Photon Capital
Photonics & Semiconductor Research
Keysight remains a favored test name as its planned OCP showcase extends AI infrastructure validation across scale-up, scale-out and scale-across networks and inference. The breadth suggests growing traction in testing varied AI architectures as data-center operators need to validate more complex deployments, extending its earlier silicon-photonics design-to-test opportunity. x.comÂ
Keysight is moving upstream from test instruments into silicon photonics design, which should pull downstream metrology demand into its ecosystem. It added a GlobalFoundries photonics PDK to its ADS Photonic Designer, linking chip design and system simulation so engineers can validate optical links before tapeout for transceivers and CPO, owning the design-to-test workflow. GlobalFoundries tying the work to AI infrastructure suggests its photonics platform is nearing volume production. x.comÂ
Keysight's gross margin expanding on fast revenue growth shows real chokepoint pricing power in AI test. Q3 FY26 gross margin reached about 69%, up roughly 140-150 bp from Q2's clean 67.6%, as revenue grew 36.5% to $1,846M; operating margin hit 33.2%, above its 31-32% long-term target. Orders rose 56% to $2,091M, though book-to-bill narrowed to 1.13, and Q4 EPS guidance of $3.34-3.40 tops the ~$2.68 consensus. Management says it spans pre-silicon design through manufacturing. x.com x.com x.comÂ
Keysight has an overlooked moat: before any hardware innovation such as AI chips, 5G/6G or EVs reaches the market, it must be tested and meet strict standards, and Keysight owns that critical testing market with a unique mix of hardware and software. Rather than guessing which tech company will win, investors can own the one that tests them all. The stock is up 69% YTD. x.comÂ
Aehr Test Systems is shifting from niche test provider to a required layer in scaling chip systems, and revaluation begins when the market catches up. As AI failure costs rise, testing becomes required infrastructure; Aehr reports follow-on AI processor orders and a major hyperscaler win for next-gen AI ASIC burn-in. It also tests silicon photonics as photonics replaces copper interconnects, and its wafer-level burn-in serves rising packaging complexity. x.comÂ
Keysight is the photonics test stock to buy now because the 1.6T production ramp, where the big checks get cashed, has not yet peaked and Keysight is the least exposed to a major correction. It reported $1.6B in revenue with 30% year-over-year growth projected for Q2, and is buying back $1.5B of stock because it sees the 1.6T cycle lasting through 2028. Its diversified $57B market cap makes it safer than Aehr, which has run too far too fast; Goldman has a $384 target. x.comÂ
Credo's interconnect opportunity grows as campus-scale AI clusters require more links to validate and integrate. The OCP 2026 meeting takeaway is that complex requirements put interconnect at the center, with both copper and optical links set to benefit. x.comÂ
Coherent's pivot to AI optics is paying off as copper gives way to optical links inside large AI clusters. Quarterly revenue topped $2 billion for the first time, up 42%, with datacenter and communications up 59% to 79% of sales. Multi-year contracts with AI leaders including Nvidia extend order visibility into 2028. It launched its PhotonLink interconnect platform at ECOC 2026, and its indium phosphide, VCSEL and 6-inch wafer strengths give it a moat in 800G and 1.6T transceivers. x.comÂ
All high-bandwidth data center interconnects will be optical within a few years, making photonics one of the most important AI infrastructure themes. The opportunity spans seven layers: materials and wafers such as AXT and IQE, tools such as Aixtron and Applied Materials, lasers such as Sivers and MACOM, foundries such as Tower and Fabrinet, testing such as Aehr and Teradyne, optics such as Lumentum, Coherent and POET, and networking such as Credo, Marvell and Ciena. x.comÂ
AI is accelerating quantum computing by attacking its real bottleneck, real-time error correction, a pattern-recognition task deep learning excels at. DeepMind's AlphaQubit cuts errors 30% versus the best fast classical decoder, and Nvidia is productizing the layer with Ising, NVQLink and CUDA-Q as the core toll road. Alphabet, IBM and Microsoft add lab exposure, Rigetti and IonQ pure-play upside, and FormFactor, MKS and Keysight sell shovels. AI may also eat quantum's early chemistry uses. … Show more
Shorter qualification windows make transceiver testing a critical bottleneck for photonics, favoring Viavi, Aehr Test Systems, Cohu, FormFactor and Teradyne. Celestica's OCP slide shows windows falling from 36 months for 400G to 24 for 800G and 18 for 1.6T. With GPU architectures refreshing every 12 months and 224G per lane leaving little margin for error, laser fab capacity alone cannot match the silicon cadence without faster validation. x.comÂ
Marvell looks well positioned for scale-up compute domains, where interconnect is the wall that limits performance. Scale-up domains should be thought of differently from simply networking racks together, and Marvell's own commentary now focuses on exactly that opportunity. x.com x.comÂ
