Earnings call transcript: Sivers Semiconductors posts Q2 2026 revenue drop

Sivers Semiconductors reported Q2 2026 revenue of 53.8M SEK (down 12% YoY), while product revenue rose 13% YoY (18% at constant FX) as the company shifted resources from NRE engineering to production ramps. Adjusted EBITDA loss widened to -35.5M SEK from -20.9M SEK, but management attributed the step-up partly to a non-cash 42.9M SEK share-based incentive accounting charge and emphasized accelerating customer programs into production (revenue inflection expected in Q4 2026, broader product inflection in 2027). The company strengthened its balance sheet with 825M SEK gross equity capital and a $12M convertible loan converted to equity, and said its opportunity pipeline jumped to $1.2B in July (+268% from end-2025). Shares rose 2.26% to $3.35, reflecting constructive but still cautious investor reaction given ongoing profitability pressure.

Analysis

The important read-through is not the microcap’s operating noise, but that AI infrastructure spend is still strong enough to keep the optical bottleneck unresolved. That favors the companies selling into the interconnect stack—COHR, LITE, and JBL—more than the headline GPU leader itself, because the next marginal dollar of capex is increasingly constrained by lasers, packaging, and qualification capacity rather than wafer availability. If that bottleneck persists, pricing and mix in photonics can improve before unit volumes do.

Near term, the risk is that management narratives about pipelines and future ramps get treated as backlog; conversion cycles in this part of the chain can slip by quarters. The key catalyst window is 1-3 months, when customer order cadence and hyperscaler capex guides either confirm or puncture the thesis. Six to eighteen months out, the real falsifier is not macro noise but evidence that pluggables/NPO gain share while CPO remains slower and more capital intensive.

Consensus may be underestimating how much upside accrues to manufacturing and integration rather than design wins. If the market continues to price AI beta indiscriminately, that is an opportunity to own the better supply-chain beneficiaries and avoid the weakest balance sheets that still need repeated capital raises to bridge to production. For NVDA, the trend remains constructive, but the incremental edge is in the ecosystem, not the core name.

Key Decisions for Investors

• Long NVDA on 3-5% pullbacks over the next 2-4 weeks; use recent earnings strength as confirmation that AI capex is still accelerating. Falsify if hyperscaler capex commentary or next-quarter guide turns flat.
• Add COHR and LITE as a 1-3 month basket long on dips to express the optical bottleneck thesis; these names are cleaner beneficiaries of laser and interconnect scarcity than speculative design-win stories. Exit if order conversion or pricing commentary fails to improve by the next print.
• Small long JBL as a secondary beneficiary of module build-outs and packaging intensity, but only on evidence that pluggable/NPO demand is translating into actual production volume. This is a lower-conviction satellite position, not a core book item.
• Avoid initiating a direct long in the small-cap photonics story cited here; balance-sheet risk and non-binding pipeline make it more of a watch item than a high-probability trade. Reassess only if revenue inflection and gross margin stabilization show up by the Q4 2026/Q1 2027 window.