Talk Around Town: Onsemi Claims GaNEXUS GaN Portfolio Could Shake Up AI Data Center Power

THE QUICK TAKE

Onsemi announced its GaNEXUS gallium nitride portfolio on June 9, 2026, targeting AI data centers and robotics, with devices available for sampling across a 40V to 650V range, according to the company's press release.
The company claims GaNEXUS can shrink magnetic component sizes by up to 60% and double power density, but those figures are self-reported and have not been independently benchmarked by any third party.
The broader GaN race for AI infrastructure is real and confirmed by multiple sources, with rivals including Navitas Semiconductor, Infineon, Texas Instruments, and STMicroelectronics all chasing the same market.

What Folks Are Jawing About: The GaNEXUS Announcement

Well, slap a catfish and call it Tuesday — onsemi has rolled out what it describes as the GaNEXUS gallium nitride power portfolio, announced June 9, 2026, according to the company's own GlobeNewswire press release. The company says the lineup is aimed squarely at AI data centers, industrial automation, robotics, and energy infrastructure, which is about as broad a target as a barn door in a field. According to onsemi, devices are currently available for sampling across voltage ranges spanning 40V all the way to 650V, including what the company calls GaNEXUS Smart devices at the 650V tier.

Onsemi also says GaNEXUS can be paired with its own Treo Platform for sensing, control, protection, and power management functions, framing the combination as a more complete solution — though that is the company's own description of its own product lineup, not an independently verified assessment. The company further discloses that its overall GaN solutions design funnel has surpassed $1.5 billion spanning products from 40V to 1,200V, per management disclosures cited in analyst reports. Now, a funnel is just a funnel until something actually pours out the bottom, and whether that pipeline converts to real revenue remains an open question no outside party has answered.

What We Actually Know for Certain

Here is the part of the barn that ain't on fire yet. The announcement itself is confirmed: onsemi did introduce GaNEXUS on June 9, 2026, and devices are genuinely available for sampling, per the press release and trade coverage from Power Semiconductors Weekly. The voltage range — 40V to 650V — is a confirmed specification from the company's own disclosure.

On the financial side, onsemi's AI data center revenues rose more than 30% sequentially and doubled year over year in Q1 2026, per company disclosures cited in multiple analyst reports including Zacks coverage republished by The Globe and Mail and Yahoo Finance. Those numbers come from management, so they carry the usual self-reported limitation, but they are cited across several separate analyst write-ups. Separately, the competitive landscape is genuinely confirmed by independent sources: rival Navitas Semiconductor has also repositioned toward high-power AI infrastructure markets using GaN and silicon carbide products, and analysts at multiple outlets independently name Infineon, Texas Instruments, and STMicroelectronics as additional players in this space.

What Nobody Has Verified Yet

Lord have mercy, here is where it gets murkier than a catfish pond in August. Every single performance figure attached to GaNEXUS — the claim that it can reduce magnetic component sizes by up to 60%, and the assertion that power density can improve by as much as 2x — originates entirely from onsemi's own press materials, according to the company. Not one independent lab, trade publication, or third-party engineer has reproduced or benchmarked those numbers in the available sources. That does not mean the claims are wrong, but it means you are currently taking onsemi's word for it the same way you take a used-truck salesman's word about the mileage.

The $1.5 billion GaN design funnel figure is a management disclosure, and analysts note plainly that qualification cycles in AI infrastructure and industrial markets are long and unpredictable, meaning pipeline and payday are very different animals. Mass production timelines are unspecified. Hyperscaler qualification processes — the actual hoops a chip must clear before a major cloud operator will drop it into a real data center — are not addressed in any available source. Until those cycles complete, GaNEXUS remains a promising sample, not a proven revenue stream.

The Competition Is Real, and It Is Hungry

One thing that is not in dispute is that the GaN-for-AI-infrastructure race is absolutely underway, confirmed independently by multiple analysts and rival company disclosures. Navitas Semiconductor, covered by Yahoo Finance across multiple pieces in May and June 2026, has made a deliberate pivot toward higher-power, infrastructure-grade markets where GaN and silicon carbide are already attracting serious attention from larger competitors. Navitas argues that offering both GaN and SiC under one roof gives it a strategic edge — a positioning claim that, to be fair, onsemi's data does not currently refute, and Navitas's data does not currently prove.

There is a cautionary note worth nailing to the barn door, though: Navitas, despite operating in the exact same market onsemi is targeting, reported a non-GAAP net loss of $13.8 million on just $8.6 million in revenue in Q1 2026, per Yahoo Finance coverage. That gap between enthusiasm and profitability suggests the GaN-for-AI opportunity is still early-stage and genuinely uncertain, even for a company already deeply committed to the market. Onsemi is considerably larger and better capitalized, but the data point is a useful reminder that promising funnels and real margins are two separate counties.

Analysis: Is GaN the New Silicon Battleground, or Just the New Buzzword?

This next part is analysis, not settled reporting, so put on your thinking overalls. The GaNEXUS launch fits a recognizable pattern: a major semiconductor company plants its flag in a high-growth segment with a freshly branded portfolio, backs it up with impressive self-reported specs, and points to a fat design funnel as evidence the market agrees. Whether that pattern resolves into genuine disruption of silicon-based power delivery in AI data centers — or quietly fades the way some previous 'next silicon killer' campaigns have — depends almost entirely on qualification outcomes nobody outside onsemi currently knows.

What does seem analytically clear is that gallium nitride is becoming a genuine battleground for AI infrastructure power semiconductors, not just marketing fluff. The simultaneous moves by onsemi, Navitas, and the broader competitive set named by analysts — Infineon, Texas Instruments, STMicroelectronics — suggest multiple serious players believe this transition is real and imminent enough to invest heavily. Onsemi's confirmed sequential and year-over-year revenue growth in AI data center segments adds at least some financial credibility to the broader thesis. But until GaNEXUS clears hyperscaler qualification and ships at volume, the distance between the company's claims and confirmed market impact is, as the old-timers say, a long row to hoe.

Who is doing the hollering

These links show where the chatter came from. A link is attribution, not our endorsement or independent confirmation.

Revision record

Last checked Jun 21, 2026, 5:07 AM EDT. Talk Around Town: GaNEXUS devices are currently sampling only, not in mass production. All claimed efficiency and size-reduction figures come exclusively from onsemi. Design-win timelines, hyperscaler qualification cycles, and whether the $1.5B funnel converts to revenue are unverified by independent parties.