Aeluma, Inc. (ALMU) Earnings Call Transcript & Summary

Good afternoon, and thanks for joining us for Aeluma's earnings call covering the first quarter of fiscal 2027. Joining me today are Dr. Jonathan Klamkin, our founder and CEO, and Christopher Stewart, our CFO. You can find a presentation that goes along with today's remarks in the investor section of our website. During today's discussion and in our responses to questions, we may make forward-looking statements, and those carry various risks and uncertainties that could make our actual results differ materially from what those statements suggest. You can find these risks and uncertainties detailed in the earnings press release we issued today, as well as in the reports filed with the United States Securities and Exchange Commission. Those reports, together with today's earnings release, are available under the Investors section of our website. Aeluma undertakes no obligation to update or revise any forward-looking statements to reflect events or circumstances that may arise after the date of this call. During the discussion, the company will refer to non-GAAP financial measures, including EBITDA and adjusted EBITDA. A reconciliation of non-GAAP financial measures to the most directly comparable GAAP measures is included in our earnings press release and SEC filings. With that, I'll hand the call over to Aeluma's CEO, Jonathan Klamkin.

Thanks, Moira, and thanks to everyone for being on today's call. We'll start today by reviewing how much has been happening lately across the AI Datacom sector, including announcements that demonstrate the strength of Aeluma's technology and its go-to-market strategy. At the European Conference on Optical Communication, or ECOC, which took place not long ago in September in Malaga, Spain, the slow and wide transceiver concept for rack-scale AI came up again and again in discussions. As hyperscalers look to add more compute in chasing 1,000-unit XPU/GPU clusters and beyond, scaling up inside the rack and between racks becomes necessary. Traditional copper cables hit a ceiling at high data rates once you go beyond a few meters. Optical interconnects are the answer there. We expect several architectures to exist side by side as data center build-out continues at its massive pace. What separates these formats is chiefly the proximity of the optics to each switch ASIC or XPU/GPU, along with how far the data must go. Pluggable optics modules have advanced beyond 100G per lane to 200G and 400G per lane, covering distances ranging from tens of meters up to kilometers; their primary role is scaling across the data center. Near-packaged optics (NPO) and co-packaged optics (CPO) handle scaling up inside racks and between adjacent racks, and each shortens electrical paths dramatically to reduce the burden on power-hungry electronics. Transceivers built around a slow-and-wide design are being considered for links reaching 50 meters, with 100 meters perhaps attainable. The idea behind slow and wide is straightforward. Instead of cramming the full data load into a small handful of lanes, it is distributed across a moderate count of lanes so that power can be saved. That method brings built-in benefits. No single lane is as vital, redundancy comes for free, and there is headroom for future scaling. Outside of ECOC, the conversation has continued at the Open Compute Project, or OCP, Global Summit, as well as at the Photonic-enabled Cloud Computing Summit. Credo has unveiled short-reach micro-emitter optics, known as SRMO — an OCP workstream that unites know-how across the entire optical connectivity stack so that development of the wide and parallel ecosystem can move forward. Hyperscalers and ecosystem partners keep applauding the Optical Compute Interconnect Multi-Source Agreement, or OCI-MSA, for setting an open, interoperable optical interconnect spec aimed at AI scale-up. That spec lays out wavelength channels and groups for a bi-directional transceiver architecture, underscoring how essential lasers are to AI datacom. Supply of indium phosphide substrates is still tight. Indium phosphide fabs can't match the throughput of gallium arsenide and silicon fabs. Earlier this year, NVIDIA put money into Lumentum and Coherent to help expand indium phosphide manufacturing. Earlier this month, Marvell said it had reached a deal to buy Lumentum. By joining forces, the two companies combine top-tier technology to address demand for highly integrated optical solutions serving AI infrastructure. In total, the U.S. AI infrastructure build-out that got underway in 2025 is projected to hit an aggregate investment of $10.3 trillion by 2032. Photonics accounts for a rising portion of that spending, and the opportunities available to Aeluma's technologies are huge, spanning slow-and-wide along with fast-and-narrow transceivers, which are important for both scale-up and scale-out AI architectures. As a reminder to everyone, Aeluma relies on no indium phosphide substrates and no indium phosphide fabs. Aeluma's MOCVD technology, which scales well, can be used to make micro-emitters and micro-photodetectors for slow-and-wide transceivers capable of integrating directly with CMOS. Aeluma is also developing quantum dot lasers on gallium arsenide substrates, aimed at scale-up and scale-out architectures. No new fabs need to be built and qualified for Aeluma. We're able to work with gallium arsenide and silicon fabs that are already qualified. Our goal is to be the world's top supplier of semiconductor photonics components, and our heterogeneous integration platform — pictured on slide 4 — could take us there. On slide 6, you can see our LINX S-series photodetectors, which are in development for slow-and-wide architectures. Our Quasar quantum dot lasers are being developed for both scale-up and scale-out interconnects. Indium phosphide substrates aren't needed for either, nor are indium phosphide fabs, and each could potentially back efforts like the SRMO-OCP workstream and the OCI-MSA; we continue collaborating with Sumitomo Chemical Advanced Technologies, our scaling partner for MOCVD wafers, along with fab partners. In addition, we announced an agreement with AIXTRON to buy the first of what may be multiple new MOCVD systems, aimed at further capacity expansion and at producing 150 mm and 200 mm substrates, covering both gallium arsenide and silicon. The AIXTRON G10 arsenide-phosphide system has grown popular among indium-phosphide manufacturers for scaling and for meeting laser demand. Aeluma has been collaborating with AIXTRON over a relationship spanning multiple years to set up these more recent tool models for our proprietary processes, which avoid the use of indium phosphide substrates and are able to scale beyond 150 mm, or 6-inch, substrates. Talks with customers in datacom and additional markets have broadened even more; they now involve negotiating potential non-recurring engineering investment as well as product opportunities, several of which could carry large volumes to support the build-out from 2028 through 2032. Aeluma's core technologies — Lynx high-speed photodetectors and Quasar quantum dot lasers — are designed to take advantage of our growing pipeline of commercial opportunities. To sum up, the size of what's being invested in AI infrastructure is putting strain on the supply chain, and we're glad that so many customers and partners have shown interest in our approach of pairing top-tier materials with manufacturing that can scale. As we keep working through fiscal 2027, our attention stays on commercialization and on creating shareholder value over the long term. I'll now hand the call to Chris Stewart, our CFO, who will go over the financials.

Thanks, Jonathan. Now to the financial results for our first fiscal quarter of 2027. Revenue came in at $568,000, versus $1.4 million in the first fiscal quarter of 2026, and it was generated mainly from government R&D contracts. As we talked about on our year-end call, we're currently putting priority on funded R&D programs that line up with our commercial priorities. GAAP net loss for the first quarter was $3.8 million, or 19 cents per basic and diluted share, compared with a net loss of $1.5 million, or 9 cents per basic and diluted share, in the same period a year ago. Adjusted EBITDA loss for the quarter was $2.7 million, compared with an adjusted EBITDA loss of $450,000 in the first quarter of fiscal 2026. That year-over-year change mainly reflects greater investment in headcount, product development, and the operational capabilities needed to support customer programs and commercialization. We finished the first quarter with a strong balance sheet, including $53.9 million of cash and cash equivalents, compared with $56 million at June 30, 2026, and our balance sheet is still debt-free. On the CHIPS program letter of intent for up to $30 million in proposed funding, we're continuing our diligence and contract negotiations. Terms aren't final yet, and any potential award is still subject to definitive agreements and applicable conditions. Over the rest of the year, we expect to keep increasing investments in the people, systems, and supply chain initiatives that support our customer-driven technology development and to prepare for scaling our business. Our goal is to respond to urgent market demand for new solutions while keeping fiscal discipline as we pursue the significant opportunity we see ahead. I'll hand the call back to Jonathan for his closing remarks before we open the call to your questions.

Thanks, Chris. In summary, we're happy about the steady strides we've been making toward commercialization, and about the rising excitement in the AI datacom market that is powering our go-to-market strategy. As ever, I'd like to express appreciation to our amazing team for the effort and commitment they bring, and to all of you for the backing and enthusiasm you've shown. Thank you.

Hi, Tyler here, filling in for Richard Shannon. I was wondering about that MOCVD tool you signed a purchase agreement for — when will it be installed and running? And once it's in place, could you walk us through the capacity you'll have compared with what you had before?

Thanks, Tyler, for asking that. We’re not able to comment on the timeline for that new system, though what we can say is this: for commercialization, we don’t need it; that tool is meant for expanding capacity for next-generation products as well as for some advanced development. It would definitely raise our capacity, but given the capacity we have in-house at Aeluma today and the capacity commitment we have from our partner Sumitomo Chemical Advanced Technologies, we have enough capacity for the foreseeable future as we make the transition to commercialization.

Understood, that clears things up. So it's really more of a catalyst. And regarding the customers with whom you've broadened those conversations, how would you characterize them relative to the ones you mentioned in the prior quarter? Is there any distinction, whether in terms of what they're focused on, or simply another way of describing them? Thanks.

Yes, as you're aware, we're still engaged across a number of markets, though the AI datacom opportunity is genuinely front and center. We're also working to move our technologies into products, and that covers a range of photonics built on our large diameter wafer platform, such as high-speed photodetectors suited to scale-up and scale-out applications, along with emitter technologies, quantum dot lasers among them. The bulk of near-term opportunities probably lie in what the industry calls micro-detectors and micro-emitters for wide and parallel applications; that refers to our LINX high-speed photodetectors; we're seeing substantial demand for this technology; and the timeline we've discussed, where a significant ramp is expected, is really 2028 and beyond. The figures we're hearing, as roughly cited in our earnings release, are, you know, an aggregate of something like $10.3 trillion for this build-out that began in 2025 and runs through 2032. So we'll continue pointing to that timeline, 2028 through 2032, where we see significant ramps taking place. You know, even as we hear that laser suppliers are sold out, to some degree, we can read that as, you know, supply being more or less fulfilled for the coming year. But a significant ramp is expected in 2028 and beyond, and that timing works well for us on commercialization.

Hi everyone, I appreciate you having me here. Regarding that NRE topic you were just discussing, is there any conversation about upfront payments or capacity reservations, or are these payments tied to milestones instead?

Appreciate that question, Danial. Here's what we can share: looking at our fiscal '27, all incoming revenue — as we mentioned a couple of months back during our fiscal year-end earnings — will be generated by R&D contracts along with commercial NRE. As the year progresses, we anticipate sharing additional detail on those development efforts. Once more, I want to highlight the technologies we are converting into products — high-speed photodetectors along with quantum dot lasers. As for how these arrangements might be structured, they could range from funding development work aimed at advancing the technology for later scaling, to development funding for delivering samples, or to development funding for delivering samples meant for qualification ahead of possible product ramps beginning around 2028. That is all we are able to share right now.

All right, yes. Speaking of that AIXTRON MOCVD reactor you bought, could you tell us about the wafer size it can handle? Does it do 200 mm or 300 mm wafers, and do you have any expectations for the next generation reactor you plan to purchase?