The market is pricing Sivers Photonics as a marginal specialty foundry. The order book, however, is pricing like a bottleneck monopoly. That divergence is not a minor anomaly. It is the entire trade.
I have spent the better part of the last decade reading supply chains the way other people read tea leaves. When a company reports supply constraints and rising average selling prices while its valuation barely moves, the first question is not “Is it cheap?” The first question is “Which layer of the stack does the market actually understand?”
Sivers Photonomics (SIVE) is a London-listed III-V compound semiconductor and silicon photonics foundry. It is not a logic chip maker. It does not compete on 3nm or 5nm. Its competitive battlefield is the junction between indium phosphide lasers and silicon photonic waveguides. That is the precise location where AI data centers will be won or lost in the coming years. Co-packaged optics (CPO) and external laser sources (ELS) are the two growth vectors. Both depend on the kind of active-passive integration that Sivers does well.
Yet Sivers trades like a company waiting for permission to exist. The market assigns a price-to-sales multiple in a range that would be generous for a slow-growing tool company but stingy for a supplier sitting directly under a multi-billion-dollar CPO expansion curve. The Discord between physical demand and financial valuation is not a failure of information. It is a failure of framing.
The core insight is simple: the silicon photonics industry is about to hit a wall that only InP active integration can break through. And Sivers is one of the few pure-play foundries that can actually deliver the InP to silicon photonics bridge at scale. The constraints are real. So is the competition. But the market is focusing on the wrong constraints.
Let me start with the technology, because everything else follows from the physics.
InP Integration Is the Moat That Nobody Talks About
Silicon photonics is a tool for moving photons. But photons cannot be generated efficiently in silicon. The bandgap is indirect. The material simply does not emit light well. That is why every serious silicon photonics platform needs a light source. For years, the industry tried to make do with external lasers. That worked for pluggable transceivers where the laser could sit outside the package. But as we move toward co-packaged optics, the external laser becomes a system-level headache. You need to couple light from a discrete laser into the package, then into the waveguides. Each coupling interface costs power, alignment tolerance, and reliability.
The better path is hybrid integration: wafer-bond or epitaxially grow indium phosphide on silicon, then process both the active laser and the passive waveguide on the same substrate. This is not theoretical. Intel, Broadcom, and TSMC have all invested heavily in hybrid silicon photonics. But they are all giant IDMs or foundries with revenue bases that make photonics a rounding error. Sivers has no such luxury. It must win on the basis of process know-how.
Based on the public disclosures and the market commentary around Sivers, the company's core process advantage lies in InP epitaxial growth and the integration of active InP devices with silicon photonic passive components. This is the single hardest part of CPO. The yield curves are unforgiving. InP lasers need precise epitaxy, precise lithography, and a process that does not destroy the silicon photonics layer when you add the III-V layer. A two-degree temperature excursion can shift the emission wavelength. A slight lattice mismatch can create dark-line defects. The fact that Sivers has any yield at all in this domain is a testament to years of accumulated engineering.
But the market cannot see this. It sees a small foundry with limited revenue and a CEO who talks about strategic pivots. It does not see the technical library of process recipes that allow Sivers to co-design a laser with a CPO fabricator like Ayar Labs.
The Competence Trap: TSMC Is Not the Real Benchmark
Many analysts point to TSMC's COUPE platform and claim Sivers will be crushed. That is a misunderstanding of the market structure. TSMC is a giant of logic manufacturing. Its silicon photonics platform will be excellent at integrating photonics into a standard CMOS process flow. But TSMC does not have decades of InP epitaxy know-how. It will need to partner or license that capability. Sivers, by contrast, has InP active integration as its core competency. The problem is scale, not technology.
In a world where CPO volumes explode to tens of millions of units per year, TSMC will likely become the high-volume supplier for the mass market. Sivers will likely remain the specialty house that delivers the high-reliability, high-performance InP light engines that go into the first-generation CPO designs. That is not a bad place to be. Specialty foundries in the III-V space, like IQE, have survived by serving niches that the giants ignore. Sivers' edge is that its niche is the heart of the AI connectivity war.
The real risk is not TSMC. It is the ability of Sivers to maintain yield discipline as it scales. InP integration yield at 70-85% today might improve with process maturation, but it is structurally lower than CMOS logic because the physics is messier. If Sivers cannot get yield above 90% for its CPO-qualified devices, unit economics will remain fragile. The company needs to convert its current in-demand products into a self-sustaining manufacturing loop.
Capacity Is the Strategic Bottleneck
Sivers appears to be running at high capacity utilization. The market commentary explicitly mentions supply bottlenecks and rising ASPs. That combination means demand exceeds supply. In any normal industrial pricing framework, that is the moment to expand. But capacity expansion for a photonic foundry is not like buying a new line of extruders. It means committing to new cleanroom space, new epitaxy tools, and a multi-year qualification cycle with every customer.
The company's situation is complicated by the fact that it potentially operates two wafer fabs: one in Sweden and one in the United States. Sweden has historically been the center of gravity. But the strategic logic now points to the United States. Not because American engineers are superior, but because the customers are there. Ayar Labs, the high-profile CPO startup with Intel heritage, is American. The largest cloud operators and networking equipment buyers are American. And the U.S. government wants photonics supply chains inside its borders. The CHIPS Act money is not just for logic; it is for advanced packaging, photonics, and anything else that supports AI infrastructure.
If Sivers continues to allocate its best engineering resources to Sweden while the sales traction is in the U.S., it will face two separate misalignments: a time zone misalignment with customers, and a capital allocation misalignment with its valuation story. The senior management needs to treat the U.S. fab not as an option but as the logical endpoint of its current order book. Speed is an illusion if the exit door is locked.
The capacity question also has a capital expenditure component. Building or expanding a wafer fab is expensive. The market already senses this. Sivers trades at a high price-to-sales multiple, but its return on invested capital is likely below its cost of capital. That is normal for a growth-stage specialty foundry, but it creates valuation vulnerability. If the company announces a large equity raise to fund U.S. expansion, the stock will likely drop in the short term, even if the long-term narrative is positive. The market has little tolerance for dilution in a company that is not yet solidly profitable.
Customer Concentration Is the Hidden Claw
Every deep dive on Sivers should start with the honest assessment that its customer base is dangerously concentrated. The public data suggests that the top five customers account for an estimated 70-80% of revenue. A single CPO customer like Ayar Labs could represent 20-30% of revenue. That is not inherently fatal, but it is a structural risk that is easily underestimated in a bull market.
What makes customer concentration especially tricky in photonics is the qualification cycle. Switching a foundry is not like switching a software vendor. A CPO module design that has been qualified with Sivers' laser architecture cannot simply be moved to another fab without a total redesign. The switching cost is enormous. That is a moat, but it is also a leash: Sivers is tied to the success of its customers. If Ayar Labs stumbles, Sivers stumbles. If the broader CPO adoption timeline slips by a year, Sivers' revenue visibility slips with it.
There is evidence that Sivers is trying to diversify. The mention of six new pluggable transceiver customers suggests an intentional move to broaden the revenue base beyond the CPO pioneers. Pluggable transceivers are the bridge between the current 800G/1.6T generation and the CPO future. They will generate cash flow while CPO matures. That is a sound strategy. But it also means Sivers is split between two markets: one that pays today (pluggable) and one that will pay tomorrow (CPO). The market hates mixed narratives because it forces the analyst community to argue about which part of the story deserves the multiple.
Supply Chain Dependencies and Geopolitical In-Betweenness
Sivers is not directly exposed to the hardest export controls. Its process uses DUV lithography, not EUV. The equipment is available to non-Chinese companies without special licensing. Its core material is indium phosphide substrates, which are predominantly sourced from Japan. The InP substrate market is concentrated, but InP is not a dual-use item that triggers the same national security panic as advanced logic tech. The practical risk of an immediate supply cutoff is low.
However, the broader geopolitical picture is not zero. Chinese gallium and germanium export restrictions do not directly hit InP, but they signal that materials supply chains can be weaponized. If the U.S. and China continue to escalate over AI infrastructure, CPO could be pulled into the export control net. Sivers, as a British/Swedish company, would likely be in the “friendly” column. But its relationship with O-Net, a Chinese optical module manufacturer, is an interesting counterbalance. The ELS product collaboration with O-Net gives Sivers access to the Chinese market but also creates a perception problem for U.S.-centric investors.
Serenity, the activist or advisor mentioned in the original analysis, seems to be pushing Sivers toward the U.S. market. That is not just about investor relations. It is about derisking the country mix in the customer base. If Sivers can credibly claim that the U.S. is its primary growth market, its valuation should be benchmarked against U.S. photonics and AI infrastructure names rather than European small-cap technology stocks. That is a meaningful multiple expansion opportunity.
Yet I would caution against the naïve story that a geographic pivot is sufficient. The fundamentals still need to work. Sivers' return on invested capital is likely in the 3-5% range, below the weighted average cost of capital. That is not a sustainable steady state. The market will eventually demand evidence that the order book translates into cash flow. If the U.S. expansion requires heavy capex, the financial leverage will rise. The company will need to execute flawlessly in production, not just in presentations.
The Valuation Puzzle: Why the Market Refuses to Pay for the Future
Let us talk about the actual numbers. Sivers is estimated to trade at a price-to-sales of 5-8x. That is high for a foundry that is barely profitable. IQE, its closest competitor, trades at around 3-4x sales. A premium is justified because Sivers has a higher technical position in InP active integration and a richer backlog. But a 2x premium to IQE implies a phenomenal confidence in execution.
The contra argument says the valuation is still cheap if you discount the CPO business as a call option. If CPO reaches 20-30% penetration of the high-speed optical module market by 2028, a meaningful portion of that growth will flow through the InP photonic engine supply chain. Sivers is one of the few independent foundries that can serve that market without vertically integrating with an end-customer. That gives it a partner-of-choice status for multiple system vendors. In contrast, Intel and Broadcom will primarily use their photonics internally. TSMC may open its COUPE platform to customers, but its laser strategy is still being built. Sivers has more optionality than its larger competitors in terms of who it can sell to.
The market is essentially being asked to decide between two narratives. The bear narrative: Sivers is a small, capital-hungry foundry with a concentrated customer base and thin margins. The bull narrative: Sivers is the only pure-play frontier foundry for the InP hybrid integration that CPO requires. Both narratives are true. The outcome depends on timing. If CPO adoption accelerates in 2025-2026, Sivers will likely be supply-constrained and will have pricing power. If CPO slips to 2028-2029, the company will burn cash while waiting for the curve to arrive.
In my experience, when a company has a real technological moat and a real market opportunity, the biggest risk is not competition; it is the willingness of the management team to make the hard strategic choices. Sivers has to decide whether it wants to be a diversified specialty III-V foundry or the definitive InP-to-silicon bridge for the AI datacenter. The latter is more valuable. The former is safer. It cannot be both successfully because the process development roadmap and the sales org structure are different.
The Contrarian Angle: The American Pivot Is a Double-Edged Sword
Everyone seems to agree that Sivers should pivot harder to the U.S. market. I am less convinced that a geographic pivot is the highest-value action. The market's headquarters location affects the conversation, but it does not change the photonics physics. The real strategic imperative is to convert the existing order book into a repeatable foundry process. That is a technology execution problem, not a real estate problem.
Moreover, the U.S. is becoming a political football for all semiconductor-related public companies. A Brexit-related UK company with a Swedish production base might find that American capital appreciates its technology but resents its lack of domestic production. The CHIPS Act favors companies that commit to U.S. fabs and U.S. workers. If Sivers wants to win U.S. investors, it will need to show a clear plan for U.S.-based manufacturing. That raises the capex question and the dilution question. Logic prevails, but bias hides in the edge cases.
The edge case here is the possibility that Sivers simultaneously announces a U.S. fab plan and a large equity raise. The optics of that could be bad because the company's existing investors will see their ownership stake diluted just before the biggest market opportunity in photonics in a generation. That could create a sell-the-news event. The more prudent path would be to secure a strategic investment from a U.S-based customer or a financial partner before committing to large capex. A cozy relationship with Ayar Labs is not the same as having a strategic equity investment from a hyperscaler's venture arm.
Another contrarian point is that the pluggable transceiver business might be more valuable than the CPO business in the immediate future. The market is excited about CPO because it is a paradigm shift. But pluggable optical modules are still the workhorses of data centers. The 800G upgrade cycle is running at high volumes, and 1.6T is just around the corner. Sivers' new pluggable customers could provide a stable revenue base that funds the CPO bet. The problem is that investors often discount product lines they consider transitional. If Sivers is too successful in pluggables, the market might label it as a legacy supplier. If it is too focused on CPO, the near-term financial metrics will be weak. This is the classic growth vs. profitability tension.
Management needs to communicate both tracks with equal clarity. The revenue from pluggables buys time. The technology roadmap for CPO buys the future. The market hates ambiguity, so the narrative must be binary: we are building the InP integration foundry for the AI era, and we are using current pluggable revenue to finance that journey. That is a clear, investable story.
Sivers vs. the Big Boys: The Competition Matrix
Let me quantify the landscape. TSMC is expected to scale its COUPE photonic engine platform to production status around 2025. Intel has already built silicon photonics into its product lineup. Broadcom is active in CPO. All three have significant resources. But each has an Achilles heel. TSMC is a logic foundry first; it lacks deep InP epitaxy internally. Intel has the technology but historically has not opened its process to outside customers. Broadcom develops in-house solutions for its own networking products. None of them are independent specialty foundries.
IQE is a III-V epitaxy foundry, but its focus is broader, and its photonics business is not as tightly aligned with the CPO market as Sivers. Companies like Marvell and Ayar Labs are designers, not foundries. The real competitive threat might be a vertical integration move by a hyperscaler. If Microsoft or Google decides to build its own photonics supply chain, it could wash away any independent foundry. That is why Sivers must lock in customer relationships and long-term supply agreements now.
The good news for Sivers is that the photonics talent pool is small, and the process know-how for InP integration is not codified in a textbook. It is embedded in the team's individual judgment. Sivers has an experienced team. But the risk of key-person dependency cannot be ignored. If the chief epitaxy engineer leaves, the company loses a decade of learning. That is an off-balance-sheet risk that is hard to price.
The Financial Reality Check
Sivers' financial profile is the reason the valuation cannot be argued purely on narrative. The company is likely generating gross margin in the 30-40% range. That is acceptable for a specialty foundry, but not impressive. Net profitability is marginal. Operating cash flow is probably weak, and free cash flow is likely negative due to capex. The return on invested capital is below the cost of capital. Unless the company achieves higher capacity utilization and better pricing, its financial model does not stand on its own.
The order book, however, suggests that pricing power is improving. Supply bottlenecks and rising ASPs are not just hopeful statements; they reflect a demand overhang. The six new pluggable customers, the O-Net ELS partnership, and the Ayar Labs CPO relationship all point to a diversified backlog. The question is how much of that backlog converts to revenue within the next two years. In the semiconductor world, qualification cycles can take 18 months or more. A customer name in the press release does not equal revenue in the next quarter.
For investors, the right approach is to model Sivers as a long-duration optionality. The base case is that CPO becomes a meaningful part of the data center landscape by 2027-2028. In that scenario, Sivers will likely see revenue growth acceleration and margin expansion. The bear case is that CPO disappoints due to reliability issues, thermal management complexity, and a lack of industry standardization. In that scenario, Sivers will continue to rely on pluggables, and the valuation might converge to a lower multiple.
The Risk Register: What Could Go Wrong
The highest-priority risk is simply competitive displacement from TSMC. If COUPE becomes a production-ready platform that directly integrates InP lasers via a partnership with a European or Japanese III-V house, Sivers could lose its unique position. The second risk is customer concentration cascading into revenue volatility. The third risk is dilution from a capital raise for U.S. expansion. The fourth risk is an execution failure in yield ramp. The fifth risk is a broader AI infrastructure spending slowdown that pushes CPO deployment to the right.
Each of these risks is manageable, but managing them all simultaneously in a small-cap company is not easy. The management team has to make choices with limited resources. That is why I believe the market's valuation is actually rational: it reflects a wide distribution of possible future outcomes. The share price will respond violently to quarterly updates that tip the scale in either direction.
Speed is an illusion if the exit door is locked. Sivers can grow rapidly, but if the exit door is a liquidity event on the London AIM or a sale of the company, the valuation geography is constrained. A move to the U.S. listing would open the door to deeper capital pools and higher multiples. But the lock on the door is the company's ability to hit its revenue milestones.
Forward-Looking Judgment: The Semiconductor-Scale Answer
The next 12-18 months will define Sivers Photonics. If the company can demonstrate that its InP integration process is ramp-ready for CPO applications, and if it can expand capacity without completely diluting existing shareholders, then the valuation gap to its U.S.-listed peers will likely close. If it fails on either front, the stock will continue to be a minority-interest idea with occasional spikes on headlines.
I believe the technology is real, but the execution risk is high. The market is correctly waiting for proof. The honest analysis is that Sivers is a high-risk, high-reward story with a genuine technological asset. The payoff depends on whether the management team can make the proverbial right choice when every choice involves trade-offs.
The question that matters is not whether Sivers is a good company. It is whether the market's current valuation has already priced in the failure scenario. Given the modest market cap and the scale of the CPO opportunity, I suspect the market is still underpricing the technical moat. But the moat will not protect the company from a weak balance sheet. Watch the upcoming financial disclosures for gross margin expansion and capacity utilization. The yield numbers, if ever disclosed, would be the most informative data point.
Logic prevails, but bias hides in the edge cases. The bias here is the belief that a small UK/Swedish foundry cannot out-execute the giants. The edge case is that specialty processes are not easily replicated, and the giants have other priorities. If Sivers can stay alive financially until the CPO wave crests, its shareholders will be rewarded disproportionately. The alternative is that a bigger player acquires it for the InP integration expertise, providing a floor for the valuation.
In a world where AI clusters consume and generate petabytes of data, the light engine is the nucleus of the network. Sivers is one of the few companies that can make that nucleus. The market should not ignore the physics because the mechanics of the balance sheet are messy. Sometimes the right investment is the one that the market refuses to love until it has no choice.
I will be watching the capacity announcements and the customer qualification lists. Those will tell me far more than any valuation multiple. The race to supply the InP light engines for the AI era is just beginning. Sivers Photonics is a dark horse, but a dark horse with a strong pedigree. Let the tape confirm or deny the thesis. Until then, the asymmetry favors the patient observer.