Sivers Semiconductors: The Laser Bottleneck Is Real, the Earnings Are Not

Sivers Semiconductors (STO:SIVE; OTC:SIVEF) is the story the AI hardware complex keeps telling about itself, compressed into one stock. The narrative is that the bottleneck in the data-center value chain has migrated — from silicon wafers to advanced packaging to, now, the lasers and optical amplifiers that connect accelerators — and that the Swedish chipmaker owns a seat at exactly that last point. In the second quarter the market bought the narrative hard: the shares ran from about SEK 10.7 to SEK 63, a roughly sixfold move. Then the economics caught up. Sivers sold new stock to institutions at SEK 57 in early July; by early September the shares sat near SEK 31, below that placement price. The distinction between a "sell" and a "buy" here is a distinction without a difference, because both sides are betting on a conversion the financials have not yet produced.

The structural case is worth taking seriously before it is dismissed. It is not a demand story; it is a supply story, and it is real.

The constraint Sivers is chasing

The logic runs through co-packaged optics (CPO), which folds optical components directly into the chip package to cut the power that pluggable transceivers burn. Nvidia is ramping production of its Vera Rubin compute platform around this technology, citing roughly five times better power efficiency than conventional transceivers, and its Spectrum-X Photonics CPO switches are being built with Lumentum, Coherent, and Corning. The input everyone now needs is the indium phosphide (InP) laser, the light source, and the optical amplifier.

This is where Sivers positioned itself. The company struck a strategic collaboration with GlobalFoundries in which Sivers' laser arrays are integrated into GF's silicon-photonics reference designs, including the SCALE optical module used for CPO and linear pluggable optics. Being inside a reference design is not the same as being the awarded production supplier, but it is exactly the kind of qualification slot that, in a supply-constrained market, decides who captures the value. It explains why capital is flowing in: to win the bottleneck, you first have to build capacity for it.

The economics have not arrived

The financials, however, are the opposite of the story. In Q2 2026 net sales were SEK 53.8 million, down 12% year over year (10% adjusted for currency). Product revenue rose 18% at constant currency — but off a small base. Gross margin was negative: roughly −2.4% over the trailing twelve months. Adjusted EBITDA widened to a loss of SEK 35.5 million from SEK 20.9 million a year earlier.

The revenue decline is the telling part. Sivers says it deliberately cut its non-recurring engineering (NRE) business — design-and-development work — to reallocate engineers to production ramps. That is a pure capital-reallocation move: the company is giving up current, recognized revenue to build the product capacity it believes will pay later. In the terminology of this industry, it is not a cost cut; it is a revenue cut made in service of a supply bet. That is a coherent strategy. It is also a statement that the inflection is still in the future.

There is a second, quieter distortion in the quarter. The income statement absorbed a SEK 42.9 million non-cash social-security charge driven purely by the share price's rise — employee stock compensation that scales with the stock itself. The reported EBITDA of SEK −98.3 million is therefore largely an accounting echo of the run-up, not an operating outcome. It is worth flagging because it shows how much of this company's reported loss now swings on its own equity price.

Capital went where the bottleneck is

The company then funded the bet. Across the second quarter it raised roughly SEK 825 million in gross equity, including a July directed share issue of about SEK 700 million at SEK 57 per share, which was multiple times oversubscribed and added roughly 3.3% to the fully diluted count. In September it committed USD 30 million to expand its Glasgow, Scotland InP facility, moving from a fab-lite model to a hybrid one — about a third internal capacity, two-thirds from foundry and packaging partners — with a stated goal of more than 100 million continuous-wave DFB lasers of annual capacity by the fourth quarter of 2027.

Read this correctly: the capital raises and the Glasgow expansion are supply-side responses to a perceived constraint. Management says the opportunity pipeline reached USD 1.2 billion in July, up 268% from December, and cites a USD 4 billion serviceable market for optical amplifiers in AI data-center circuit switches. None of that is booked revenue. It is a pipeline, and the production orders cited — ALL.SPACE, SemiNex, a LiDAR customer, Jabil — are all scheduled for a 2027 ramp, with the quarterly revenue inflection promised for Q4 2026 at the earliest.

What the price already assumes

The problem with Sivers is not that the constraint story is implausible. It is that the market has already paid for success, on evidence that is not yet in the numbers. At about SEK 10.5 billion market value, even the friendliest published bull case concedes the stock trades near 37 times a forecast of 2026 revenue.

The established, profitable optical suppliers provide the reference point. As of this writing, Lumentum traded near 105 times EV/EBITDA and MACOM near 74 times; Broadcom, at a lower 34 times, carries the scale and the actual CPO design wins. Those incumbents are real businesses with positive margins trading at extreme multiples because the AI-optics complex is already priced for a multiyear boom. Sivers is a money-losing, negative-gross-margin, roughly billion-dollar micro-cap asking investors to pay the same kind of multiple for an inflection that still has to happen — on a course where total revenue actually declined in the latest quarter.

The condition that decides it

The honest reading is that both a long and a short are underinformed until one number appears: positive gross margin, confirmed by unit shipments rather than a pipeline. The structural supply logic is genuinely intact — that is what separates this from a pure narrative stock, and why it could run again on a single customer win. But the economics that would justify the current valuation, or a higher one, are exactly what the company has deferred to 2027. Until the Q4 2026 inflection converts the pipeline into shipped units at a positive margin, the price is paying for a bottleneck Sivers is positioned for but has not yet proven it owns. That is not a sell thesis and not a buy thesis. It is a waiting position on evidence that management itself says is a year away.