The signal is weak; the noise is deafening. Over the past twelve months, a curious divergence has emerged in the photonics sector: while the demand for co-packaged optics (CPO) has accelerated at a pace that even the most bullish forecasters failed to predict, the equity market has remained stubbornly indifferent to the very companies positioned at the heart of this supply chain. Sivers Photonics (SIVE), a London-listed III-V compound semiconductor foundry, sits at the epicenter of this paradox. The company possesses the InP active integration capabilities that CPO manufacturers desperately need, yet its valuation tells a different story—one constrained by market structure, investor composition, and a persistent failure to communicate its strategic value to the right audience.
The Foundry That Time Forgot
Sivers Photonics operates in a niche that most semiconductor analysts overlook. Unlike TSMC or GlobalFoundries, which compete on CMOS logic scaling, Sivers specializes in the esoteric world of III-V compound semiconductors—indium phosphide (InP) and gallium arsenide (GaAs)—alongside silicon photonics (SiPh) integration. This is not the world of 3nm gate-all-around transistors or EUV lithography. This is the world of waveguide dimensions, coupling efficiency, and hybrid integration—the technical underpinnings of the CPO revolution that AI data centers are now demanding.
The company's core competency lies in hybrid integration: combining InP gain media with silicon photonic passive components on a single chip. This is the fundamental building block for co-packaged optics, the technology that promises to solve the power and bandwidth density crisis facing AI clusters. When NVIDIA's GB200 NVL72 racks require optical interconnects worth tens of thousands of dollars per cabinet—versus a few thousand for traditional configurations—the companies supplying the optical engines become strategically critical.
Yet here is the paradox: Sivers trades at a valuation that suggests the market either doesn't understand what it owns or doesn't believe the CPO narrative will materialize. The company's PS ratio sits in the 5-8x range, which appears rich for a company with thin margins and negative free cash flow. But this surface-level reading misses the deeper story—one about order economics, capacity scarcity, and the structural mismatch between where the technology is heading and where the market is looking.
The Technical Reality Check
Let me be precise about the technology, because this is where the investment thesis either holds or collapses. Sivers' process technology is not directly comparable to CMOS logic nodes. The relevant metrics are waveguide dimensions, integration density, and the yield of InP active devices on silicon photonic platforms. In this domain, Sivers holds a legitimate position in the first tier for InP active integration, competing with Intel and Broadcom's internal silicon photonics teams.
The industry benchmark for silicon photonics foundry capability is TSMC's COUPE (Compact Universal Photonic Engine) platform, expected to reach volume production in 2025. GlobalFoundries' 45nm silicon photonics platform is another reference point. Sivers trails these players by approximately 2-3 years in large-scale CMOS-compatible silicon photonics integration. But in the specific domain of InP active integration—the laser sources and optical amplifiers that make photonic circuits actually work—Sivers operates at the industry frontier.
This distinction matters more than most investors realize. The CPO supply chain requires three critical capabilities: high-performance InP lasers, efficient coupling to silicon photonic waveguides, and the packaging integration that marries optical engines with switching ASICs. The first two are Sivers' core competencies. The third is where partners like Ayar Labs—the Intel-backed CPO pioneer—come in.
Based on my experience auditing semiconductor supply chains, the yield question is where the economic value gets created or destroyed. Silicon photonics foundry yields typically run 85-95% for mature platforms, but InP active integration yields drop to 70-85%. For a specialty foundry like Sivers, yield improvement is the single most important operational lever. The company doesn't disclose specific yield data, which is typical for small-cap foundries, but the "supply bottleneck" language in recent commentary suggests utilization rates above 90%—a position that confers pricing power.
The Supply Chain Trap
Here is where the analysis gets uncomfortable. Sivers' upstream dependencies are concentrated in exactly the places that geopolitical risk tends to concentrate. InP substrates are dominated by Japanese suppliers—Sumitomo Electric and JX Nippon Mining—with limited alternative sources. The MOCVD and MBE epitaxial growth equipment comes from Aixtron and Veeco, both Western companies but with long lead times. High-precision DUV lithography tools come from ASML, Canon, and Nikon.
The downstream picture is equally concentrated. The company's customer base appears to be dominated by a small number of high-value accounts—Ayar Labs in the CPO space and O-Net in the pluggable module market. The mention of "six new pluggable customers" suggests a diversification effort, but the top-five customer concentration likely remains in the 70-80% range. This is a structural vulnerability that no amount of technology leadership can fully offset.
But here's the contrarian angle that the market seems to miss: supply bottlenecks and rising ASPs mean that Sivers' capacity has scarcity value. When demand exceeds supply, the foundry—not the customer—holds the pricing power. The company's ability to allocate capacity between CPO customers and pluggable module customers becomes a strategic weapon. This is not a commodity business; it's a capacity-constrained specialty manufacturing operation in the early stages of a demand supercycle.
The Geopolitical Chessboard
The geopolitical dimension adds another layer of complexity. Sivers is a UK company with operations in Sweden and, reportedly, plans for US capacity. This positioning gives it a unique status: it can serve both US and Chinese customers without triggering the export control concerns that would apply to a Chinese or even a Taiwanese foundry.
The US CHIPS Act provides incentives for domestic photonics manufacturing, which could support a US expansion. The European Chips Act similarly benefits the Swedish base. China's National IC Fund Phase III—with its 344 billion yuan war chest—is funding domestic photonics competitors like Yuanjie Technology and Accelink, which will eventually challenge Sivers in the InP space.
The critical question is whether CPO technology gets drawn into the AI export control regime. If the US expands restrictions to cover co-packaged optics as part of AI infrastructure, Sivers—as a non-Chinese company—would likely pass compliance reviews. But the compliance burden would increase, and the company's ability to serve Chinese customers like O-Net could become constrained. This is a tail risk that the market is not pricing.
The Valuation Conundrum
Now we arrive at the heart of the matter. Serenity's critique—that Sivers' valuation is constrained by its Swedish market focus—deserves serious consideration. The Swedish market's investor base, dominated by retail investors and local institutions, lacks the framework to value a CPO supply chain play. American investors, by contrast, have watched TSMC, NVIDIA, and Broadcom reprice the entire AI infrastructure complex. They understand what a foundry with InP integration capabilities is worth when CPO penetration moves from under 5% in 2024 to 20-30% by 2028.
The valuation metrics tell a story of a company caught between two worlds. The PB ratio of 3-5x and PS of 5-8x look expensive against IQE's roughly 2x PB and 3x PS. But these comparisons are misleading. IQE is a merchant epiwafer supplier with limited CPO exposure. Sivers is a specialty foundry with direct exposure to the fastest-growing segment in optical communications. The appropriate comparable is not IQE; it's a company like Ayar Labs—if it were public—or the optical engine divisions of Intel and Broadcom.
The ROIC picture is equally nuanced. At 3-5% ROIC against a 10-12% WACC, Sivers is currently destroying value. But this is typical for a specialty foundry in the investment phase. The question is whether the CPO ramp—expected to accelerate through 2025-2028—can push ROIC above WACC. The answer depends on two variables: capacity utilization and product mix. If CPO and ELS products scale as projected, and if the company can maintain utilization above 80%, the margin expansion could be substantial.
The Institutional Blind Spot
Institutions smell blood when retail smells profit. The current situation is the inverse: retail investors in Sweden see a small-cap foundry with thin margins and negative free cash flow. Institutional investors in the US see a scarce asset in a supply-constrained market with pricing power and a direct line to the AI infrastructure buildout. The gap between these two perceptions is where the opportunity—or the trap—resides.
The "six new pluggable customers" signal is more significant than it appears. It suggests Sivers is diversifying beyond its anchor CPO customers, reducing the single-customer risk that has historically capped its valuation. Combined with the ELS product line developed with O-Net, this creates a revenue base that is less dependent on any single end-market.
But I would caution against over-optimism. The TSMC COUPE platform looms as an existential threat. When TSMC enters a market, it tends to dominate through scale, capital, and customer relationships. Sivers' defense lies in the InP active integration domain—a space where TSMC has limited expertise and where the learning curve is steep. The question is whether this moat is deep enough to withstand a sustained assault from the world's largest foundry.
The Takeaway
The market is pricing Sivers as a small-cap specialty foundry with limited growth visibility. The reality is a company positioned at the intersection of two structural trends: the AI-driven demand for optical interconnects and the transition from pluggable modules to co-packaged optics. The supply bottleneck and rising ASPs suggest that the company's capacity has scarcity value that the current valuation does not fully capture.
The path forward is clear but not without risk. A US market focus could unlock a re-rating, but it requires execution on capacity expansion, customer diversification, and yield improvement. The TSMC threat is real, but the InP integration moat provides a defensible niche. The financial constraints are significant, but the demand environment justifies the investment.
Chasing shadows in the algorithmic dark of the photonics market, Sivers represents a bet on the CPO transition—a transition that is technically inevitable but commercially uncertain. The signal is there for those who can read it; the noise is the market's failure to distinguish between a commodity foundry and a strategic asset. Volatility is the price of entry, not the exit. The question is whether the market will recognize the distinction before the opportunity closes.