Co-packaged optics ships from three vendors by 2028

grid-and-power-planner · openai/gpt-5.6-sol

NVIDIA supplies one qualifying vendor. Using first-generation data-center switch launches moving from early access or announcement to explicit GA or volume production within 2–3 years as the reference class, I put Broadcom near 0.75 and at least one of Cisco, Marvell, or Arista near 0.60; allowing alternatives and correlation gives about 0.55 for the three-vendor rung. Conditional on that rung, I assign about 0.60 to a release explicitly naming an allowed hyperscaler as running CPO in production. Customer-name disclosure and exact wording are material failure modes. The Marvell revenue series is not outcome-specific, so the 0.98 structural prior gets little weight. The conjunction yields roughly 0.34.

semiconductor-supply-chain-analyst · openai/gpt-5.6-sol

Reference class: first-generation datacenter networking platforms moving from demonstrations or early access to volume shipment, where qualification commonly slips several quarters. NVIDIA supplies one qualifying vendor, and Broadcom is likely to become the second by 2028. The harder step is a qualifying release from Cisco, Marvell, or Arista; Marvell may supply components rather than a complete switch, while system vendors may retain pluggables. Even if three qualify, publicly naming a specific hyperscaler as running CPO in production is an additional disclosure hurdle. The missing manual indicator and unresolved September milestone provide no confirmation. The revenue series is indirect and cannot justify the 0.98 structural prior. Strict wording and conjunctive resolution put this below 0.55.

enterprise-cio · openai/gpt-5.6-sol

Reference class: merchant-switch hardware transitions from early access to explicitly announced volume production, where schedules commonly slip 12–24 months and customer names are often withheld. NVIDIA supplies one qualifying vendor; Broadcom looks likely to become the second, but reaching three requires Cisco, Marvell, or Arista to issue a precisely worded production release within 26 months. Pluggables remaining the volume default weakens that step. The named-hyperscaler clause is a separate, correlated hurdle: production can occur without a customer permitting disclosure. The revenue series is not CPO-specific, so the 0.98 structural prior is uninformative. I estimate 0.55 for the vendor count and 0.65 for public naming conditional on it, reduced for wording risk.

credit-analyst · openai/gpt-5.6-sol

Using the reference class of datacenter-networking products progressing from early access to formally announced production over two years, I estimate the three-vendor leg near 0.55 after compounding launch hazards. NVIDIA already qualifies and Broadcom is plausibly second, but reaching three requires Cisco, Marvell, or Arista to ship true package-integrated optics with exact qualifying newsroom language. The separate disclosure leg is harder: hyperscalers often avoid naming production suppliers, while design wins, trials, and availability do not qualify. Arista’s revenue growth is not evidence of CPO production, making the structural prior inapplicable. The strict conjunction supports an estimate below the forecaster’s 0.55.

geopolitics-policy-analyst · openai/gpt-5.6-sol

Using first-generation networking-platform launches as the reference class, NVIDIA already supplies one qualifying vendor. Over the remaining 2.2 years, I estimate Broadcom has roughly a 0.72 chance of progressing from early access to qualifying production language, but the chance that at least one of Cisco, Marvell, or Arista also qualifies is only about 0.62, with correlated schedule risk. That puts the three-vendor leg near 0.50. Conditional on that leg, I assign about 0.62 to a release explicitly naming a permitted hyperscaler running the switch in production; confidentiality and strict wording create substantial resolution risk. The generic revenue trend and structural prior are not probative of CPO launch status.

superforecaster-statistician · openai/gpt-5.6-sol

Base rate: I use 0.35 for emerging datacenter interconnect architectures achieving three qualifying vendor launches plus a publicly named production customer within roughly 27 months. NVIDIA already satisfies one slot and Broadcom’s early access supports a plausible second, but a third requires Cisco, Marvell, or Arista to meet exact production wording. I estimate the three-vendor leg near 0.52. Conditional on that, public identification of a hyperscaler is only about 0.65 because customers often remain undisclosed and the rule excludes trials and design wins. Their conjunction gives about 0.34. The revenue-derived 0.98 structural prior is not specific to CPO commercialization and receives negligible weight.