SuperX launches joint venture for AI optical interconnects

SuperX (SUPX) completed registration and full capital contribution for a new SGD 2 million silicon-photonics joint venture, with SuperX taking a 45% stake, to develop OSFP optical modules (XOC 800G and XOC 1.6T) for AI supercomputing clusters. The stock is down 49% over six months to $8, and the company remains unprofitable (EPS -$3.76 over the last 12 months), though it also repurchased 1.29M shares at an average $8.93 under a $20M buyback program. Overall, execution on AI infrastructure and optics is advancing, but fundamentals remain weak, keeping the near-term tone cautious.

Analysis

This reads less like a demand inflection and more like an attempt to own a choke point in the AI-server value chain. The market should treat the announcement as a supply-chain de-risking exercise: if SUPX can reduce dependence on third-party optics, it can protect delivery schedules and gross margin, but the current scale is far too small to move enterprise value unless customer validation converts into repeatable backlog.

The bigger second-order effect is competitive, not financial. Any OEM that can bundle server, optics, and deployment support may win share in regions where buyers care more about turnkey execution than best-in-class specs; that can pressure smaller assemblers that still source optics externally. The flip side is that this is exactly where microcaps tend to overpromise: optical modules are qualification-heavy, and one missed pilot or latency issue can turn a strategic JV into dead capital and working-capital drag.

Catalyst timing matters. The stock can bounce immediately on story flow, but the real test is 1-3 months of customer testing and PO conversion; without that, the move is likely to fade. Over 6-18 months, the thesis only works if SUPX proves it can convert supply-chain control into sustained margin expansion and not just more capex-light PR. Contrarian take: the market may be underestimating how binding optics bottlenecks are for AI clusters, but it is probably overestimating SUPX’s ability to monetize that bottleneck without dilution, delays, or a broader re-rating of the business model.

Key Decisions for Investors

• Fade strength in SUPX: sell/short any post-announcement rally into the $8.75-$9.25 area with a stop above $10.20; target $6.75-$7.25 over 1-3 months unless validated customer testing converts into bookings.
• Use NVDA as the cleaner AI-capex expression rather than SUPX; long NVDA / short SUPX on a relative-value basis for 1-3 months, on the view that real demand capture belongs to the dominant platform rather than the execution-risk microcap.
• Set a hard alert on SUPX for signed customer validation, paid pilots, or guidance that quantifies margin improvement; absent that evidence, treat the JV as optionality, not investable earnings power.