The data is unambiguous. Supply is constrained. ASPs are rising. The end market grows at 50% annually. Yet Sivers Photonics (LON: SIVE) trades at a valuation that ignores its order book's economic value. This is not a technology problem. It is a market structure problem. The activist investor Serenity has identified the issue correctly. Its proposed solution, however, misses the deeper operational reality.
Sivers Photonics is a specialty foundry. Not a CMOS logic house. Not a memory manufacturer. The company fabricates III-V compound semiconductor devices: indium phosphide (InP) lasers, optical amplifiers, and silicon photonics (SiPh) integrated circuits. Its customers build the optical engines that move data inside AI clusters. Ayar Labs. O-Net. Every CPO manufacturer scaling toward 2028.
The technology stack matters. Silicon photonics handles passive routing: waveguides, modulators, photodetectors. But silicon does not lase. You need InP for the active gain medium. That is Sivers' lane. Hybrid integration, bonding InP gain elements onto silicon photonic dies, is the critical path for co-packaged optics. Sivers is one of the few foundries that does this at scale.
The company operates from two fabrication sites. One in Sweden. One in the United States, based on Serenity's push to pivot toward American customers. The dual-site structure matters for geopolitical reasons. Sivers is a UK-headquartered company with Swedish fabrication and potential US capacity. It sits in a neutral position between US and Chinese supply chains. That neutrality has strategic value.
The activist thesis is straightforward. Sivers' valuation is constrained by its Swedish market listing and investor base. Swedish retail investors cannot price CPO growth stories. American institutional investors can. Therefore, a US pivot unlocks valuation. The thesis has merit. It is also incomplete.
The Technical Position
Sivers' InP epitaxial growth and active device integration sits in the industry's first tier. This is not marketing language. The company's hybrid integration capability, placing InP gain media on silicon photonic platforms, is the exact process CPO requires. Its silicon photonics platform lags TSMC's COUPE (Compact Universal Photonic Engine) by roughly two to three years. But that gap is less relevant than it appears. TSMC's COUPE targets 2025 mass production. Sivers does not compete there. It supplies the InP layer that makes silicon photonics functional. Different layer. Different moat.
Yield data is not publicly disclosed. Industry benchmarks suggest silicon photonics foundry yields run 85-95% for mature platforms. InP active integration runs lower, 70-85%. Sivers' yield performance is unverified. This is a critical unknown. Based on my audit experience across semiconductor and blockchain infrastructure projects, yield data is the first metric operators obscure and the last one investors verify. Yield directly determines unit economics. In a supply-constrained market with rising ASPs, yield improvement is the single highest-leverage operational metric. If Sivers pushes InP integration yields toward the upper end of the range, the economic value of its existing capacity increases substantially.
The packaging dimension matters equally. CPO is the most advanced packaging direction in optical communications. The core barrier is co-packaging yield, thermal management, and test methodology for optical engines combined with electrical switch chips. Sivers' contribution sits upstream: InP laser and optical amplifier performance consistency, plus co-design capability with CPO packaging processes. This is not commodity manufacturing. It is precision engineering with a multi-year learning curve.
Supply Chain Dependency
The dependency structure is clear. InP substrates come from Japan. Sumitomo Electric and JX Nippon Mining dominate. MOCVD epitaxial equipment comes from Aixtron and Veeco. DUV lithography from ASML. None of this is export-controlled in the way advanced logic is. No EUV. No sub-3nm. The supply chain is constrained but not weaponized.
The real risk is customer concentration. Sivers' top five customers likely represent 70-80% of revenue. Ayar Labs and O-Net anchor the book. The company is adding six new pluggable module customers, which diversifies the base. But concentration remains the structural vulnerability. One customer loss would materially impact revenue. The six new customers are the right strategic move, but they need to convert from design wins to production volume.
The O-Net relationship deserves scrutiny. O-Net is a Chinese optical module manufacturer. This creates a geopolitical tension. Sivers serves both American CPO innovators and Chinese module makers. In a technology decoupling scenario, this dual exposure becomes a liability. Serenity's push toward the US market is partly a hedge against this risk.
Supply chain security assessment: medium-high vulnerability. Core equipment and materials depend on imports. But optical communications chip manufacturing avoids advanced process export controls. The disruption risk is lower than for logic foundries. The InP substrate supply is concentrated in Japan, but InP is not a military-sensitive material. Actual export control probability is low.
Capacity and Capital Expenditure
Sivers' capacity utilization is not disclosed. But the supply bottleneck narrative implies utilization above 90%. High utilization plus rising ASPs equals volume-price growth. This is the core investment logic. The constraint: capacity limits revenue growth. The company cannot sell what it cannot produce.
The two-fab structure suggests a strategic allocation problem. Which customers get priority? High-value CPO customers or pluggable module customers? This allocation decision determines revenue mix and margin trajectory. If Sivers prioritizes CPO and ELS products, gross margins improve. If it prioritizes pluggable modules, volume grows but margins compress.
Any US fab expansion requires capital. Sivers' free cash flow is negative. The company will need external funding, likely equity issuance, which dilutes existing shareholders. The depreciation from new fab capacity will suppress gross margins by 3-5 percentage points until utilization reaches 60-70%. This is the financial cost of the US pivot. Semiconductor equipment depreciates over 5-7 years on a straight-line basis. The new fab needs to reach 60-70% utilization just to cover depreciation costs.
Equipment delivery timelines favor Sivers. Optical communications manufacturing equipment, DUV lithography and MOCVD systems, has delivery cycles of 6-12 months. This is shorter than advanced process equipment at 12-18 months. If Sivers commits to a US fab, equipment can arrive within a year. Production ramp takes another 12-18 months. Realistic revenue contribution: 2025-2026.
Market Demand
The demand picture is unambiguous. AI clusters need optical interconnects. NVIDIA's GB200 NVL72 racks require thousands of dollars in optical modules per unit. CPO penetration moves from under 5% in 2024 to 20-30% by 2028. LightCounting projects a multi-billion dollar CPO market by 2028. Sivers sits upstream of all of it.
The company's ELS (external laser source) products, developed with O-Net, directly serve CPO architectures. Ayar Labs' expansion plans through 2028 signal that CPO is moving from pilot to scale deployment. Supply bottlenecks persist through 2025-2026. ASPs are rising. This is a textbook supply-demand imbalance.
The inventory cycle supports the thesis. The optical communications industry runs 2-3 year inventory cycles. 2019 saw destocking. 2021 saw restocking. 2023 saw destocking. 2024 entered restocking. We are in the early phase of a restocking cycle driven by AI demand. The cycle typically lasts 2-3 years. This positions Sivers favorably through 2026-2027.
Revenue mix estimates: pluggable optical modules for data centers represent 50-60% of revenue, growing 15-20% annually. CPO represents 10-15%, growing over 50%. ELS represents 10-15%, growing over 30%. Other applications, sensing and medical, represent 10-20%. The growth is concentrated in CPO and ELS, which are exactly the segments where Sivers has technical differentiation.
Geopolitics and Export Controls
Sivers is not on any US entity list. Its technology, optical communications chips, falls outside current export control scope. No EUV. No GAA. No sub-3nm. The company's DUV lithography needs are met through standard commercial channels.
The risk is indirect. If the US expands AI-related export controls to include CPO infrastructure, Sivers becomes a compliance checkpoint. As a UK company with Swedish fabrication, it would likely pass review. But the compliance burden adds friction.
China's export controls on gallium and germanium do not directly affect Sivers. The company's core material is InP, not GaAs or Ge. However, China's National Fund III, with 344 billion yuan, targets optical communications chips. This will spawn domestic competitors. Yuanjie Technology. Accelink. These competitors will target the Chinese market first, then expand globally. Sivers faces increasing competition from Chinese specialty foundries over the next 3-5 years.
The localization trend favors Sivers' dual-site strategy. The US CHIPS Act provides 52 billion dollars for domestic semiconductor manufacturing. The European Chips Act provides 43 billion euros. Japan's semiconductor revival plan strengthens material and equipment advantages. Sivers' Swedish base benefits from European support. A US fab would benefit from CHIPS Act funding. The dual-site structure aligns with localization incentives.
Competitive Landscape
The threat matrix is real. TSMC's COUPE platform enters mass production in 2025. Intel and Broadcom have in-house silicon photonics teams with billions in R&D. IQE dominates III-V epitaxy with 30%+ share. Sivers' defense is specialization.
InP active integration is not TSMC's core competency. The company's know-how in epitaxial growth and hybrid integration creates a barrier that scale alone does not breach. But the window is narrow. If TSMC's COUPE ramps successfully and captures the CPO optical engine market, Sivers' addressable space compresses.
The competitive positioning breaks down by segment. In III-V foundry, Sivers ranks third or fourth behind IQE and Win Semiconductors. In silicon photonics foundry, Sivers ranks fifth or lower behind TSMC and GlobalFoundries. But in InP active integration, the specific capability CPO requires, Sivers ranks in the top three alongside Intel and Broadcom. This is the segment that matters.
R&D spending tells the story. Sivers runs 15-20% of revenue into R&D. The absolute amount is in the millions of dollars. Intel and Broadcom spend billions. Sivers cannot outspend them. It must out-specialize them. The InP epitaxial growth know-how, accumulated over years of process development, is not replicable through capital expenditure alone. It requires time, iteration, and customer feedback loops.
The five forces analysis is sobering. Industry competition is intense. Buyer bargaining power is strong given customer concentration. Supplier bargaining power is strong given InP substrate and equipment concentration. Substitution threat is moderate, with VCSELs competing in short-reach applications. New entrant threat is high, with TSMC entering CPO. Sivers survives through niche positioning. That is the core survival logic.
Financial Reality
The numbers are thin. Gross margins likely run 30-40%. Respectable for specialty foundry. Far below TSMC's 55-60%. R&D is likely fully expensed, which is conservative and improves earnings quality. But it suppresses current profits.
ROIC sits below WACC. The company is not creating value at current returns. Free cash flow is negative. The balance sheet will need capital for any US expansion. This is the fundamental tension: the technology is right, the market is right, but the balance sheet is thin.
Valuation multiples suggest the market is pricing growth expectations. Price-to-sales runs 5-8x. EV/EBITDA runs 20-30x. These are not distressed valuations. The market has already assigned a growth premium. The question is whether that premium is sufficient given the order book's economic value. Serenity argues it is not. The evidence is mixed.
The process executes, not the promise. Sivers has the technology. It has the market position. What it lacks is demonstrated financial execution. The order book is real. The bottleneck is real. But converting that into shareholder value requires capital discipline, yield improvement, and customer diversification. None of these are guaranteed.
The Contrarian View
Here is where the consensus narrative breaks down. The activist argument is that Sivers should pivot from Sweden to the US market. The logic: Swedish retail investors cannot price CPO growth stories. American institutional investors can. Therefore, a US listing or US operational focus unlocks valuation.
This is partially correct and fundamentally incomplete. Moving headquarters does not change the balance sheet. A US listing does not fix negative free cash flow. The valuation discount is not purely a function of listing venue. It is a function of demonstrated execution. Sivers needs to show revenue growth from its six new pluggable customers. It needs to show InP integration yield improvements. It needs to show that the Ayar Labs relationship converts into production volume. Market structure matters. It is secondary to operational proof.
The second contrarian angle: the TSMC threat is overstated. COUPE is real. But TSMC's model is high-volume, standardized manufacturing. CPO optical engines require customization, tight customer co-design, and InP integration expertise. These are not TSMC's strengths. The company's COUPE platform will serve hyperscalers with standardized needs. Sivers' specialty foundry model serves the long tail of CPO innovators. Different customers. Different requirements. The competitive overlap is narrower than the market assumes.
The third contrarian angle: the US pivot narrative ignores the value of Sivers' neutral positioning. In a decoupling scenario, a UK-headquartered, Swedish-fabricated, InP-specialized foundry can serve both US and Chinese customers. This neutrality is an asset. Pivoting entirely to the US market forfeits that optionality. The optimal strategy is not a pivot. It is a hedge.
What to Watch
The next twelve months determine the thesis. Watch three signals. First: TSMC COUPE production status. If it slips, Sivers' window extends. Second: Sivers' US fab announcements. If capital is committed, dilution follows. Third: the conversion of six new pluggable customers into recurring revenue. If this happens, the customer concentration risk recedes.
Audit first, invest later. The wafer does not lie. Right now, the wafer is telling us the bottleneck is real. The question is whether the company can convert that bottleneck into shareholder value before the competitive window closes. The InP substrate is immutable. That is a feature, not a flaw. The question is whether Sivers can make it a profitable one.