What Jonathan Klamkin actually said on Schwab, the ugliest reading, the more interesting reading, and why everyone is suddenly an expert on gallium arsenide.

There are moments in micro-cap investing when a CEO utters three innocent-sounding words and half the shareholder base reaches for the defibrillator. On August 12, during a brief appearance on Schwab Network, Aeluma CEO Jonathan Klamkin said the company had brought its technologies to large-volume manufacturing environments including “silicon manufacturing, Gallium Arsenide manufacturing.”

For a company whose public identity has been built heavily around integrating high-performance III-V materials with large-diameter silicon manufacturing, those three words landed with a thud. Within days we had a high-profile shareholder exit note, elevated anxiety, and a sudden outbreak of GaAs expertise among people who, until approximately last Wednesday, could not have picked gallium arsenide out of a police lineup.

So let’s do this properly—not as a tribal bull-versus-bear food fight, and not as another round of the bulls are delusional / the bears are bitter. Let’s treat it like a crime scene: establish what was actually said, separate observations from inferences, examine the competing explanations, and identify exactly what management needs to tell us before anybody declares the patient either cured or dead.

Exhibit A: What Was Actually Said

First, watch the interview yourself:

Jonathan Klamkin’s August 12 Schwab Network interview

Here is the critical passage, which I have listened to repeatedly and transcribed carefully:

“We’ve taken those materials, those technologies, that are manufactured on these Indium phosphide substrates that companies are having a hard time procuring now, and we brought those technologies to large-volume manufacturing, like, silicon manufacturing, Gallium Arsenide manufacturing, the type of manufacturing capacity that builds chips for mobile phones and other consumer electronics.”

The market has understandably become fascinated with three words—“Gallium Arsenide manufacturing.” Fair enough. But those three words occurred inside several minutes containing some of the strongest manufacturing language I have heard Dr. Klamkin use publicly, and any serious interpretation of the interview needs to explain all of that language rather than isolating the three words that caused the heartburn.

In the same interview, CEO Klamkin said Aeluma has “transitioned this to a production environment,” is “now transitioning to large-volume production with manufacturing partners,” and has “brought those technologies to large-volume manufacturing.” Earlier, when the Schwab host explicitly asked whether Aeluma was effectively challenging the incumbent approach used by companies such as Coherent and Lumentum by bypassing their InP constraints, Klamkin’s answer was unusually direct: “That’s correct.”

That is the public record. Everything beyond it—including what follows below—is interpretation, inference and probability weighting. There is currently plenty of all three.

Feeding the Bears: The Cuts That Actually Drew Blood

Several bearish observations are legitimate. Ignoring them would be cheerleading, and micro-cap investing already produces enough cheerleading to power a medium-sized municipality.

1. GaAs is genuinely new language. To my knowledge, this was the first time Klamkin publicly framed GaAs manufacturing as part of Aeluma’s commercialization discussion. When a CEO unexpectedly introduces another material system on live television, shareholders are allowed to raise an eyebrow. Preferably one eyebrow; two starts looking panicked.

2. The GaAs ecosystem is already mature. Gallium arsenide is not an undiscovered manufacturing frontier. Experienced manufacturers, particularly in Taiwan, possess decades of process knowledge and existing device capability. If Aeluma is simply entering that ecosystem without differentiated epitaxy, device architecture, performance, integration capability, IP, domestic-manufacturing value or some other defensible advantage, then the scarcity value of the platform becomes less compelling. That is a real question, and management should answer it.

3. The wafer economics are not silicon economics. Much of established GaAs manufacturing remains concentrated on smaller wafer diameters than the 200 mm and 300 mm silicon environments that make Aeluma’s scalability proposition so interesting. If the endgame were simply conventional small-diameter GaAs manufacturing, investors would be entitled to ask what happened to the extraordinary cost and scalability argument associated with large-diameter silicon.

4. Customer qualification remains insufficiently visible. Aeluma has discussed approximately 30 customer engagements, which is encouraging, but “engagement” remains a soft metric. At some point the funnel needs depth rather than merely width: evaluation must become qualification, qualification must become design wins, and design wins ultimately need to become production. Thirty courtships are interesting; eventually somebody needs to get married.

5. The reliability-engineering hire deserves attention. Recent hiring around photonics reliability and qualification naturally raises questions about where specific products currently sit in that process. Hiring the capability is not evidence that qualification has not begun, but neither is it evidence that qualification is substantially complete. Investors need more visibility.

6. The business-development turnover and disappearing LiDAR narrative deserve explanation. A heralded commercial executive arriving and leaving within months is ugly optics during precisely the period when commercialization supposedly matters most, while early Aeluma materials emphasized LiDAR far more heavily than the company does today. Neither development proves technological failure—people leave, markets change, and rational companies reallocate resources toward better opportunities—but shareholders should not have to perform forensic archaeology to understand major changes in commercial emphasis.

Those are legitimate cuts. Now let’s stop bleeding theatrically and inspect the wounds.

Where the Bears May Be Overfeeding

The largest leap currently being made is straightforward: Klamkin mentioned GaAs; therefore silicon must be failing; therefore GaAs is a fallback; therefore commercialization has slipped materially. That chain is possible. What it is not is established.

Klamkin explicitly paired “silicon manufacturing, Gallium Arsenide manufacturing” in the same sentence; he did not say Aeluma was abandoning silicon, replacing silicon, or resorting to GaAs because silicon had failed. Treating the mention as a quiet surrender therefore requires importing a confession that was never actually made. Silicon could certainly be progressing more slowly than investors hoped, and GaAs could be part of the response, but platform semiconductor companies routinely pursue multiple substrates, device architectures and foundry routes, so optionality only becomes evidence of desperation once we know why the additional route was chosen.

The qualification argument deserves similar discipline. GR-468 and related reliability requirements can be lengthy and demanding; nobody serious should trivialize them. But converting the duration of one qualification regime into a deterministic statement that meaningful commercial revenue cannot occur until late 2027 assumes several facts we do not possess: which products require which protocols, what testing has already begun internally or externally, whether customers and manufacturing partners are conducting work simultaneously, whether defense products, lasers, photodetectors and standardized products travel through identical qualification paths, and whether an established GaAs manufacturing route could accelerate qualification for particular applications. An engineering requirement can be precise without the investment inference built upon it being equally precise.

And then there is the manufacturing language itself. If Aeluma were simply stuck in an R&D cul-de-sac, with meaningful commercialization years away and no credible manufacturing path, Klamkin’s repeated statements about having transitioned into a production environment and now transitioning toward large-volume production would be remarkably aggressive language for this particular CEO. Nobody has ever confused Jonathan Klamkin with the semiconductor industry’s P.T. Barnum. If those words merely mean we hope someday to manufacture something in volume, management has created a communications problem of its own making; if they mean what ordinary English suggests they mean, something has advanced materially. Either way, the statements deserve at least as much analytical attention as the words “Gallium Arsenide.”

Four Ways to Read GaAs

Today, anyone assigning 100% probability to one of those four explanations knows more than management has publicly disclosed. Congratulations to them. I don’t.

The GaAs Paradox

This is where the story becomes commercially interesting. GaAs has an established ecosystem, incumbent competition, less attractive substrate economics than silicon and less compelling wafer-scale economics, so the obvious question is: why would Aeluma pursue it at all? The bearish answer is that silicon is taking too long. That answer could be right—but it is hardly the only economically rational explanation.

Imagine, purely as an illustration, that a prospective customer already has access to an established or qualified GaAs manufacturing ecosystem and tells Aeluma: We like your device architecture; give us this product through that manufacturing route now. Should management refuse because shareholders originally modeled silicon? Of course not. Customers do not care what substrate I put into my investment spreadsheet eighteen months ago; they care whether the device works, qualifies, ships reliably, satisfies supply-chain requirements and produces attractive economics.

If GaAs can move particular products toward customers faster while large-diameter silicon continues toward superior long-term scaling economics, GaAs is not necessarily a detour; it could be the entrance ramp. Conversely, if management tells us that GaAs has become necessary because the silicon architecture has encountered material technical or manufacturing obstacles, then the bears will have identified something important. The distinction is enormous, which is precisely why jumping from GaAs exists to silicon failed is premature.

There is another possibility worth considering. Perhaps investors have defined Aeluma’s platform too narrowly. If the company’s genuinely differentiated asset is not simply “III-V on silicon,” but rather proprietary materials, epitaxial and device technology capable of delivering high-performance compound-semiconductor functionality across multiple scalable manufacturing environments, then demonstrating compatibility with both silicon and GaAs could broaden rather than dilute the platform proposition. That is the most bullish interpretation, and it requires considerably more evidence before I would underwrite it—but it belongs in the decision tree.

One phrase from Klamkin is particularly intriguing here. He did not merely mention GaAs; he described “the type of manufacturing capacity that builds chips for mobile phones and other consumer electronics.” Why choose that example? I don’t know, and neither does anyone outside the company who has not been told publicly. Turning it into a secret Apple design win would be message-board fan fiction. But asking what commercial or manufacturing strategy Klamkin was trying to describe is entirely legitimate.

The Sentence That May Matter More Than “GaAs”

I keep returning to one statement:

“Now we’re transitioning to large-volume production with manufacturing partners.”

That sentence arguably matters more than the substrate debate because manufacturing readiness and customer qualification are separate gates. A process can be transferred into a volume-capable manufacturing environment while individual customer products remain in evaluation, reliability testing or qualification; those activities can overlap rather than occurring as a neat sequential relay race.

That possibility would reconcile two facts that otherwise appear contradictory: Aeluma could be advancing materially toward scalable manufacturing while commercial customer qualification remains incomplete. It does not mean large-volume commercial revenue is arriving tomorrow, but it does mean investors need to understand exactly what Klamkin means by “production environment,” what has already been transferred to manufacturing partners, what remains under development, and which gates separate today’s manufacturing status from actual production orders.

The verbs have changed. I want to know why.

The Evidence That Doesn’t Fit the Funeral

A good bearish thesis has to explain the bullish evidence too, just as a credible bullish thesis must explain the yellow flags above. That is where I struggle with the darkest interpretation.

Aeluma’s government relationships have expanded rather than contracted; government programs have increasingly targeted manufacturing, non-InP technology, photonics and strategically important applications, while the company has simultaneously expanded customer engagements, developed manufacturing relationships and hired Jason Taylor into a senior role explicitly oriented toward customer-driven commercialization and manufacturing execution. None of that proves commercial success, but it is an awkward collection of facts to reconcile with the idea that the underlying platform has simply stalled.

Government money, obviously, is not commercial validation. A Pentagon contract does not magically become a hyperscaler purchase order because both organizations own computers. But technical diligence by sophisticated counterparties is not economically meaningless either, and Mister Stockwell himself deserves credit for previously digging into the underlying government documentation and finding that Aeluma had arguably understated the significance of certain awards. That is almost performance art for a micro-cap: most promotional companies can turn a modest feasibility award into the Manhattan Project, while Aeluma occasionally appears capable of receiving something consequential and describing it like reimbursement for toner cartridges.

The same discipline applies to personnel. Bouch’s departure deserves an explanation, but Jason Taylor’s arrival may ultimately matter more. Perhaps there was insufficient commercial work for the former role; perhaps it was performance or personality; perhaps management concluded that the bottleneck had shifted from business development toward program execution and industrialization. We do not know. What matters now is whether Taylor’s appointment corresponds to an organization preparing to execute customer programs—or rearranging deck chairs. The next several quarters should make that distinction considerably easier.

As for LiDAR, the issue is simpler. Early Aeluma materials emphasized it heavily; today AI datacom, defense, mobile and other photonics applications dominate the discussion. That shift deserves explanation, but changing priorities because larger markets emerge is not thesis failure—it is capital allocation. The problem is not that LiDAR became less central; the problem is that shareholders should not need forensic archaeology to understand why.

What I Want Asked on the FY2026 Call

Forget spending twelve minutes debating whether quarterly government R&D revenue missed somebody’s spreadsheet by $400,000. These are the questions that actually affect my estimate of intrinsic value:

1. Is GaAs additive to silicon, sequential to silicon, or substitutive for silicon?

2. Is the large-diameter silicon commercialization roadmap intact and progressing as expected?

3. What exactly does “transitioning to large-volume production with manufacturing partners” mean operationally?

4. How many leading AI-datacom laser and high-speed-photodetector programs have entered formal customer qualification?

5. What major reliability, manufacturing or qualification gates remain before production orders can begin?

6. What differentiated capability does Aeluma bring to an already mature GaAs ecosystem—materials/epitaxial IP, device architecture, performance, integration, domestic manufacturing, or something else?

7. Of the approximately 30 customer engagements, how many have materially advanced beyond initial performance evaluation?

8. What capability does Jason Taylor add that management concluded was necessary at this particular stage of commercialization?

And there is one broader question I particularly want answered:

What changed over the last twelve months that gave management sufficient confidence to use substantially stronger manufacturing language on Schwab than it has historically used with investors?

That question does not require customer names, revenue guidance, or confidential information. Klamkin has already put the language into the public domain.

I simply want to understand what sits behind it.

My Read Before Earnings

I have not changed the core investment thesis; I have changed the questions that must now be answered, which is precisely what investors should do when new information arrives. The bearish interpretation of GaAs is plausible but not established, while the bullish interpretations—a commercialization bridge, an application-specific branch, or evidence of a broader manufacturing platform—are plausible but equally unproven.

What I reject is the peculiar market habit of treating ambiguity as proof of whichever conclusion best matches today’s stock chart. A falling stock does not make GaAs evidence of a failed silicon program any more than a rising stock would make it evidence of a secret Apple design win. Price tells us what investors are doing; it does not tell us what atoms are doing inside a wafer.

For now, the evidence presents a fascinating contradiction. Customer engagements have expanded, government support has become more substantial, manufacturing relationships have advanced, Aeluma has hired specifically around commercialization and program execution, and its historically restrained CEO has publicly described the technology as having entered a production environment and now transitioning toward large-volume manufacturing. At the same time, commercial qualification remains insufficiently visible, management communication has created unnecessary ambiguity, personnel turnover deserves scrutiny, and the first explicit public reference to GaAs raises a legitimate question about how the manufacturing roadmap is evolving.

Both sets of facts are real. The FY2026 call needs to connect them.

I am not here to tell you that the GaAs comment is bullish, and I am equally unwilling to promote it into a quiet admission of defeat without considerably better evidence. Maybe the bears are right and GaAs represents a detour because silicon is taking longer than expected; maybe it is a commercialization bridge, maybe it is application-specific, or maybe Aeluma’s platform is simply broader than investors understood. We should know considerably more soon.

The difference between a yellow flag and a red flag is often just the quality of the next explanation.

Aeluma owes investors that explanation.

I’ll bring the Band-Aids. The bears can bring the tourniquet.

Let’s see what actually bleeds.

Subscribe

If you follow Aeluma, semiconductors, photonics, or asymmetric small-cap situations, subscribe.

I write when I think there is something worth saying—not because the calendar says I owe the internet another article. The objective is to separate signal from noise, challenge both bullish and bearish narratives, and identify the handful of facts that can actually change intrinsic value.

Aeluma is entering a particularly consequential stretch, and I expect there will be considerably more to dissect after FY2026 results and management’s next commentary.

Subscribe if you want the follow-up. The next chapter should be more interesting than the GaAs autopsy.

Disclaimer

Disclosure: I am long Aeluma (ALMU). This article reflects my personal opinions and analysis and is provided for informational purposes only. It is not investment advice or a recommendation to buy or sell any security. I have a financial interest in ALMU and therefore am not an impartial observer. Readers should conduct their own due diligence and reach their own conclusions. Micro-cap securities involve substantial risk, including the potential loss of principal. Unless otherwise indicated, information discussed here is derived from public sources. Any forward-looking statements, scenarios, or interpretations are inherently uncertain and may prove incorrect.

APPENDIX: FULL SCHWAB NETWORK INTERVIEW TRANSCRIPT — AUGUST 12, 2026

The transcript below is provided so readers can review Klamkin’s remarks in full and draw their own conclusions. I have reviewed the interview repeatedly for transcription accuracy. Any emphasis added below is mine.

12 August 2026 - AFTER Market CLOSE - Schwab - Aeluma - Cleaned and Triple-checked Transcript

HOST:

Jonathan, thank you so much for your time. Just explain exactly where you sit in the semiconductor supply chain

Jonathan Klamkin:

Aeluma, thank you very much. Ticker: ALMU, builds transformative semiconductors. We specialize in the high-performance compound semiconductors, but what we do is we’ve invented ways to manufacture those components at scale. So think of components like lasers and photodetectors that are used in specialty applications, but are also being used in this massive AI data center buildup, manufactured by companies like Coherent and Lumentum. There are supply chain issues today: raw materials sourced from China, substrates produced in Japan and in China. There are export bans, so much of the supply can’t come to the U.S. to some of those manufacturers. We use a different technology. We build those same components on other substrates, more mainstream substrates, like silicon that’s used in mainstream microelectronics. So we don’t need those Indium phosphide semiconductors, and again, these are so critical for infrastructure projects like the Massive AI Data Center buildout.

HOST:

Yeah, it’s interesting. You just mentioned Coherent because we heard from that company just in the last few minutes or so, Lumentum is another one we saw today. I mean, these companies have been getting a lot of praise, particularly from Nvidia, which kind of validated their kind of picks-and-shovels positioning here right now. But are you sort of a direct challenge to them? Are you saying, are you sort of engineering a way to sort of bypass the constraints that they face entirely?

Jonathan Klamkin:

That’s correct. We don’t use indium phosphide substrates, the incumbent technology, and you mentioned NVIDIA. Nvidia made multibillion-dollar investments in both of those companies and a few others to secure the supply of critical components and materials. We don’t use those indium phosphide substrates. We’ve invented ways to manufacture those core components, lasers, photodetectors, and so on, on other substrates, like silicon, that aren’t subject to the same supply chain constraints. And so this is, in part, why the government is so interested in partnering with us, and we’ve been working with the government since our inception about five years ago, because our technology not only is valuable for defense and aerospace applications, but to really support this massive AI and high-performance computing infrastructure build-out. So our technology is about supply chain resilience, national security, economic security, and also innovation security. There’s a lot of IP behind what we’ve developed.

HOST:

Right, and so last month, the Department of Commerce announced these letters of intent with seven companies, am I right? For $874 million to accelerate R&D for AI and advanced compute supply chain. You were one of them. What is exactly the result of that? What does that mean?

Jonathan Klamkin:

So, letter of intent means that we are in negotiations with the government right now. They announced those seven companies that they executed letters of intent with, and this cohort of companies that was announced are intended to develop critical technologies for AI and advanced compute. And we’re one of the few photonics companies, part of that cohort. And if you look at the announcement made by the government, they speak to indium phosphide-free substrate technology that we could use to build photodetectors and lasers, so critical to optical networking inside data centers. So we’re very happy to partner with the government. We’ve been working with them, as I mentioned, for a number of years, to build some of these technologies for defense systems, and now they’ve become so critical for AI data centers.

HOST:

How easy is it to do all of this? As a matter of national security, this onshoring, obviously, companies wanting to align themselves with U.S. policy to take on China’s hegemony, but also because of some of the risks involved in doing this offshore. I mean, how easy is it to manufacture this stuff here on US soil and compete with China and not have to rely on China?

Jonathan Klamkin:

That’s a great question, and I would say it’s not very easy at all. Unfortunately, we used to do a lot of this manufacturing onshore, and some of it, you know, sort of went away. I mean, that’s the impact of globalization. And to sort of build some of it back up again does take time. The good news is that we invent a lot of these technologies. We invented, we pioneered a lot of these technologies. We invest so heavily in the development of these technologies. And the more manufacturing we do here, the more secure it will be. Not only because we need the supply; we don’t want to be blocked by export bans and the like, but we need to protect that intellectual property. So, I would say not so easy, but the good news is that our government seems to be very committed to supporting this. And it is nice to see not only investments being made in U.S. companies, but incentives being provided for some of those foreign companies to do the manufacturing within U.S. borders as well. It’s a little easier for us to protect it if the chips are built here.

HOST:

Absolutely. It’s a two-way street, but I guess part of the intent, all those decades ago, to go offshore was because it costs less. I mean, that’s obviously why products haven’t cost as much over the decades, but, you know, with respect to what you’re doing, you’re saying you can manufacture this high-performance stuff in large volumes and at a low cost. How do you do that? How do you keep costs low?

Jonathan Klamkin:

So this technology wasn’t invented overnight. We’ve been working on this for a number of years. It builds on more than a decade of research at some, you know, fantastic top universities in the US, and it’s really been a focused 5 years where we’ve transitioned this to a production environment. And now we’re transitioning to large-volume production with manufacturing partners. We’ve made some announcements, limited announcements, about who some of our partners are. We mentioned Tower Semiconductor in a press release a few months ago. We have the IP and the technology, and that’s so critical. So the good news is that the U.S. has continued to innovate and invent new ways of doing things, and what we’ve done is we’ve taken those materials, those technologies, that are manufactured on these Indium phosphide substrates that companies are having a hard time procuring now, and we brought those technologies to large-volume manufacturing, like, silicon manufacturing, Gallium Arsenide manufacturing, the type of manufacturing capacity that builds chips for mobile phones and other consumer electronics. So for us, it was the combination of building strong IP and identifying and partnering with solid manufacturing partners that we get to go build this out with, and not have to make huge investments in capital equipment.

HOST:

Understood. Jonathan, great to have you on the program. Thanks so much for jumping on with us. Really appreciate your time today. Jonathan Klamkin, there, CEO and director of Aeluma.

This article is not about whether Aeluma succeeds. It is about why the market and the U.S. government are currently evaluating two very different questions.

Today’s announcement is not primarily about the headline dollar amount. It is about what the U.S. Department of Commerce chose to evaluate and potentially fund.

The equity market continues to treat Aeluma as a speculative, low-revenue semiconductor microcap: volatile, lightly covered, execution-dependent, and easy to punish during a news vacuum.

Commerce is looking at something different. It is evaluating whether Aeluma’s manufacturing platform could become a strategically important part of the domestic photonics supply chain for artificial intelligence and advanced computing.

Aeluma announced on July 29 that it signed a Letter of Intent with the Department of Commerce’s CHIPS Research & Development Office for up to $30 million in proposed funding. The contemplated award would support the development and commercialization of Aeluma’s scalable, non-InP semiconductor manufacturing platform for photonics, including high-speed photodetectors and lasers for next-generation AI interconnects, communications, and sensing.

The proposal remains preliminary and conditional. It is subject to further diligence, government approvals, and the execution of definitive award documents. The $30 million should not be treated as cash already awarded.

Even at the LOI stage, however, the announcement provides valuable information about how Commerce is evaluating the technology.

KEY TAKEAWAY

The stock market continues to value Aeluma as a speculative microcap with limited revenue and substantial execution risk. The Department of Commerce is evaluating something different: whether Aeluma’s manufacturing platform can become part of America’s strategic AI and photonics infrastructure.

Those are fundamentally different questions. That divergence is the central point.

What Commerce Is Actually Funding

Strip away the headline and the promotional adjectives, and the government’s priorities become remarkably clear.

COMMERCE REPEATEDLY EMPHASIZED

• Scalable manufacturing

• Manufacturing readiness

• Commercialization

• Large-diameter, non-InP substrates

• AI datacom and advanced computing

• Domestic production

• Supply-chain resilience

• National security

This is industrial-policy language, not merely research language.

Aeluma’s proposed work centers on manufacturing high-performance photonic devices on alternative, large-diameter substrates—up to 300 millimeters—rather than relying exclusively on conventional indium-phosphide substrates.

Traditional InP substrates remain essential to modern photonics, but they also carry practical constraints. They are relatively expensive, limited in manufacturing throughput, and exposed to capacity, geopolitical, and supply-chain risks. Those limitations become more consequential as AI infrastructure demands enormous volumes of optical connectivity, higher bandwidth, and better energy efficiency.

Commerce is not merely evaluating whether Aeluma can produce an impressive laboratory device. It is evaluating whether the company’s architecture could support a more scalable domestic manufacturing model for critical photonic components.

Notice what is absent from the government’s framing. Commerce does not emphasize record-setting device physics, academic novelty, or another laboratory breakthrough. It emphasizes manufacturing.

That shift—from scientific capability to industrial capability—is precisely what makes this announcement consequential.

The proposed funding targets manufacturing capability rather than laboratory capability.

Commerce official Bill Frauenhofer described the contemplated program as supporting a “breakthrough in manufacturing of photonic devices” for compute and communications networks. He tied the work directly to the bandwidth and energy-efficiency requirements of increasingly complex AI workloads.

Government language is often ceremonial. This language is unusually specific. It identifies a manufacturing bottleneck, connects that bottleneck to AI infrastructure, and positions Aeluma’s substrate architecture as a potential solution worthy of substantial technical and financial diligence.

Why This Matters

This is the strongest manufacturing-validation event Aeluma has disclosed to date.

That does not mean the government has validated every technical claim, guaranteed commercial success, or concluded that Aeluma will become a major supplier. It means the company’s core manufacturing thesis has advanced far enough to warrant consideration for a potentially transformative federal program.

The distinction matters because Aeluma has historically been easy to categorize as an interesting photonics research company: quantum dots, photodetectors, lasers, epitaxy, and heterogeneous integration.

The vocabulary has now matured:

• manufacturing platform

• production scaling

• commercialization

• partner sites

• domestic capacity

• AI infrastructure

• supply-chain resilience

That is the linguistic migration from a device-development story toward an industrial-capability story.

Aeluma is no longer presenting the problem merely as:

Can we make a high-performance photonic device?

The larger question is now:

Can Aeluma manufacture high-performance photonics on a platform that scales more efficiently, integrates with established semiconductor infrastructure, reduces dependence on constrained substrates, and supports the volume requirements of future AI systems?

One question produces papers and prototypes.

The other can produce a business.

The Silence Was Not Inactivity

The announcement also changes how the recent news vacuum should be interpreted.

Public-market investors often treat silence as evidence that nothing material is happening. That is understandable in a small company with limited revenue and sparse analyst coverage. When management says little and the stock declines, the market fills the vacuum with its own conclusions.

But a federal funding process of this scale does not materialize overnight.

The LOI presumably followed substantial work involving technical submissions, project plans, budgets, manufacturing milestones, commercialization pathways, government review, and discussions over funding structure. Management could not narrate that process publicly while it was underway.

This does not mean every period of corporate silence conceals a major positive development. That would be wishful thinking dressed up as analysis.

It does mean that Aeluma was advancing a consequential strategic initiative while the public market was largely evaluating the company through stock-price weakness, limited near-term revenue, and an absence of frequent announcements.

Silence did not equal inactivity.

The reference to “Aeluma’s facilities in Goleta, California, and partner sites within the U.S.” is also worth noting. It suggests that the contemplated program extends beyond an isolated internal research project and into a broader domestic manufacturing chain.

The release does not identify those partner sites. Investors should not invent names, production volumes, or customer relationships that have not been disclosed.

Still, the ecosystem language appears deliberate.

The Unresolved Economics

THIS IS THE MOST IMPORTANT OPEN QUESTION.

The strategic implications of the announcement are significant. The economics of the proposed funding remain unknown.

The release states that, in connection with definitive award documents, Aeluma would issue equity securities to the Department of Commerce with an aggregate value equal to the award amount.

That is not a decorative footnote.

It means the proposed package cannot yet be treated as uncomplicated grant capital or as $30 million of value accruing directly to existing shareholders. Depending on the eventual structure, the issuance could materially affect Aeluma’s capitalization and the economics of the transaction.

The definitive documents need to answer several questions:

• What type of securities would Commerce receive?

• How would those securities be valued?

• When would they be issued?

• Would they carry voting rights?

• Would there be transfer restrictions?

• How would milestone payments interact with the issuance?

• Could securities be issued or earned in stages?

• Would the arrangement include clawbacks, repayment obligations, or other conditions?

• What would the fully diluted impact be at the time of issuance?

Until those terms are disclosed, investors should not treat the full proposed award as free capital.

The proper framing is more disciplined:

The LOI represents strong strategic and manufacturing validation, while the capital-structure consequences remain unknown.

Both statements can be true simultaneously.

What This Does—and Does Not—Mean

Government selection is not the same as commercial validation.

Governments fund technologies because they may have strategic importance, national-security relevance, supply-chain value, or long-term economic potential.

Customers buy products because those products meet performance requirements, pass qualification, arrive reliably, scale economically, and solve a real operating problem.

Aeluma still needs to demonstrate:

• repeatable manufacturing

• acceptable yield

• device reliability

• customer qualification

• production economics

• commercial demand

• meaningful revenue

The LOI does not eliminate those requirements. It raises the credibility of the platform while leaving the commercial burden of proof intact.

That is exactly why the announcement matters. The company has not merely received another small research contract tied to a narrow experiment. Commerce is evaluating the manufacturing platform itself—the same platform Aeluma believes can address substrate constraints and scale photonics for AI and advanced computing.

The market is still asking whether Aeluma can become a real commercial semiconductor company.

Washington is asking whether Aeluma’s technology could become part of America’s strategic manufacturing infrastructure.

Those are different questions, but they are no longer unrelated.

What I Will Be Watching

The next meaningful information will come from the documentation and execution—not from another round of celebratory adjectives.

Documentation

• Definitive award documents

• Any related SEC filing or Form 8-K

• Disclosure of the final milestone schedule

• Conditions attached to upfront and subsequent payments

Capital Structure

• Type of securities issued to Commerce

• Valuation methodology

• Timing of issuance

• Dilution mechanics

• Voting rights

• Transfer restrictions

• Clawback or repayment provisions

Execution

• Manufacturing-readiness milestones

• Partner-site responsibilities

• Process-transfer or scale-up progress

• Customer qualification

• Commercialization milestones

• Evidence that the funding accelerates production rather than extending research indefinitely

The $30 million headline will attract attention.

The structure will determine the economics.

The milestones will determine whether the program advances Aeluma toward commercial scale.

Bottom Line

The equity market may continue to price Aeluma as a speculative microcap. Limited revenue, execution risk, thin coverage, and an opaque commercialization timeline give investors ample reason to remain demanding.

But it is no longer credible to dismiss Aeluma as merely an obscure laboratory with an ambitious presentation deck.

The Department of Commerce is evaluating the company’s core manufacturing platform as a potential contributor to U.S. leadership in AI, advanced computing, photonics, and semiconductor supply-chain resilience.

That does not settle the investment case.

It materially upgrades it.

The most important aspect of the announcement is not the headline amount. It is that Commerce selected the exact thesis:

Non-InP substrates. Large-diameter scalability. Domestic photonics production. Commercialization. Manufacturing readiness.

Public markets are pricing today’s financial statements.

Commerce is evaluating tomorrow’s manufacturing capability.

Eventually, those two views converge.

The only question is whether they converge because the market catches up—or because execution falls short.

Updating the Thesis

That is one of the most important disciplines in investing.

Today’s announcement does not eliminate execution risk. Aeluma still must demonstrate repeatable manufacturing, customer qualification, commercial adoption, and meaningful revenue growth. Government validation is not commercial validation.

But good investing is not about waiting until every uncertainty has disappeared. It is about continuously updating the probability distribution as new information becomes available.

From that perspective, I believe today’s announcement is highly significant.

Had this been another modest research grant or SBIR award, I would not view it as thesis-changing. Instead, the Department of Commerce is proposing to invest up to $30 million in the development and commercialization of Aeluma’s manufacturing platform. More importantly, the announcement repeatedly emphasizes scalable manufacturing, manufacturing readiness, commercialization, domestic production, AI infrastructure, and supply-chain resilience.

That tells me the government is evaluating precisely the issues I believe investors should be focused on.

The muted market reaction suggests many investors either stopped at the headline or interpreted this as just another government funding announcement. I believe that materially understates its significance. The headline is not the story. The manufacturing thesis is.

Does today’s announcement guarantee commercial success? Of course not.

Does it materially increase my confidence that Aeluma’s platform has strategic value and is progressing in the right direction?

Yes.

Subscribe

Subscribe to Prati Management

I publish when there is something worth analyzing—not to satisfy a content calendar. The focus is underfollowed companies, primary-source research, manufacturing and commercialization risk, and situations where the market may be mispricing a change in probabilities.

Paid subscribers receive the full research archive and future deep dives as material developments emerge.

Disclosure:

I am long Aeluma (NASDAQ: ALMU). This article reflects my personal views and is for informational purposes only. It is not investment advice or a recommendation to buy or sell any security. I may change my position without notice. Micro-cap securities carry substantial risk, including loss of principal. Readers should conduct their own due diligence.

Deep Fission fabricating and testing its prototype reactor canister. Unlike many advanced nuclear developers still centered on conceptual designs, Deep Fission is fabricating and testing hardware for commercial deployment. Its next major milestone is a commercial-scale borehole proof-of-concept in 2H of 2026.

What the Market Is Discounting

That distinction matters because Deep Fission’s valuation should not be judged only against companies pursuing new reactor architectures or novel fuel cycles. The central question is narrower, but still difficult: whether the company can make underground deployment work at commercial scale.

At roughly a $650 million market capitalization, the market is assigning limited value to a company that has disclosed an 18.5 GW commercial development pipeline and is targeting commercial deployment beginning in 2027. On top of that, Deep Fission is using the most common pressurized water reactor (PWR), which allows them to bypass the criticality tests required for other experimental reactor types.

Deep Fission is not attempting to reinvent reactor physics. It is attempting to solve the two problems that have constrained nuclear deployment for decades: time-to-market and cost at scale.

Three weeks ago, I argued that Deep Fission represented one of the most asymmetric opportunities in the advanced nuclear sector. The IPO has since disappointed; the shares have declined materially, and sentiment has deteriorated. They have not materially changed the underlying investment thesis. Rather than revisit the original investment case, this article focuses on a more important question: What is the market still missing?

Why the Deployment Model Matters

Deep Fission’s approach places a conventional pressurized water reactor roughly one mile underground in a borehole. Management believes this approach could substantially reduce the amount of surface infrastructure and engineered safety systems required relative to conventional reactor deployments.

The company uses standard low-enriched uranium (LEU) fuel—the same fuel already operating in commercial reactors throughout the United States. This could prove to be one of the company’s most important strategic advantages if commercial deployment proceeds as planned.

Much of the valuation assigned to several peers reflects expectations that they will ultimately solve both new reactor technology and new fuel supply chains. Deep Fission is attempting to solve only the deployment problem while relying on reactor technology and fuel that already exist commercially.

This distinction matters for valuation. Scaling from one reactor to many—or even to tens of gigawatts—becomes materially easier when the fuel supply chain already exists at commercial volumes.

The Most Important Near-Term Catalyst

Because the novel part of Deep Fission’s design is the underground deployment rather than the reactor itself, the company’s next major milestone directly addresses its biggest remaining execution questions.

In the second half of 2026, Deep Fission plans to drill and test a larger-diameter commercial prototype borehole. This is a non-nuclear test designed to validate:

• Drilling performance at the required diameter

• Cuttings management

• Casing support

• Prototype reactor emplacement and retrieval

• Sensor integration

• Heat transfer behavior in an actual borehole environment

In other words, this is the first real-world engineering test of the principal integration element that has not already been demonstrated through decades of conventional nuclear operation: the underground deployment model.

Successful results would provide concrete evidence that the system functions as modeled before the company moves to its demonstration reactor.

In my view, this is the milestone that has received the least attention relative to its importance.

Regulatory Path and Timeline

DOE leadership has increasingly emphasized commercial output rather than technical demonstration alone. That policy backdrop matters because Deep Fission’s stated 2027–2028 commercialization roadmap focuses on revenue-generating operations rather than on a standalone criticality milestone. Policy support alone will not be enough if the company fails to execute technically, but an administration focused on accelerating nuclear deployment can reduce regulatory friction, improve confidence in financing, and increase the relevance of companies pursuing near-term commercial operations rather than isolated technical demonstrations.

CEO Liz Muller with U.S. Energy Secretary Chris Wright during the DOE Reactor Pilot Program meeting.

Unlike several peers pursuing separate criticality demonstrations, Deep Fission is participating in the DOE Reactor Pilot Program to advance validation of its commercial deployment approach from the outset. Management believes this could compress the path to commercial operation, though ultimate NRC review timelines remain uncertain. Rather than spending resources on a standalone criticality experiment using a novel fuel form, the company is using proven pressurized water reactor technology and standard LEU fuel, allowing management to focus on demonstrating the principal novel aspects of the system—the borehole deployment model and its associated operations.

U.S. Energy Secretary Chris Wright has repeatedly emphasized the goal of commercial power—not merely criticality demonstrations—by 2027/2028. Based on currently available public disclosures, Deep Fission appears to have one of the more direct stated paths toward that objective because its DOE Reactor Pilot Program work is tied to the reactor configuration it intends to commercialize. That does not eliminate regulatory or execution risk, but it makes the company’s upcoming proof-of-concept milestones especially important.

Why Criticality Is the Wrong Comparison

One reason Deep Fission is difficult to compare with other advanced nuclear companies is that DOE Reactor Pilot Program progress is not necessarily apples-to-apples across the sector. Several participants are pursuing prototype or transitional criticality demonstrations. Those milestones may be valuable, but they do not necessarily represent the same step toward revenue-generating commercial operations.

Deep Fission’s stated path is different. The company is advancing the reactor configuration it intends to commercialize, while focusing near-term work on the borehole deployment model, emplacement and retrieval, heat-transfer behavior, and system integration. That does not eliminate regulatory or execution risk. It does, however, mean that investors should be careful about comparing “criticality” milestones across companies as if they all carry the same commercial significance.

Read more

The Transition Investors May Be Missing

Over the past several months, my research on Aeluma gradually led me to a broader realization about how investors evaluate deep-technology companies.

Most investors spend years asking one question:

Does the technology work?

Eventually, however, the question changes:

Can it be manufactured economically, repeatedly, and at commercial scale?

Those are fundamentally different challenges. I believe some of the largest valuation changes occur during that transition.

In my latest Seeking Alpha article, I introduce a framework describing the three phases of deep-tech commercialization:

• Scientific validation
• Production validation
• Commercial validation

Rather than focusing on any single press release or catalyst, the article argues that investors should pay attention to the gradual accumulation of manufacturing evidence—production hires, foundry partnerships, qualification milestones, government programs, and commercialization progress.

One sentence probably captures the article better than anything else:

No single announcement convinced me. The pattern did—gradually.

Read more

Editor’s note: This post has been revised to remove overlapping language from my Seeking Alpha Exclusive article on Deep Fission, “Deep Fission: Built Faster, Priced Cheaper, Ready Sooner.” This piece discusses the same company and investment thesis but has been substantially rewritten in a different format, structure, and voice. Readers can find the Seeking Alpha article here: Deep Fission: Built Faster, Priced Cheaper, Ready Sooner

Deep Fission caught my attention because it is approaching advanced nuclear from a different angle. The company is not trying to win by inventing the most exotic reactor. It is trying to combine conventional pressurized water reactor technology, standard low-enriched uranium fuel, and subsurface drilling experience into a new deployment architecture.

That distinction matters. In nuclear, investors often focus on reactor design. I am increasingly focused on deployment. Who can get permitted? Who can finance projects? Who can build repeatedly? Who can turn customer interest into actual power generation?

Deep Fission’s answer is unusual. Instead of building a traditional above-ground reactor facility, the company wants to place a reactor deep underground in a borehole. The commercial logic is straightforward: reduce surface complexity, use the subsurface environment as part of the safety architecture, and make deployment more repeatable.

The technical and regulatory questions are also obvious. This approach remains unproven at commercial scale. The company still has to demonstrate field execution, navigate regulatory review, secure financing, and convert customer interest into real projects.

That is why I am watching the next phase closely. The investment case is not about certainty. It is about whether upcoming evidence points can begin to shift the probability that Deep Fission’s deployment model works.

Demand Is the Backdrop, Not the Thesis

The demand case for advanced nuclear is no longer obscure. Data centers, industrial electrification, grid reliability concerns, and reshoring all point in the same direction: the world needs more reliable power.

But demand alone is not enough to make any nuclear company interesting. The harder question is supply. Who can actually bring new generation online within a timeframe customers and capital providers can use?

That is where Deep Fission becomes worth watching. The company is not merely arguing that nuclear power will be needed. It is arguing that deployment architecture can change the timeline, cost structure, and repeatability of new nuclear projects.

That remains to be proven. But it is a more specific question than “will power demand grow?” The market already understands demand. The unresolved question is who can execute.

The Race May Matter More Than the Reactor

The advanced nuclear sector has become increasingly crowded, with numerous companies pursuing different technologies, fuel types, cooling systems, and reactor architectures. Investors spend considerable time debating which design is superior. The market may ultimately care far less. If power shortages emerge as many forecasts suggest, customers are unlikely to prioritize elegant engineering solutions. They will prioritize reliable electricity delivered on schedule.

That distinction is important because most advanced nuclear companies are still targeting broad commercial deployment sometime in the next decade. Deep Fission’s timeline is notably different. The company is targeting commercial operations in the 2027-2028 timeframe, several years ahead of many publicly traded peers. Whether management ultimately achieves that objective remains uncertain, but the potential implications are significant.

The first company capable of deploying commercial advanced nuclear power at scale may enjoy advantages that extend well beyond any individual project. Utilities, hyperscalers, industrial customers, regulators, financing partners, and supply-chain providers tend to gravitate toward demonstrated success. Once a technology is operating commercially, the conversation changes from “Can this work?” to “How quickly can it be replicated?” That transition is often where the largest value creation occurs.

For this reason, I increasingly view Deep Fission not simply as a lower-cost reactor story, but as a potential first-mover story. If the company can validate its borehole approach and execute against its development roadmap, the most valuable part of the investment thesis may not be that Deep Fission eventually builds reactors more cheaply than competitors. It may be that Deep Fission is building reactors while many competitors are still pursuing their first commercial deployment.

Why Being First May Matter More Than Being Best

Investors frequently assume that the advanced nuclear industry will evolve like the semiconductor industry, where technical superiority ultimately determines the winners. History suggests infrastructure markets often behave differently. Railroads, pipelines, cellular networks, cloud infrastructure, and hyperscale data centers were not always won by the technically elegant solution. They were frequently won by the platform that reached commercial scale first and then compounded its advantage through financing relationships, customer adoption, regulatory familiarity, and supply-chain dominance. If power demand develops as many industry participants expect, advanced nuclear may ultimately prove to be less of a science competition and more of a deployment race.

There is another aspect of the Deep Fission story that I believe investors may be underappreciating. The real competition may not be over reactor design. It may be that Deep Fission begins producing commercial evidence while many competitors are still working toward first deployment.

Deep Fission is making a different bet than most advanced nuclear companies. Instead of trying to invent a more exotic reactor, management is attempting to prove a model that could be deployed faster, more cheaply, and at greater scale. If they are right, the market may ultimately care less about which reactor design is theoretically superior and more about which company actually delivers power first.

Why This Is No Longer a Fringe Idea

One reason I became increasingly interested in Deep Fission is that the concept is no longer being discussed solely within niche nuclear circles. The idea of placing reactors deep underground has begun attracting attention from mainstream publications, including Forbes and The Economist, both of which recently highlighted the potential advantages of underground reactor deployment.

From Forbes: Acclaimed Physicist And His Daughter Are Burying Tiny Nuclear Reactors A Mile Underground

From the Economist:

Burying nuclear reactors might make them cleaner and cheaper

An American firm hopes to test the theory

Investors do not need to agree with every claim made in these articles. What matters is that the concept is beginning to move from an obscure engineering discussion to a topic that is receiving broader attention from investors, policymakers, and the energy industry.

The Bottleneck Isn’t Demand. It’s Deployment.

The investment backdrop for advanced nuclear has changed dramatically over the last several years. For decades, electricity demand growth in developed economies was relatively modest and largely predictable. Today, that assumption is being challenged by the rapid buildout of AI infrastructure, hyperscale data centers, industrial electrification, domestic manufacturing initiatives, and the growing power requirements of an increasingly digital economy.

Major technology companies are now discussing power needs in gigawatts rather than megawatts. Utilities across many regions are warning of future supply constraints, while large industrial and data center operators are increasingly seeking reliable long-term electricity sources. The challenge is no longer whether demand exists. The challenge is bringing sufficient new generation online quickly enough to meet it.

This is where advanced nuclear has re-emerged as an attractive solution. Unlike intermittent renewable sources, nuclear provides reliable baseload generation and can operate around the clock regardless of weather conditions. The problem, however, is that many advanced nuclear designs remain years away from commercialization and often require new fuel supply chains, lengthy construction schedules, or significant capital investment.

Deep Fission’s approach is interesting because it attempts to address what may be the industry’s biggest challenge: deployment. Rather than reinventing reactor physics, the company is combining proven pressurized water reactor technology with established drilling techniques in an effort to reduce construction costs, shorten deployment timelines, and accelerate commercialization. Whether management ultimately succeeds remains to be seen, but if their assumptions prove correct, the company’s approach could represent one of the fastest paths to bringing meaningful new nuclear generation online.

How I Think About the Architecture

The reason Deep Fission remains interesting to me is not that every claim in the investor deck must prove correct. Investor decks are selling documents. The useful question is narrower: does the company’s architecture attack the parts of nuclear deployment that have historically made projects slow, expensive, and hard to repeat?

Deep Fission’s model tries to shift part of the burden away from surface construction and into subsurface deployment. That does not make execution easy. It simply changes the problem. Instead of asking whether a new reactor physics concept can work, investors are being asked whether a known reactor type can be integrated into a borehole environment and operated in a repeatable way.

That is a better question for investors because it creates clearer evidence points. Field work, drilling progress, emplacement and retrieval testing, heat-transfer behavior, regulatory movement, customer conversion, and financing progress should all provide real signals over time.

The company still has to prove the model. But if it can, the significance is larger than one reactor. It would suggest that nuclear deployment may be improved not only through better reactor design, but through a different physical deployment architecture.

Why Execution Evidence Matters

Deep Fission’s team and advisory group are part of the reason the story is worth following, but I do not think management résumés are the central issue anymore. In development-stage nuclear, credentials matter, but field execution matters more.

The company has already moved beyond a purely conceptual story. It has selected a Kansas site, advanced work tied to the DOE Reactor Pilot Program, entered into a fuel arrangement with URENCO, and continued building toward its proof-of-concept and demonstration milestones. Those are useful signals, but they are still not the same thing as proving the commercial model.

What matters now is whether Deep Fission can convert planning milestones into physical evidence. That means borehole work, emplacement and retrieval testing, thermal behavior, regulatory progress, customer conversion, and eventually project financing. Each step should either strengthen or weaken the thesis.

That is the part I like as an investor. This is not a story where the market has to wait ten years for the first meaningful datapoint. The company’s own roadmap suggests that investors should begin receiving evidence much sooner. That evidence may be positive or negative, but it should make the investment case less theoretical.

Commercial Interest and Capital Signals

Deep Fission’s disclosed commercial interest is part of what makes the story worth following, but I do not treat LOIs or MOUs as revenue. They are signals of customer interest, not contracts. The company still has to convert interest into financeable projects, and that will require technical progress, regulatory confidence, and credible project economics.

The more interesting point to me is that the company has attracted attention from capital providers and infrastructure-oriented partners. That does not validate the technology. It does, however, suggest that sophisticated investors are willing to spend time evaluating whether this deployment model could eventually support real project finance.

That distinction matters. A nuclear concept can be intellectually interesting and still be commercially irrelevant. For Deep Fission, the question is whether its model can become underwritable. Customers need confidence in timing. Lenders need confidence in construction risk. Infrastructure investors need visibility into repeatability and returns.

If the company can show real field progress, the commercial conversation should become more concrete. If it cannot, LOIs will remain exactly what skeptics say they are: non-binding expressions of interest.

Valuation Is Still a Probability Question

Deep Fission trades as though the market is assigning a low probability to successful execution. That is understandable. This is a development-stage nuclear company with major technical, regulatory, financing, and customer-conversion risk.

What interests me is not that the stock is “cheap” in a traditional sense. It is not. There is no mature revenue base to underwrite. The valuation argument is really about probability: if the company begins validating the borehole deployment model, the market may assign a higher likelihood to the commercial roadmap.

That is where the asymmetry comes from. The market does not need to believe the full bull case today. It only needs evidence that the probability of success exceeds the currently assumed value. The upcoming technical and regulatory milestones should begin testing that question.

What I Am Watching Next

The next phase should be less about investor presentations and more about field evidence.

For Deep Fission, the most important near-term question is whether the company can continue moving from engineering logic to physical validation. That means progress on borehole execution, integration work, regulatory review, and the commercial steps needed to turn interest into actual projects.

I am not looking for every milestone to eliminate risk. That is unrealistic. I am looking for evidence that the deployment model is becoming less theoretical.

The thesis strengthens if field execution confirms that the company can translate its subsurface concept into repeatable operations. The thesis weakens if technical work slips, regulatory progress slows materially, or commercial interest fails to move toward financeable commitments.

Execution Remains the Critical Variable

No serious investor should treat Deep Fission as de-risked. The company is still attempting something that has not yet been proven at commercial scale.

That is the point. The opportunity exists because the market does not yet know whether the deployment architecture works. If the answer were obvious, the valuation would already reflect it.

The main risk is not conventional reactor physics. The main risk is integration. Deep Fission has to show that large-diameter boreholes can be drilled, completed, instrumented, used for deployment, and eventually repeated in a way that supports commercial economics. That is a very different question from whether a pressurized water reactor can operate.

The regulatory path also matters. DOE support is helpful, but it does not remove the need for safety review, site-specific requirements, financing discipline, and commercial execution. Nuclear timelines have a long history of disappointing investors who mistake policy support for completed execution.

The economics also remain projections until they are supported by real-world evidence. Cost, timing, deployment cadence, customer conversion, and financing terms all have to move from presentation logic to project reality.

That is why I view the next phase as a probability test. Each milestone should either increase confidence or expose where the model is weaker than expected.

What Would Change My Mind

The thesis would weaken if borehole testing slips materially, if field work shows that emplacement or retrieval is more difficult than expected, or if the larger-diameter drilling program proves harder to execute than management anticipates.

I would also become more cautious if customer interest does not begin moving toward firmer commercial commitments as the technical work progresses. LOIs are useful signals, but they are not projects.

Financing is another important marker. If the model is as attractive as the company believes, capital providers should become more engaged as the technical and regulatory picture improves. If that does not happen, it would tell me something.

The most important question is simple: does each step make Deep Fission look more like a repeatable infrastructure platform, or does it remain an interesting nuclear concept that is difficult to finance and deploy?

Conclusion

Deep Fission remains one of the more interesting advanced nuclear stories I follow, not because success is certain, but because the evidence should arrive sooner than in many development-stage energy stories.

The company is attempting to solve a specific problem: deployment. If it can prove that known nuclear technology can be combined with subsurface deployment in a repeatable commercial model, the market may reassess the opportunity. If it cannot, the thesis breaks.

That is a clean setup. There are real risks, but there are also identifiable milestones. Investors should not have to wait until the 2030s to learn whether the story is strengthening or weakening.

For now, I remain interested because Deep Fission is not just asking investors to believe in a future nuclear market. It is asking them to watch whether a different deployment architecture can produce real field evidence over the next several years.

Disclosure: I am long FISN and participated in the company’s final private financing round prior to the IPO. This post reflects my personal opinions and is provided for informational purposes only. It is not investment advice.

Subscribe For More Research

Deep Fission is exactly the type of opportunity I spend my time searching for: companies operating at the intersection of major secular trends, misunderstood by the market, and possessing the potential for highly asymmetric outcomes.

Most of my research focuses on identifying opportunities before they become widely recognized. That includes emerging technologies, photonics, AI infrastructure, advanced energy, special situations, and companies where I believe the market is mispricing future optionality.

Paid subscribers receive:

• Full-length investment research reports.

• Deep dives into emerging opportunities before they reach mainstream coverage.

• Ongoing updates to active investment theses.

• Portfolio commentary and position disclosures.

• Access to the full archive of premium research.

The objective is simple: find a small number of exceptional opportunities where the upside materially exceeds the downside.

Thank you for reading and for supporting independent research.

Disclaimer

This article reflects my personal opinions and is provided solely for informational and educational purposes. Nothing contained herein should be construed as investment advice, legal advice, tax advice, or a recommendation to buy, sell, or hold any security.

I am not acting as an investment adviser, broker, or fiduciary. Readers should conduct their own independent research and consult qualified professional advisers before making any investment decisions.

Investing involves risk, including the potential loss of principal. Many of the companies discussed may be early-stage, pre-revenue, speculative, highly volatile, and subject to significant technological, regulatory, financing, and execution risks.

I may own positions in securities mentioned in this article and may buy, sell, or otherwise change my holdings at any time without notice. Any views expressed are current only as of the date of publication and are subject to change without update.

Past performance is not indicative of future results.

If you find this research valuable, please consider subscribing to support future work.

Indium phosphide (InP) is essential for many of the high-performance lasers, modulators, and optical components that power next-generation AI interconnects. Yet native InP substrates come with structural limitations that silicon never faced:

• Small wafer sizes. Standard InP wafers are typically 2–4 inches, with 6-inch wafers only now becoming more widely available. By comparison, silicon manufacturing has spent decades scaling around mature 300 mm (12-inch) wafers.

• High cost and lower yields. InP remains significantly more expensive and difficult to manufacture than silicon due to material and process complexity.

• Concentrated and geopolitically fragile supply chains. A small number of suppliers dominate production, with meaningful exposure to China.

• Long qualification cycles. New suppliers and processes often require years of validation before deployment in mission-critical datacenter environments.

At the same time, demand for optical bandwidth is accelerating rapidly.

200G/lane deployments are underway. 400G/lane platforms are already being prototyped. AI infrastructure roadmaps continue pushing bandwidth requirements higher.

As a result, performance is no longer the scarce resource.

Manufacturing throughput and scalability increasingly are.

This distinction matters because industries are rarely constrained by the thing investors are focused on.

Over the last several weeks, multiple independent signals have converged on the same conclusion:

• Reuters highlighted China’s tightening controls on InP exports.

• Industry analysts began mapping winners and losers based on access to InP supply chains and manufacturing capacity.

• Tower Semiconductor announced a multi-year InP epiwafer supply agreement with IQE, including minimum purchase commitments, to support its silicon photonics roadmap for AI datacenters.

• Engineers and industry participants continue discussing wafer-size limitations, qualification challenges, and supply constraints across the optical ecosystem.

• Research efforts from major industry players continue moving toward heterogeneous integration of III-V materials onto silicon manufacturing platforms.

Taken individually, these developments may appear unrelated.

Viewed together, they tell a consistent story.

The Tower-IQE Signal

The most interesting aspect of the recent Tower-IQE announcement is not that IQE secured a customer.

It is that one of the world’s leading silicon photonics foundries is explicitly describing a future built from high-volume silicon platforms augmented by selectively integrated InP functionality.

Tower's own words underscore the shift: performance and scale are increasingly inseparable.

Notice the emphasis is not simply on performance. Tower specifically highlights the combination of performance and high-volume manufacturing. That distinction appears increasingly important as AI infrastructure scales from millions to potentially billions of optical components.

That framing appears increasingly common across the industry.

The debate is no longer whether III-V materials matter.

The debate is how to scale them.

Every major player appears to be converging on the same core challenge:

How do you achieve InP/GaAs-level performance while maintaining silicon-scale economics, throughput, and manufacturability?

That is fundamentally a manufacturing question rather than a device-performance question.

Why This Matters

For investors, the most interesting opportunities may not be the companies building the fastest individual optical devices.

They may be the companies attempting to solve the manufacturing constraint itself.

One example is Aeluma (NASDAQ: ALMU), which is developing heterogeneous integration of III-V semiconductors directly onto 300 mm silicon wafers — a manufacturing architecture designed to bypass many of the native InP substrate constraints the industry is now highlighting.

The potential significance is the manufacturing architecture.

If the industry’s central challenge increasingly becomes delivering InP/GaAs-level performance with silicon-scale economics and throughput, then approaches that reduce dependence on native InP substrates deserve close attention.

If you found this analysis useful, consider subscribing.

My research focuses on underfollowed opportunities at the intersection of semiconductors, photonics, AI infrastructure, advanced manufacturing, and emerging technologies.

Most of my work begins with a simple question:

Where is the real bottleneck?

Paid subscriptions help support deeper research, industry diligence, conference coverage, management interviews, and long-form investment analysis.

Thank you to everyone who reads, shares, and supports the publication.

Disclosure:

I own shares of Aeluma (NASDAQ: ALMU).

This article reflects my personal opinions and is provided for informational and educational purposes only. It should not be construed as investment advice or a recommendation to buy or sell any security.

Investing involves risk, including the potential loss of principal. Readers should conduct their own research and consult appropriate professional advisers before making investment decisions.

Closing Thought

The next AI bottleneck is shifting from compute performance toward the physical systems required to move, package, connect, and deploy photons at scale.

TL;DR

The next bottleneck in AI is no longer compute. It is the physical infrastructure required to move, package, integrate, and manufacture photonics at scale.

Investors remain focused on device performance, but history suggests the larger opportunity may sit closer to packaging, manufacturing, portability, qualification, and ecosystem orchestration.

The winners may not be the companies with the best demos. They may be the companies that make deployment possible.

=========================================

AI’s next bottleneck is no longer compute. It’s in the physical systems that move, package, and sense photons.

The same bottleneck is about to break two multi-billion-dollar markets at the same time: AI data-center optics and LiDAR-class sensing.

Everyone is still staring at GPUs, memory, and “AI chips.”

That’s already a lagging indicator.

The real constraint is at the layer below, in the physical infrastructure that moves and senses photons.

In my recent free Substack,
“Why Photonics Scaling Is an Orchestration Challenge — Not a Capacity Crunch”
(https://open.substack.com/pub/pratimanagement/p/why-photonics-scaling-is-an-orchestration?utm_campaign=post-expanded-share&utm_medium=web),
I laid out why the bottleneck is shifting from compute → memory → networking → optics, and why this is fundamentally an orchestration problem, not a simple capacity issue.

This piece picks up where that left off.

Because once you accept that framing, the next question is unavoidable:

Where does the system actually break?

In the data center, this shows up as co-packaged and near-package optics hitting power and density limits, and optical fabrics struggling to scale to tens of thousands of links per cluster.

At the edge, it shows up as LiDAR and sensing stacks trying to move from demo fleets to millions of units — with autonomy programs discovering that “cool sensor” is not the same as “manufacturable platform.”

On paper, these look like separate markets.

In practice, they share the same failure mode:

• The same III–V materials

• The same packaging and test bottlenecks

• The same underlying question almost nobody is pricing correctly:

Can this be manufactured — reliably, repeatedly, and at scale?

Most commentary still treats this as a capacity story: add tools, add wafers, add fabs.

That’s the wrong lens.

This is not a supply problem.

It’s a synchronization problem.

And in synchronization problems, the money rarely ends up where the slide decks say it will.

Read more

TL;DR

Aeluma is shifting from a technology-validation story toward a manufacturing-validation story. While investors remain focused on photonic devices and near-term revenue, management increasingly appears focused on foundry integration, process transfer, manufacturing scalability, and addressing potential manufacturing bottlenecks emerging across AI optics, silicon photonics, and quantum technologies.

• Aeluma is increasingly transitioning from a technology-validation story toward a manufacturing-validation story.

• Industry participants, including Coherent, have publicly discussed Indium Phosphide (InP) capacity constraints as optical demand accelerates.

• Aeluma’s large-diameter silicon-compatible manufacturing approach requires dramatically fewer wafers than traditional InP production methods.

• Recent partnerships, executive hires, and management commentary increasingly focus on commercialization, process transfer, qualification, and manufacturing scale.

• The market continues to evaluate Aeluma primarily as a photonics device company, while management increasingly appears focused on building a scalable semiconductor manufacturing platform.

For nearly the past two years, investors have analyzed Aeluma as though it were simply another small photonics company attempting to commercialize an interesting technology. The discussion generally revolved around familiar questions: Does the science work? Can the company attract customers? Will meaningful revenue ever emerge? Those were reasonable questions at the time.

I increasingly believe they may no longer be the right questions.

In fact, investors may be asking the wrong question entirely.

The reason I say that is not because the technology suddenly changed. The conversation surrounding the company changed.

In 2024, Aeluma was primarily a technology story. The debate centered on whether heterogeneous integration on mismatched substrates could work reliably enough to matter. Investors focused on device performance, laboratory results, technical feasibility, and scientific validation. The question was straightforward: Can this be done?

By 2025, however, the discussion began to evolve. Customer engagements expanded, government contract activity increased, and commercial as well as ecosystem relationships began to emerge. At the same time, management’s public commentary increasingly focused on manufacturability, qualification pathways, commercialization initiatives, and the practical challenges associated with bringing advanced photonics technologies into broader production environments.

Those are not the discussions typically associated with a company attempting to prove a scientific concept. They are the discussions associated with a company attempting to commercialize one.

As a result, the debate gradually shifted from a question of technical feasibility to a question of commercial viability. The conversation was no longer simply whether Aeluma’s technology could work. Increasingly, it became whether the technology could be qualified, adopted, and commercialized at a scale large enough to matter.

The debate evolved from:

Can this be done?

to:

Can this be commercialized?

Today, in 2026, the discussion appears to be evolving once again.

Listen to management presentations, earnings calls, conference appearances, and investor discussions.

The language is different.

Increasingly, the conversation revolves around foundry integration, packaging economics, process transfer, manufacturing scalability, supply chain considerations, qualification cycles, and commercialization readiness.

Those are not the discussions typically associated with a company still attempting to prove a scientific concept.

Those are the discussions associated with a company attempting to industrialize one.

Investors should pay attention when a technology discussion becomes a manufacturing story.

Over the past eighteen months, nearly every major development surrounding the company from foundry relationships and commercialization-focused hiring to ecosystem partnerships, qualification initiatives, and manufacturing discussions has pointed in the same direction.

History suggests that some of the semiconductor industry’s most important companies emerged not because they produced the best individual devices, but because they solved manufacturing constraints that became increasingly important as industries scaled.

That observation does not mean Aeluma will succeed.

Commercial adoption remains uncertain.

Qualification risk remains substantial.

Execution remains critical.

But it does suggest investors should at least consider the possibility that the market narrative has not kept pace with the company’s evolution.

The market largely continues to view Aeluma through a product lens.

Over the past eighteen months, the company’s partnerships, hiring decisions, manufacturing initiatives, and public commentary have increasingly centered on scale rather than feasibility. That distinction may ultimately prove more important than the market currently appreciates.

The question may no longer be whether Aeluma can build an interesting photonic device.

The more important question may be whether the optical industry is approaching a manufacturing bottleneck at precisely the moment Aeluma’s architecture becomes commercially relevant.

The InP Bottleneck Is Becoming Strategic

The most important part of the Aeluma story may have very little to do with SWIR sensing, quantum photonics, defense applications, or even individual devices.

Read more

Disclosure: I am long ALMU. This article reflects my personal opinions and is not investment advice. I may buy or sell shares at any time without notice.

I came away from Aeluma’s LD Micro Invitational XVI presentation materially more constructive on the company’s strategic positioning.

Not because management suddenly announced hyperscale customers or explosive near-term revenue. They did not.

And not because the company revealed some dramatic new technological breakthrough. The core technology story has existed for years.

What changed — and what I believe the market is still underappreciating — is the nature of the discussion itself.

The presentation increasingly sounded less like a company trying to convince investors that the underlying physics works and more like a company attempting to solve an emerging manufacturing and supply-chain bottleneck inside next-generation optical infrastructure.

The market still largely views Aeluma as a tiny speculative photonics company with minimal revenue and uncertain commercialization prospects. But after listening carefully to both the recent earnings call and the LD Micro presentation, I believe the company is gradually transitioning into something far more commercially relevant: a scalable heterogeneous integration platform potentially aligned with where AI optical infrastructure, sensing, and advanced photonics manufacturing are heading.

The Discussion Has Shifted

Much of the earlier debate surrounding Aeluma centered on whether III-V materials could realistically be integrated onto large-diameter silicon substrates with commercially viable performance characteristics.

That no longer appears to be the central debate.

Jonathan Klamkin spent remarkably little time defending the underlying science during the LD Micro presentation. Instead, the discussion repeatedly focused on manufacturing scalability, substrate shortages, wafer throughput, packaging infrastructure, supply-chain constraints, deployment economics, and foundry integration.

Importantly, management repeatedly emphasized that customer discussions are increasingly evolving from “prove the science works” toward “can this scale economically and operationally?”

That is probably the single most important takeaway from both the earnings call and this presentation.

A year or two ago, the debate was:
“Can this technology work at all?”

Today, the debate increasingly appears to be:
“Can this technology be industrialized, qualified, and scaled fast enough to matter?”

That is a materially different commercial stage.

The InP Bottleneck Is No Longer Theoretical

One of the strongest portions of the presentation was management’s increasingly direct discussion around indium phosphide (“InP”) supply-chain limitations.

CEO Jonathan Klamkin openly discussed growing InP substrate shortages, limited manufacturing throughput, packaging bottlenecks, and the broader difficulty of scaling traditional architectures fast enough for accelerating AI optical infrastructure demand.

Importantly, this aligns with broader industry developments.

Over the last year, investors have increasingly heard companies like Coherent and Lumentum discussing InP constraints and capacity limitations while NVIDIA and hyperscalers continue pouring capital into optical infrastructure expansion.

The market opportunity here is enormous.

Management referenced more than $300 billion of hyperscaler AI infrastructure spending in 2025 alone, with optical networking representing a critical enabling layer.

The core Aeluma thesis is not simply:
“better photonics.”

Increasingly, the thesis is becoming whether advanced III-V optical functionality can ultimately be manufactured on scalable silicon infrastructure at materially lower cost and far greater throughput than traditional InP approaches.

That is a much bigger idea.

The Manufacturing Economics Are Starting To Look Compelling

Read more

The market reacted to Aeluma’s fiscal third-quarter results as though a speculative microcap photonics company had failed to deliver a meaningful catalyst. Revenue remained small at $1.2 million. Full-year guidance narrowed modestly to $4.2–4.6 million. There were no hyperscaler logos, no blockbuster commercial announcements, and no dramatic hockey-stick guidance designed to satisfy momentum traders searching for immediate monetization signals.

After a rapid move higher in the stock, many investors were clearly positioned for explosive near-term revenue acceleration, a major commercial customer reveal, or another headline-grabbing partnership. They did not get one. The stock reacted accordingly.

Yet listening carefully to the conference call itself produced a very different impression.

The discussion no longer sounded centered on proving whether the technology works. Instead, it increasingly sounded like an industry beginning to recognize that existing optical infrastructure may not scale economically fast enough for what artificial intelligence is about to require.

That distinction may ultimately prove far more important than this quarter’s revenue line.

The Market Focused on Revenue. The Call Focused on Infrastructure Constraints.

Momentum investors wanted immediate commercialization proof.

The call instead focused on manufacturing transfer, qualification readiness, photodiode scaling, foundry integration, optical supply constraints, and the growing limitations of traditional indium phosphide infrastructure.

Ironically, those subjects may matter far more over the next several years.

The central theme was clear: the AI infrastructure bottleneck is shifting.

The first phase of the AI buildout centered primarily on compute. GPUs became a scarce resource. Capital flooded toward accelerators, memory, and networking hardware supporting ever-larger training clusters.

Increasingly, however, the constraint appears to be migrating toward optics, interconnects, photodiodes, packaging, substrates, and scalable manufacturing itself.

Traditional indium phosphide supply chains were never designed for the volumes AI may ultimately demand. Substrate availability is constrained. Expansion timelines are long and expensive. Packaging complexity is rising. Geopolitical concentration adds another layer of risk. Meanwhile, requirements continue accelerating toward 200G- and 400G-per-lane architectures, co-packaged optics, and denser interconnect environments.

That context matters enormously when analyzing Aeluma.

Because what sounded most important on the call was not a tiny quarterly revenue figure. It was the repeated emphasis on customers evaluating the company’s technology to address near-term supply gaps.

Why This Call Felt Different

The most striking aspect of the call was not promotional enthusiasm but operational maturity.

Management increasingly spoke as though the industry is converging on the exact problems Aeluma was originally built to solve.

Historically, many investors viewed Aeluma as an early-stage photonics science project. Prior calls largely centered on technological capability, research validation, and future opportunity.

This quarter’s discussion felt different.

The emphasis repeatedly returned to manufacturing transfer, foundry compatibility, qualification, repeatability, ecosystem integration, customer prioritization, and commercialization sequencing.

Those are not the dominant concerns of a company still trying to prove the science.

They are the concerns of a company attempting to industrialize it.

That distinction materially changes the nature of the story.

Importantly, the conversation around customer engagements also appeared to evolve meaningfully. Management indicated active engagements have grown to roughly thirty, with increasing participation from AI datacenter and optical infrastructure customers. More notably, many of these discussions now appear to carry clearer technical requirements, operational objectives, and commercialization pathways than in prior periods.

Again, none of this guarantees imminent commercial scale.

The posture is shifting from “please evaluate our technology” toward “how quickly can the ecosystem qualify and deploy this?”

The Nearer-Term Opportunity May Be More Practical Than Investors Realize

One of the most underappreciated elements of the call may involve photodiodes.

Investors naturally focus on the most futuristic portions of the story: quantum dot lasers, heterogeneous integration, nonlinear photonics, and next-generation optical architectures.

Those opportunities may ultimately become extremely important.

But the nearer-term commercial bridge may prove more practical and immediate.

High-speed photodiodes and detector arrays already exist throughout the optical networking infrastructure today. They are understood components with established deployment pathways. The challenge is accelerating demand while supply remains constrained.

That distinction matters because photodiodes are generally easier to qualify and integrate than entirely new laser architectures. Reliability, thermal management, and qualification hurdles are often less severe than for higher-power laser systems.

In practical terms, photodiodes may provide a realistic pathway toward earlier commercial adoption while the broader platform opportunity continues maturing.

The call strongly suggested management understands this dynamic.

The emphasis on standard products, qualification readiness, customer-specific requirements, and accelerating demand increasingly sounded like a company preparing for repeatable production opportunities rather than purely exploratory development work.

This may ultimately become one of the most important bridges between today’s modest revenue profile and tomorrow’s larger opportunity.

Aeluma Increasingly Looks Like a Platform Company

The market still appears to value Aeluma primarily as a speculative component supplier.

Increasingly, however, management’s language suggests something much more strategic may be emerging: an enabling manufacturing and integration architecture aligned with the future needs of optical infrastructure scaling.

That distinction is profound.

Semiconductor history shows that companies enabling ecosystem scale often become strategically more important than individual component suppliers.

This quarter’s call repeatedly reinforced that possibility.

Aeluma is not attempting to vertically integrate every aspect of semiconductor manufacturing internally. Instead, the company increasingly appears focused on leveraging proprietary materials expertise, scalable silicon-compatible manufacturing approaches, foundry partnerships, and flexible commercialization pathways, including direct supply, licensing, and ecosystem integration.

This flexibility aligns precisely with the varying requirements across AI datacom, mobile short-wave infrared sensing, defense, and quantum.

Viewed through that lens, Aeluma increasingly resembles a platform architecture aligned with multiple secular bottlenecks simultaneously.

That is a much larger opportunity than a single niche photonics component.

The Market May Still Be Using the Wrong Valuation Framework

This may be the core disconnect.

The market continues treating Aeluma as an early-stage component vendor and valuing it based on current revenue and near-term visibility.

That framework may prove increasingly incomplete.

The latest call suggested a company attempting to position itself as a scalable infrastructure enabler within optical networking, AI interconnects, and future photonic integration architectures.

Those are fundamentally different categories.

Infrastructure-enablement platforms are often valued not on current revenue but on ecosystem importance, manufacturing leverage, strategic scarcity, and long-duration optionality once commercialization credibility begins to emerge.

That does not mean Aeluma has already achieved that transition.

Execution risk remains real. Qualification cycles remain lengthy. Incumbent competitors remain powerful and deeply entrenched. Commercial scaling in semiconductors is rarely linear.

But the market may increasingly underestimate how important scalable alternatives could become if traditional optical manufacturing approaches continue to face economic and operational constraints amid AI-scale demand growth.

That possibility appears to be moving closer to the center of the story.

The Revenue Delay May Matter Less Than Investors Think

The guidance adjustment largely reflected timing delays on government contracts rather than lost demand. The company still secured six new contracts totaling roughly $5 million this fiscal year, hitting its target ahead of schedule.

Cash remained strong at $37.8 million, with only modest sequential decline despite increased hiring, manufacturing scale efforts, ecosystem expansion, and continued R&D investment. Notably, the at-the-market facility has seen little or no utilization despite periods of elevated trading volume and significantly higher share prices — a clear sign of disciplined capital stewardship and long-duration shareholder alignment.

The company had every opportunity to aggressively issue stock while the market (and its stock) was very strong. It chose not to.

Given the already strong balance sheet, recent non-dilutive funding awards, and relatively controlled burn profile, management appears increasingly focused on preserving strategic flexibility rather than maximizing short-term dilution capacity.

Why the Capital Structure Matters

Investors debating whether ALMU should trade at $10, $20, or $30 may be anchoring on the wrong variable.

There are only roughly 18 million shares outstanding.

That capitalization becomes exceptionally important if Aeluma ultimately emerges as a strategically relevant participant within the AI optical infrastructure stack.

Most investors psychologically anchor to current revenue. But infrastructure transitions often cause markets to reprice companies long before mature financial statements fully reflect their strategic importance.

The asymmetry here exists precisely because the market still appears deeply divided on what category Aeluma actually belongs in.

If this remains merely a small speculative photonics company, current skepticism may prove justified.

But if Aeluma is instead evolving into a scalable manufacturing and integration platform aligned with emerging optical bottlenecks, then the capitalization structure itself becomes unusually important.

That is why sophisticated technical investors continue paying close attention despite the company’s still-small financial profile.

The Bigger Picture

The AI buildout is not a one-year event.

Hyperscaler capital expenditures are already moving into the hundreds of billions annually, with optical networking, interconnects, and photonic integration becoming increasingly central to the next phase of infrastructure scaling.

Traditional photonics architectures are beginning to encounter real economic and operational limitations under that demand profile.

The latest Aeluma call increasingly sounded like a company positioning itself directly at the intersection of those constraints.

• Not recklessly.

• Not promotionally.

• And not with exaggerated near-term promises.

Quite the opposite.

The discussion was measured, operationally focused, and centered on manufacturing readiness, qualification progress, the quality of customer engagement, commercialization sequencing, and scalable deployment pathways.

That may ultimately prove far more important than whether quarterly revenue was $1.2 million or $1.5 million.

The coming quarters will not eliminate execution risk. Qualification cycles remain real. Revenue visibility may continue frustrating short-term traders.

But after listening carefully to this call, the more important question increasingly appears to be whether the market is still analyzing Aeluma as the company it was rather than the company it may be becoming.

Because the discussion no longer sounded like a tiny photonics startup searching for relevance.

It increasingly sounded like an industry beginning to search for scalable alternatives to infrastructure constraints that are rapidly becoming impossible to ignore.

Subscribe

This publication focuses on high-signal analysis across AI infrastructure, optical networking, semiconductors, synchronization economics, and asymmetric technology transitions.

Subscribers receive:

• Deep-dive infrastructure research

• Earnings and industry analysis

• Optical networking and photonics frameworks

• Semiconductor supply-chain positioning

• Ongoing coverage of distributed AI infrastructure and synchronization economics

The goal is simple: identify where bottlenecks, scarcity, and value migration are occurring across the AI stack before those shifts become broadly understood by the market.

No noise. Just signal.

Disclaimer

This content is provided for informational and educational purposes only and should not be construed as investment advice or a recommendation to buy, sell, or hold any security.

The views expressed are solely my own and are based on publicly available information, industry research, and personal analysis. I currently hold positions in companies mentioned throughout this publication, including Aeluma (ALMU), and may buy or sell securities at any time without notice.

Investing in equities — particularly small-cap and emerging technology companies — involves substantial risk, including the potential loss of principal. Readers should conduct their own due diligence and consult with a qualified financial advisor before making any investment decisions.

No representation or warranty is made regarding the accuracy, completeness, or timeliness of the information presented. Forward-looking statements and projections are inherently uncertain and subject to change without notice.

Aeluma reports fiscal Q3 2026 results tomorrow after the close (May 13, 2026). I would caution investors against overemphasizing near-term revenue or expecting a dramatic commercial inflection point from this particular quarter.

The stock has become extremely volatile, reaching fresh all-time highs near $31 ($30.95 today) amid sharp intraday swings (currently trading ~$27 per share). With a relatively small float and growing speculative attention around photonics and heterogeneous integration, even modest commentary could drive outsized moves in either direction. That volatility increases the risk that investors focus on the wrong metrics.

The key question is not whether Aeluma suddenly posts meaningful revenue tomorrow.

The key question is whether the company continues progressing from promising technology into repeatable, manufacturable semiconductor infrastructure.

Many semiconductor companies can produce compelling lab demonstrations. Far fewer successfully achieve repeatable, foundry-compatible manufacturing flows on standard 200mm and 300mm silicon infrastructure. Aeluma’s heterogeneous integration approach is designed to bridge that gap.

That transition is rarely linear, and it almost never reveals itself through a single quarter’s financials.

What has changed over the past several months is the broader evidence stack surrounding the company. Aeluma has continued adding experienced semiconductor talent, including Willy Rachmady as VP of Strategic Partnerships, alongside personnel with deep process and manufacturing backgrounds. The company has also expanded its ecosystem relationships through partnerships and engagement with organizations such as Tower Semiconductor and Sumitomo Chemical.

Importantly, several recently announced government contracts tied to quantum dot lasers and AlGaAs materials were specifically connected to scaling efforts involving Tower and Sumitomo. Those announcements may appear incremental individually, but collectively they reinforce the view that the company is steadily advancing toward manufacturable integration rather than remaining confined to isolated technical demonstrations.

Just as importantly, management’s commentary has increasingly shifted toward integration, scalability, repeatability, yield improvement, and foundry compatibility rather than purely technical validation.

That distinction matters.

It is also worth noting that Aeluma disclosed both the Tower and Sumitomo relationships long after the underlying engagements had already begun. That reflects a notably understated and non-promotional management style. Investors expecting aggressive investor relations campaigns, flashy projections, or exaggerated commercialization timelines will likely continue to be disappointed. That has never been how this company communicates.

For this quarter, I believe investors should focus less on headline revenue and more on commentary surrounding process maturity, customer engagement, qualification progress, scaling efforts, repeatability, yield improvement, and integration into standard semiconductor manufacturing flows.

My expectation is that the quarter itself will likely appear fairly uneventful financially. That is not necessarily bearish.

Aeluma also remains my largest position, and I do not intend to trade around this quarterly report. My focus remains on whether the underlying infrastructure transition continues progressing over a multi-quarter horizon rather than reacting to short-term volatility or a single earnings release.

Over the next several quarters, the real indicators of progress will likely involve evidence of repeat customer engagement, qualification milestones, manufacturing consistency, and broader integration into commercial semiconductor workflows.

At this stage, the real signal is whether the infrastructure transition continues advancing quietly beneath the surface.

That remains the core question.

Subscribe

This publication focuses on high-signal analysis across AI infrastructure, optical networking, semiconductors, synchronization economics, and asymmetric technology transitions.

Subscribers receive:
• Deep-dive infrastructure research
• Earnings and industry analysis
• Optical networking and photonics frameworks
• Semiconductor supply-chain positioning
• Ongoing coverage of distributed AI infrastructure and synchronization economics

The goal is simple: identify where bottlenecks, scarcity, and value migration are occurring across the AI stack before those shifts become broadly understood by the market.

No noise. Just signal.

Disclaimer

This content is provided for informational and educational purposes only and should not be construed as investment advice or a recommendation to buy, sell, or hold any security.

The views expressed are solely my own and are based on publicly available information, industry research, and personal analysis. I currently hold positions in companies mentioned throughout this publication, including Aeluma (ALMU), and may buy or sell securities at any time without notice.

Investing in equities — particularly small-cap and emerging technology companies — involves substantial risk, including the potential loss of principal. Readers should conduct their own due diligence and consult with a qualified financial advisor before making any investment decisions.

No representation or warranty is made regarding the accuracy, completeness, or timeliness of the information presented. Forward-looking statements and projections are inherently uncertain and subject to change without notice.

Introduction: The Market Still Thinks This Is About GPUs

The market still believes the AI infrastructure trade is primarily about compute. That view is already obsolete.

The first phase was compute acquisition: GPUs, accelerated servers, HBM memory, and massive training clusters. That phase created NVIDIA.

The second phase is far larger and structurally more difficult: moving data, synchronizing data, powering data, cooling data, and connecting geographically distributed compute.

This is the Synchronization Economy — the industrial architecture required once AI compute fragments across physical and geographic limits. It is where optical networking, photonics, advanced packaging, thermal systems, co-packaged optics, optical circuit switching, multi-rail architectures, and indium phosphide infrastructure become existential.

Most investors still speak the language of GPU count. Hyperscalers now speak the language of power availability, interconnect density, bandwidth scaling, cluster synchronization, optical movement, thermal constraints, packaging complexity, and distributed training fabrics.

The AI bottleneck has migrated outward from compute. That changes the entire semiconductor hierarchy.

And one company telegraphed this shift earlier and more clearly than almost anyone else: Ciena ($CIEN)

Part I — The March CIEN Call That Quietly Changed Everything

In March 2026, Ciena delivered one of the clearest early signals of the Synchronization Economy. Most investors heard “strong networking demand.” Management was describing something more profound: the shift from single-facility compute clusters to geographically synchronized compute fabrics.

The clearest signal was the repeated emphasis on “scale-across.” Hyperscalers are hitting hard physical limits inside single facilities — power availability, thermal density, cooling constraints, land, and transmission bottlenecks. The logical response is geographic fragmentation of AI workloads, creating distributed training and inference fabrics that require low-latency, high-bandwidth optical synchronization across campuses and regions.

Ciena made clear that the future AI “datacenter” is not one building. It is a geographically synchronized compute fabric.

What does that physically require?

It requires coherent optical transport systems, high-capacity switching and routing hardware, optical line systems, digital signal processors (DSPs), network timing and synchronization layers, optical control software, and bandwidth orchestration platforms capable of moving enormous volumes of data between geographically separated compute clusters with extremely low latency and high reliability.

This is where companies like Ciena operate. The company is not merely selling ‘networking equipment.’ It is selling the transport and synchronization architecture required to make distributed AI infrastructure function as a coherent system.

Part II — The Migration of Scarcity and the Birth of the Synchronization Economy

AI infrastructure is experiencing a classic bottleneck migration.

Every new compute node creates a disproportionate increase in interconnect complexity, synchronization overhead, thermal load, and bandwidth demand.

Compute was scarce → then HBM → advanced packaging → power delivery and cooling.

Now the constraints are moving outward into the optical stack: synchronization, photonics, bandwidth movement, thermal architectures, and packaging precision.

Why synchronization scales exponentially
Compute scales linearly. Interconnect complexity scales exponentially. As AI clusters grow, every additional unit of compute requires dramatically more synchronization, bandwidth, and optical movement to maintain coherence across distributed workloads. This is not a linear bandwidth problem; it is a combinatorial synchronization problem. The larger the system, the more data must move, the more precisely it must be timed, and the more power/thermal overhead it generates. Movement economics begin to rival — and eventually can dominate — compute economics. This is the Synchronization Economy in action.

AI infrastructure is no longer merely a compute problem.
It is increasingly a synchronization problem.
And synchronization at civilization scale requires optics.

What mattered about the subsequent earnings cycle was not merely strong demand. It was where the strongest demand appeared, where shortages intensified, where hyperscalers committed long-duration capital, and where management teams unexpectedly converged around synchronization, optical density, thermal constraints, and packaging complexity. The calls collectively revealed the same pattern: scarcity is migrating outward from compute into the synchronization and movement layers of AI infrastructure.

Part III — COHR, LITE, and FN Confirm the Thesis

The three subsequent calls provided powerful, independent confirmation.

Scale-Up vs. Scale-Across: The Two Engines

• Scale-up = denser, higher-bandwidth interconnects inside a facility or rack (CPO, higher-speed transceivers, advanced packaging).

• Scale-across = optical fabrics that synchronize across facilities/campuses (pump lasers, narrow-linewidth lasers, WSS, multi-rail, coherent transport, OCS).

Both are accelerating simultaneously, and both are severely supply-constrained.

Coherent (COHR)
Coherent delivered the clearest and most complete picture of the full optical supercycle. The company sits across more layers than any peer: InP lasers and photonic chips, 800G/1.6T transceivers, OCS systems, multi-rail scale-across architectures, CPO (lasers + ELS modules + fiber attach), and new thermal/XPU cooling solutions using its industrial materials platform.

Management quantified >$20 billion incremental SAM from the new growth vectors. They reported record backlog, orders extending into calendar 2028, and LTAs stretching to the end of the decade that include customer upfront CapEx commitments and minimum demand guarantees. The 6-inch InP ramp is ahead of plan: doubling next quarter (one quarter early), with yields already beating legacy 3-inch lines. First 6-inch devices are already shipping in transceivers and contributing to both revenue and gross margin. OCS is ramping in parallel facilities after a resolved bottleneck; scale-out CPO revenue begins H2 CY2026 and scale-up CPO in H2 CY2027. The NVIDIA multibillion-dollar equity investment and multi-year supply agreement through 2030 is a powerful strategic validator. Most tellingly, management stated outright that fiscal 2027 growth should exceed fiscal 2026 growth.

COHR increasingly looks like a vertically integrated optical infrastructure platform.

Lumentum (LITE)
Lumentum confirmed the supply-demand imbalance is worsening, not easing. The company highlighted EML supply-demand gaps >30% (worse than prior guidance), even tighter constraints on pump lasers and narrow-linewidth components, and “sold-out” conditions across critical categories. Non-GAAP gross margin reached 47.9% (+1,270 bps Y/Y) with strong pricing power. Demand is broadening across transceivers, coherent optics, optical switching, and external laser systems.

LITE is the high-margin scarcity monetizer sitting upstream in the most constrained nodes of the stack.

Fabrinet (FN)
Fabrinet’s call may have been the most underappreciated. As the precision manufacturer and assembly layer, FN is the toll road for AI optical industrialization. Management openly discussed Datacom revenue running “well below demand” due to laser, ASIC, and memory constraints, while highlighting aggressive capacity expansion (Building 10 adding ~$3B revenue potential with ~6-month payback at full utilization). Direct hyperscale transceiver programs are already shipping, merchant programs are ramping, and CPO/OCS packaging work is deepening.FN benefits regardless of which specific optical architecture wins. Its Thailand-scale cost advantage, fungible capacity, and operating leverage make it a uniquely high-quality business model.

Part IV — Understanding the AI Optical Stack

The ecosystem now resembles a vertically integrated industrial system. Value accrues to the layers that solve the hardest bottlenecks.

Layer 1 — Compute
NVIDIA (NVDA) remains the orchestrator. But future AI economics are migrating outward.

Layer 2 — Networking Silicon
Broadcom (AVGO) and Marvell (MRVL). Broadcom’s architectural moat in switching, custom ASICs, and high-performance connectivity is exceptionally durable.

Layer 3 — Optical Components and Photonics
Coherent (COHR), Lumentum (LITE), and others. This layer controls optical density, coherent interconnects, external lasers, and scaling architectures. It is rapidly becoming strategically central.

Layer 4 — Packaging and Manufacturing
Fabrinet (FN) and advanced packaging ecosystems. This may be the most underestimated layer. Photonic packaging is extraordinarily complex: sub-micron alignment tolerances for fiber-to-chip coupling, precise thermal coupling to prevent signal degradation, extreme yield sensitivity at scale, and the need for hermetic sealing and active alignment in high-volume production.

Thermal challenges compound exponentially as synchronization density rises. Rack power density is climbing into the hundreds of kilowatts, concentrating heat in optical components that are highly temperature-sensitive. Optical heat concentration creates new failure modes: signal drift, reduced laser efficiency, and accelerated aging. Traditional air cooling becomes insufficient; liquid cooling and advanced thermal interface materials are now required, but they introduce their own coupling, reliability, and manufacturability issues. As optics move closer to the compute (CPO, multi-rail, scale-across), thermal management and photonic packaging become existential bottlenecks.

Layer 5 — Foundries and Specialty Materials
TSMC, Tower Semiconductor (TSEM), Sumitomo Chemical, and InP substrate providers. Indium phosphide is the critical enabler for high-speed photonics, coherent systems, and advanced lasers. TSEM provides the foundry platform for scaling heterogeneous integration; Sumitomo supplies the high-quality InP wafers and process technology essential for compound semiconductor maturity. As the optical stack expands, the importance of specialty manufacturing, process maturity, and scalable InP fabrication rises materially.

Who Wins the Economics?
The highest returns accrue to companies that combine scale + vertical integration + hyperscaler trust + manufacturing depth. Pure IP players without these attributes risk commoditization as the ecosystem industrializes.

Part V — The Hyperscaler CapEx Explosion and Power-Driven Fragmentation

Microsoft, Meta, Google, and Amazon are building civilization-scale compute systems. This is driving hundreds of billions in annual AI infrastructure investment, regional AI campuses, distributed training fabrics, and optical synchronization architectures.

Power infrastructure is a key accelerant. Regional power fragmentation — availability, grid constraints, and transmission limits — is forcing hyperscalers to build distributed AI campuses rather than ever-larger single facilities. In effect, hyperscalers are becoming internal optical telecom operators, designing their own scale-across fabrics to synchronize compute across power-constrained geographies. This creates multi-year visibility for the Synchronization Economy.

Part VI — Why Past Telecom Optics Cycles Failed — And Why This One Is Different

Past optical cycles ended in overcapacity, commoditization, and painful margin collapse. This cycle differs in four structural ways:

1. Demand is tied to durable hyperscaler monetization — inference workloads, enterprise AI deployment, and model scaling economics.

2. The ecosystem is far more consolidated, favoring vertical integration, manufacturing scale, and hyperscaler relationships.

3. Supply constraints remain severe and early-to-mid cycle (widespread shortages, customer-funded LTAs, backlogs into 2027–2028).

4. The use case is fundamentally different: AI requires not just bandwidth, but low-latency synchronization and power/thermal efficiency at civilization scale.

This is not another telecom optics cycle. It is the industrialization of the Synchronization Economy.

Part VII — Risks and Specific Failure Modes

Despite the bullish backdrop, sophisticated investors must map the specific ways this thesis can fail:

• Hyperscaler internalization: Large players may bring more optical assembly, or even laser production, in-house, thereby compressing supplier margins.

• Synchronization efficiency surprises: Software or silicon-level innovations could reduce the need for raw optical bandwidth faster than expected.

• AI ROI compression or slower inference monetization: If enterprise adoption or model economics disappoint, capex could pause or reset.

• Networking standardization and commoditization: As architectures mature, pricing power may erode, and margins normalize.

• Overcapacity by 2028–2029: If multiple players successfully ramp InP, CPO, and OCS simultaneously, temporary supply gluts could emerge.

• Execution risk on massive simultaneous ramps: InP scaling, OCS, CPO, multi-rail, and packaging all happening in parallel leaves little room for missteps.

Importantly, even if the broader Synchronization Economy thesis proves directionally correct, equity outcomes across the stack are unlikely to be linear. Violent drawdowns, valuation compression, hyperscaler digestion periods, temporary overcapacity, and sharp rotations between compute, networking, optics, and packaging layers should be expected. Secular infrastructure transitions often create extraordinary long-term winners while simultaneously producing extreme volatility along the path.

Part VIII — Constructing a Long-Term Portfolio

Core Holdings

• NVIDIA (NVDA): Foundational orchestrator.

• Broadcom (AVGO): Architectural moat in networking silicon.

• Coherent (COHR): Highest-conviction pure-play optical infrastructure name — broadest exposure, explicit FY27 acceleration, and vertical integration.

• Fabrinet (FN): Manufacturing toll road with exceptional risk/reward on capacity expansion.

• Ciena (CIEN): Underappreciated scale-across / distributed AI networking leader.

Higher-Risk Optionality Layer
AEVA, ALMU, AAOI, and emerging photonics names. Real tailwinds exist, but execution risk is high.

Part IX — The ALMU Question

Aeluma (ALMU) is one of the more intellectually interesting speculative plays in photonics. The company focuses on heterogeneous integration of III-V compound semiconductor materials onto silicon-based photonics platforms for advanced sensing and high-speed communications. Its architecture is designed to enable high-performance photonic integration while eliminating dependence on traditional small-diameter indium phosphide wafer manufacturing approaches.

Rather than relying on fragile standalone InP wafer scaling, Aeluma is pursuing a differentiated approach centered around InGaAs-on-silicon integration and heterogeneous material scaling directly onto larger silicon substrates. If scalable in production, this could materially improve manufacturability, integration economics, and compatibility with broader silicon photonics ecosystems.

Its relationships with Tower Semiconductor (foundry/scaling partner) and Sumitomo Chemical (materials and compound semiconductor ecosystem exposure) position it within the broader heterogeneous integration and advanced photonics themes increasingly validated across the optical infrastructure stack. Recent non-dilutive U.S. Government contracts further support development of its photonics platform for AI infrastructure, sensing, and defense-related applications.

Part X — The Next Five Years

AI compute is fragmenting geographically. The resulting synchronization challenge may become one of the largest infrastructure buildouts in technology history — a multitrillion-dollar, decade-long transition.

The future of AI is not merely generated.
It must be moved.

Optics, photonics, thermal systems, packaging, and synchronization fabrics are becoming central to that movement.

Conclusion

The market initially treated AI infrastructure as a compute story. It is becoming clear that the larger story is the Synchronization Economy — the industrial architecture of distributed machine intelligence.

Value is shifting toward the layers that solve the hardest movement, thermal, and packaging bottlenecks.

The companies best positioned for the next decade increasingly possess scale, vertical integration, hyperscaler trust, manufacturing depth, and optical specialization.

The optical stack is no longer peripheral to AI.
It is becoming central to it.

And the market is only beginning to understand the implications.

Subscribe

Most investors react to narratives after price discovery.

This publication focuses on identifying structural shifts before consensus fully forms:

• bottleneck migration,

• capital cycle inflections,

• synchronization infrastructure,

• optical networking,

• hyperscaler architecture,

• and asymmetric positioning across the AI stack.

Subscribers receive:

• deep-dive industry frameworks,

• earnings call synthesis across the optical ecosystem,

• high-conviction infrastructure analysis,

• portfolio positioning insights,

• and early identification of where value is migrating next.

Separate signal from noise before the market reprices the implication.

If you want to understand where AI infrastructure is actually going — not where financial media says it is going — subscribe.

Disclaimer

This publication is for informational and educational purposes only and does not constitute investment advice, financial advice, or a recommendation to buy or sell any security.

The views expressed are personal opinions based on publicly available information, industry research, and independent analysis. I currently hold positions in all companies mentioned in this article and may buy, sell, increase, reduce, or exit positions at any time without notice.

Investing involves substantial risk, including potential loss of principal. Readers should conduct their own due diligence and consult qualified financial professionals before making investment decisions.

No representation or warranty is made regarding the accuracy, completeness, or timeliness of the information presented.

The market may be mispricing where risk sits in AI photonics.

The core uncertainty is less about whether the technology can work and more about whether it can scale in a repeatable, manufacturable way. Many investors have not fully adjusted to that shift, which is why companies like Aeluma are still often framed as early-stage experiments, even as the underlying story has evolved.

Aeluma appears to be valued more like an earlier-stage concept than a company that has begun assembling the infrastructure required for scale.

That gap is what makes the situation interesting.

From Breakthrough to Business

In this space, the difference between a breakthrough and a business is not device performance alone. It is foundry compatibility, packaging, yield, qualification, and the ability to build an ecosystem that customers trust at scale. This is where many photonics efforts struggle to translate promise into durable outcomes. It is also where the most meaningful value tends to be created when execution is successful.

The remaining question is not whether devices can be demonstrated, but whether they can be qualified, reproduced, and integrated into real production environments.

Recent events in the market have reinforced this distinction.

POET Technologies’ collapse following the cancellation of purchase orders by Marvell and Celestial AI was not a random shock. It was a failure at a stage where this industry has historically broken down. The issue was not simply whether the underlying technology could function. It was whether it could be translated into a trusted, repeatable, and commercially viable manufacturing pathway within a relationship-driven ecosystem.

When that trust breaks down, large customers tend to preserve optionality and move on.

That dynamic highlights what actually matters at this stage of the cycle. Technical promise alone is not enough. Execution, credibility, and integration into real production environments determine outcomes.

This is where Aeluma’s positioning begins to stand out.

This contrast highlights a broader dynamic in the space.

In photonics and advanced semiconductor systems, outcomes are often determined as much by how companies operate as by what they build.

Management Approach

Jonathan Klamkin’s approach to building Aeluma reflects a style that is easy to overlook in a market conditioned to reward promotion. He has not attempted to force recognition or accelerate the narrative ahead of the underlying progress. Instead, the company’s communication has been measured, selective, and often understated relative to the significance of the work being done.

That restraint is not accidental. In a field where credibility is built through repeatability, qualification, and long-term partnerships, disciplined communication tends to mirror disciplined execution. Companies that ultimately succeed at scale are rarely the ones that lead with aggressive claims. They are the ones that build quietly, align with credible partners, and allow external validation to accumulate over time.

Klamkin’s background and technical depth reinforce that posture. His focus appears to be less on shaping perception and more on solving the integration challenges that determine whether a platform can move from promising demonstration to manufacturable reality.

That does not guarantee success. But it is consistent with how durable companies in complex semiconductor ecosystems are typically built.

Rapidly Evolving Transition

The company’s progress over the past year reflects a steady accumulation of elements that matter for long-term commercialization. This is increasingly less about whether the technology can work and more about how effectively it can be scaled.

First, strategic relationships supporting scale

Aeluma has established relationships that directly support the transition from lab-scale demonstration to manufacturable production. Tower Semiconductor brings process discipline and foundry credibility. Sumitomo Chemical strengthens the upstream materials and epitaxy layer, which is often where III-V efforts encounter real constraints. These are not cosmetic partnerships. They address known bottlenecks in the commercialization path.

Second, targeted external validation

The company has begun to receive targeted external validation through government and NASA-funded programs. The recent $4M+ contracts tied to quantum dot lasers and high-speed datacom applications are not generic research grants. They are focused on advancing the exact areas — monolithic integration, SOAs, and PDK-level infrastructure — that tend to limit scalability.

Third, platform-level technical validation

Technical progress has started to validate the broader platform approach. The August 2025 ACS Photonics paper demonstrated, for the first time, monolithic integration of III-V quantum dot photodetectors and lasers with silicon nitride waveguides using pocket heteroepitaxy on silicon photonics coupons (fully compatible with 300 mm CMOS flows). The same layer stack serves both devices. This isn’t just another working prototype; it points to a manufacturing-friendly pathway that eliminates germanium photodetectors and supports multifunctional active components at scale.

Reference: https://pubs.acs.org/doi/full/10.1021/acsphotonics.5c01299

Reference: https://www.aeluma.com/news-media/blog/detail/21333/aelumas-quantum-dot-laser-for-silicon-photonics-technology

From Devices to Platform

What makes this particularly relevant commercially is that Aeluma is not pursuing isolated device wins. It is attempting to solve integration at the platform level. That distinction matters because the market ultimately rewards repeatability, cost structure, and manufacturability. If the platform approach holds, the commercial leverage extends beyond a single product and into a broader set of applications and customers — from AI optical interconnects to defense and sensing.

Taken together, these developments form an evidence stack that looks meaningfully different from a typical early-stage photonics narrative.

The Rachmady hire signals a shift

The recent appointment of Willy Rachmady, Ph.D., as Vice President of Strategic Partnerships and Ecosystem adds another layer to that picture. This is not a symbolic hire. A 20+ year Intel veteran and Principal Engineer who shaped roadmaps in advanced CMOS, 3D wafer stacking, heterogeneous integration, and III-V/compound semiconductor platforms, with more than 350 U.S. patents and multiple Intel Top Inventor Awards, Rachmady brings precisely the expertise needed to accelerate foundry qualification, build commercial ecosystem relationships, and translate technical capability into production reality.

Companies tend to make hires like this when the problem has clearly shifted from proving feasibility to aligning foundries, packaging, customers, and partners into a coherent path toward volume manufacturing.

That suggests a different phase of development.

That shift is also visible in the broader talent build across the organization:

Note: No additional hires identified from recent public disclosures or LinkedIn scans.

Taken together, this begins to look less like incremental hiring and more like coordinated preparation for scale.

Risk is evolving, not disappearing

None of this eliminates execution risk. The transition from promising technology to repeatable manufacturing remains difficult, and outcomes are not guaranteed. But the nature of the risk appears to be evolving. It is less about feasibility and more about whether the company can execute against the constraints that have historically limited the space.

That distinction is often where markets lag.

Macro tailwinds

The broader backdrop continues to strengthen. Goldman Sachs’ recent report on optical networking highlighted a potential 9x TAM expansion to $154 billion as AI infrastructure moves from single-chip performance to multi-chip networking at rack scale. Optical interconnects and silicon photonics are becoming increasingly central to data center scaling and next-generation bandwidth requirements. As systems become more distributed and power constraints intensify, the role of efficient, manufacturable optical solutions expands accordingly.

Aeluma is not the only company positioned to benefit from these trends, but its emphasis on wafer-scale compatibility and ecosystem integration places it in a part of the stack where durable value can emerge if execution holds.

What matters next

The more important question is when the market begins to recognize that progress.

Investors should focus less on headline narratives and more on signals of continued execution. The next proof points will come from qualification progress, partner conversion, and evidence that the platform is moving into repeatable production flows.

If the company continues to build along these lines while attention remains focused elsewhere, the repricing could occur on a different timeline than most expect.

Closing

In semiconductors, the biggest gains are often made after the technical question is answered but before the market fully understands the manufacturing one.

Aeluma appears to be entering that transition.

Subscribe

Subscribers get:

• Early identification of structural risk

• Positioning frameworks before consensus shifts

• High-conviction analysis across the optical stack

If you want to stay ahead of where this market is going:

Subscribe.

No noise. Just signal.

Disclaimer

This content is for informational and educational purposes only and should not be construed as investment advice. The views expressed are my own and are based on publicly available information, industry research, and personal analysis.

I may hold positions in companies mentioned, including Aeluma (ALMU), and may buy or sell securities at any time without notice. Investors should conduct their own due diligence and consult with a qualified financial advisor before making any investment decisions.

No representation or warranty is made as to the accuracy or completeness of the information presented. This content does not constitute a recommendation to buy, sell, or hold any security.

For a long time, Aeluma has been a story about possibility.

Compound semiconductor performance on silicon.
Larger wafers.
Lower cost.
Access to markets that were previously constrained.

That was always the pitch.

But markets don’t pay for pitches.
They pay for manufacturing reality.

And that’s where today’s announcement matters.

Not because of the $4M in government funding.
That’s helpful, but not the point.

The point is this:

Aeluma just made its path to scale more concrete.

This Is Not About Funding. It’s About Integration

The two names that matter here are:

• Tower Semiconductor

• Sumitomo Chemical Advanced Technology

This is one of the clearest signals yet of how Aeluma intends to scale through its supply chain.

Not theory. Not “strategic partners.”
Actual participants in the manufacturing stack.

That changes the conversation.

Because at this stage, the only question that matters is:

Can this platform survive contact with scale?

Where These Stories Usually Break

Every photonics company looks good in the lab.

Performance works early.
Demos impress.
Specs look competitive.

Then scale shows up—and everything fractures:

• Yield collapses under packaging

• Integration introduces variability

• Qualification timelines extend

• Systems stop being repeatable

That’s where most “platforms” die.

Not because the physics fails.
Because the manufacturing system does.

Read more

This is Part 5 of a 5-part series on where the AI infrastructure bottleneck is actually moving—and how that reshapes the competitive landscape.

By now, the pattern should be clear:

The bottleneck in AI is not compute.
It’s the ability to scale the physical layer—reliably, repeatedly, across multiple partners.

Parts 1–4 mapped the framework. Part 5 names the casualties.

Now we get to the uncomfortable part:

Which companies are exposed—and why most investors won’t see it until it’s too late.

Because every cycle has the same trap:

The names that look strongest early—
the best demos, the cleanest narratives, the highest expectations—

are often the ones that break first when scaling becomes the constraint.

Because scale is where physics replaces narrative.

🔒 Paid Begins

Read more

The market will not announce this shift.
It will leak into earnings transcripts first.
Then guidance.
Then estimates.

By the time it’s obvious, positioning is already gone.

Most investors wait for confirmation.
They get none.

What they get instead is a quiet change in language:

“Strong demand” becomes “unable to meet demand.”

That’s the moment that matters.

Not when it’s explained.
When it first appears.

“The market doesn’t price constraints when they’re explained. It prices them when they leak.”

🔒 Paid Begins

Read more

In Part 1, I laid out the core shift: that photonics scaling is not a capacity problem, but an orchestration problem—where coordination across materials, process, and packaging becomes the real constraint.
https://pratimanagement.substack.com/p/why-photonics-scaling-is-an-orchestration?r=6ild8

In Part 2, I moved down the stack to where that orchestration breaks in practice—the physical layer—and how that’s beginning to separate early winners from noise.
https://pratimanagement.substack.com/p/the-next-bottleneck-in-ai-part-2?r=6ild8

Now in Part 3, the focus shifts to what the market actually cares about:

Who captures value—and who gets left behind—as this bottleneck tightens.

Most capital will be positioned in the wrong layer when this becomes obvious.

The bottleneck in AI is moving from compute → to the physical layer.
And the constraint is not capacity. It’s coordination.

The mistake is treating this as a capacity problem.
It isn’t.
It’s a coordination problem under scale.

Now comes the part that actually matters:

Who wins. Who loses. And why most investors will be positioned incorrectly when it becomes obvious.

Because cycles always repeat:

Early → everyone owns the “best technology”
Late → only a handful of companies can actually ship at scale

The market rewards performance first.
It rewards manufacturability later.

The money is made in between.

Read more