

Great. If I could ask one more, just, Tonya, how do you see OpEx scaling with all this revenue growth? Do you have a number as maybe as a percentage of revenue growth? Or is there another way to think about it over the next year or 2?

Tonya Stevens

Yes, that's a great question. So the step-up you saw in Q4, we talked about that was our normal run rate of our guidance as well as Claros taking on Claros. And the way we think about that is still what we've talked about before, OpEx increasing at a percent of revenue. So after that initial step-up, we see that happening in '27, so as a fraction of our revenue growth. We're not really giving the full year out, but that's how we're thinking about that.

Chris Allexandre

John, I think one thing I would add on top of what Tonya said is we've been very clear since the beginning of this transformation that our eyes are on getting to a profitability. Profitability will come from scale. We have two vectors. One is scale and consistently growing double-digit the revenue, as we talked about.

Number two is gross margin expansion and the gross margin expansion will continue as we just talked about. And when IVR comes in play, we see also a step function in the combined gross margin of the company. But we remain focused on making sure we do that by increasing OpEx at a fraction of the revenue growth, so we can get to profitability at the horizon, right, which remains a very important focus for us.

Operator

[Operator Instructions] And your next question comes from the line of Richard Shannon with Craig-Hallum.

Richard Shannon

I'm going to follow up on the last question and answer, especially taking to the obvious angle here, which is thinking about a breakeven model here. Not asking for the time frame you might hit this, but how do we think about that model here?

And I presume that it would -- I guess I'd prefer an answer that would include IVR with the gross margin dynamic you described here, but kind of scale and what's the margin structure or I guess, maybe I'll just let you answer of how we can look at the breakeven model. And I have a follow-up.

Chris Allexandre

So thank you, Richard, and good to hear from you. I think the best way to think about this is I think we're going to continue organically to grow the consistent gross margin. We see the company gaining a couple of points over time. And then it's not inconceivable, as I said, once we have IVR I think that the blend of the gross margin is north of 50%.

Once you get there, if you take our OpEx in Q4 of this year, kind of multiply by four annualize it and increase -- apply a yearly increase, which is less than revenue, you're basically getting OpEx in the low 30s, right, which at 50% gross margin would basically make the breakeven point around the low 60s. So that's kind of -- if you ask me about the model, that would be kind of the model you have to keep in mind.

Richard Shannon

Okay. Perfect. That's very helpful. Chris, maybe just ask a very high-level question here since we've seen some activity both with Navitas as well as across the industry, but some patent arguments, I guess, we'll call it there, and you've -- you're on both sides of this. Maybe you can help us understand where you sit here and how you expect kind of the industry to evolve over the next couple of years in this dynamic.

Chris Allexandre

Richard, I think I kind of got the question in the last earning about the litigation. I think I was fairly explicit, okay, on the last earning on my view of this. If you don't mind, I would defer to what I said before. We'll continue to defend our IP and we'd expect IP, but we'll continue to defend our IP. And I would defer to what I said on the last earnings, if you don't mind.

Richard Shannon

No problem there. Maybe let me ask another question, a replacement of that one here, Chris. Maybe you can talk about the inflections here, and obviously, you're gaining some real traction in backlog here in inflections numbers one and two here. Maybe if you can talk about your share in any kind of qualitative terms here in any way that you think is interesting, like SiC versus GaN or any specific pieces within those inflections, that would be helpful.

Chris Allexandre

It's hard to give you a blended market share, right, of the Inflection 1, but a couple of data points, right? Number one is Inflection 1 is essentially replacing as you move to 3, 5, 10 or 8, 10-kilowatt density of PSU, right? You're replacing silicon with silicon carbide. And the beauty of this is this is replacing with 1.2 kV.

So, the content is pretty nice. And usually, you have two to three vendors, right? So, some program, we are first source, some program we are second source. And when you have 3 vendors, right, one gets 40, the other one gets 30, the last one gets 30. So, I would say, over time, it trends to a good 30% to 40% share of the market -- of the PSU we participate in.

What is interesting in the inflection 2 is, number one, it's a lot more application. So there is AC/DC, which is basically now taking it to 15, 18, 25, 27, 30 kilowatt. There is also DC-DC with 15, 30-kilowatt DC-DCs and BBU, which start to use both SiC and GaN. So, the beauty of Inflection 2 is not just a SiC play. It's also a GaN play, right? And here, we've seen, in some cases, we're sole source. In some cases, we are gain sharing, right?

So, for me, the way I view it is what you want is to be first on the print, okay, and give customers a secured path from availability standpoint and execution standpoint, so they can ramp the production because I talked about it earlier, we see an acceleration. If you would have asked me six months ago, do you see Inflection 2 and sidecar coming that fast, I would have said it's coming, but it's definitely coming faster than we expected.

Same with Inflection 1, okay? That's why you saw us kind of coming on top of the estimates, okay, that we had set for ourselves that you have set for ourselves, right? So, I think the speed is helping because if you are first and speed is core to our values, if you first, you can command more share, right? But it's difficult to give you kind of how much market share of that market we own.

What I would tell you is if you remove mobile, we grew in '26 based on the midpoint of the guidance for Q4, right? We grew 125%. I believe this is faster than the market. If you look at GaN and SiC serving highpower/AI infrastructure, it's faster than the market. So the momentum is there for us to gain share, okay? And we are not here to participate, as I mentioned. We here to gain share.

So, I think we have with our speed and the technology and the focus that we have. We're not a company that does EV on Monday, industrial on Tuesday and AI data center on Wednesday and Thursday, right? That's the only thing we do. And that focus helps us to secure more share than our competitors.

Operator

Your next question comes from the line of John Tanwanteng with CJS Securities.

Jonathan Tanwanteng

I was wondering if you could talk a little bit more about Claros. I think you disclosed on the update call you gave a week or two ago that they have an anchor customer for the IVR product. What is the potential scale of that one customer, number one? And number two, what is the pipeline of additional customers behind that? And how quickly could those come into play over the '28 to '29 timeframe?

Chris Allexandre

Thank you, John, for that follow-up question. So, the anchor customer is a vertically integrated customer that look at processor up to the system level. Claros has been working with them from the beginning. To Kevin's point earlier, it has helped to shape the technology because it's not just about the technology stack is how do you put that into a chip, into an array, into a system that delivers density, right?

So that customer, as we mentioned at the acquisition, is going to ramp in a meaningful way really late '28 and '29. Since -- and Claros before the acquisition was essentially focused on that customer to prove their technology and get scale, right? Since we combined the two companies, we've met pretty much everybody out there doing an IVR program. And we're starting to see good feedback, but also good momentum in the technology that we had.

And again, what -- the consistent feedback we have is this is a race for density, okay? And then you're going to get into argument into Claros versus Ferric versus Empower versus Parlattice and others, who has the best density and who has the best approach, right? And I don't want to bore you with details, but density without caps outside or density with caps outside is different.

But this is not what customers tell me. Customers love the modularity of Claros. Let say this is something which has been though through with the system in mind, not even the XPU in mind with the system in mind. And that is getting a lot of traction. And I think we'll, under Navitas, be able to scale that to more customers. A bit too early to declare victory, right? There's a lot of work to do.

Operator

That concludes our question-and-answer session. I will now turn the call back over to Mr. Chris Allexandre for closing remarks.

Chris Allexandre

So thank you, everybody, for listening to us. I'll leave you with something which I said, but I think it's very important that you crystallize that in your mind, right? A year ago, to this day, we said we're going to transform this company from a consumer-centric company to a high-power company. We've done that at the same OpEx, more or less.

We gained two points of gross margin and while we have been able to deliver consistent double-digit growth for the last few quarters despite the headwind of conscious moving away from mobile. Now we're entering into the second phase of the company. We've focused ourselves into AI infrastructure. We have the technology stack that we need from the ultra high-voltage SiC, high-voltage SiC, GaN and IVR. AI infrastructure, which is data center and the grid that power infrastructure is 40% of our sales at the end of Q4 is going to drive the growth in the future.

And I think we just got started. And I feel -- now there's a lot of work to do. This market is very demanding in reliability and speed, but I think we have the right team, the right technology. And I think now you have to think about us as a consistent double-digit growing company, leveraging the inflection points that come on top of each other. That's what I want to leave you with.

And I also want to take this opportunity to thank our employees. It's a lot of work to transform the company, and they've been very resilient and determined to do this. And now we are just can scale Navitas to the next level. So, my closing remark goes to them and all the work they've done for the last 12 months and what they're going to do in the future. Thank you.

Operator

Ladies and gentlemen, that concludes today's call. Thank you all for joining. You may now disconnect.