Skip to content

AI Market · Forecast desk · Live · hardware

44%

AMD publicly names a Taalas-derived product or roadmap item tied to a specific model or model class by June 30, 2027.

Probability history

0%25%50%75%100%08-0809-06deadline

Called at 46% on August 8, 2026. Repriced 1 time; now 44%. 297 days to deadline.

How the number was built

Final method: stacked-fixed-weights (resolved n=57 < 80). Each rung is an independent estimate; the final is a stacked combination.

Stated (issue date)46%
Reference class (n=29)50%
Recalibrated53%
Feature model56%
Council35%
Final44%

Drivers

  • Reference class qtype:disclosure runs 0.50 (n=29) vs stated 0.46.
  • Hierarchical calibration moves stated 0.46 to 0.53 (lens hardware, kappa_stated=24.0, kappa_lens=200.0).
  • Feature model at 0.56: conjuncts=1, named disclosure=True, horizon=326d.
  • Council of 5 at 0.35.

Cautions

  • all 5 votes came from one vendor (openai); extremization disabled, persona diversity only
  • 1 cited URL(s) failed verification and were dropped.

Council

5 personas on one model (openai/gpt-5.6-sol); other vendors were unavailable this run, so read this as one model's persona-averaged view, aggregated by log odds mean without extremization.

38%

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

The 297-day window includes likely AMD product-event opportunities, and manufacturing nodes are commonly disclosed once hardware reaches launch-level detail. However, the rule requires AMD itself to connect three elements: Taalas derivation, a named product or roadmap item, and a specific node. Acquisition updates or generic technology-integration language do not qualify. Acquired architectures often take longer than ten months to enter a public roadmap, and AMD may absorb the technology without preserving Taalas attribution. These strict documentation risks put me below the forecaster’s 0.46.

34%

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

The deadline allows only about ten months, short for AMD to convert acquired accelerator IP into a named roadmap item with a disclosed process node. A roadmap announcement is more plausible than volume productization, especially if Taalas already had working silicon. However, the resolution requires AMD itself to identify the item as Taalas-derived and specify its node; AMD could instead discuss generic integration, custom silicon, or an Instinct roadmap without publicly attributing the IP. The strict source and ambiguity rules therefore put this below the hardware and disclosure base rates.

34%

capital markets credit analyst · openai/gpt-5.6-sol

The deadline allows roughly one major AMD roadmap-event cycle. A Taalas design already near commercialization could surface, and AMD commonly identifies manufacturing nodes for sufficiently mature products. The larger obstacle is the literal attribution requirement: acquired technology is often folded into broader Instinct or custom-silicon roadmaps without naming its origin. AMD must publish both recognizable Taalas derivation and a specific node on an IR or product page; a general acquisition update, architecture discussion, or node-free product announcement misses. AMD’s elevated capex supports development activity but says little about whether this unusually specific provenance disclosure occurs by the deadline.

30%

superforecaster statistician · openai/gpt-5.6-sol

I use 0.45 as the base rate, blending the ledger’s >90-day 0.40 and disclosure 0.50 rates; hardware’s 0.615 is secondary because this rule is unusually strict. A hit needs, by a hard date, an AMD IR or product-page item that both identifies Taalas derivation and states a manufacturing node. Ordinary acquisition or integration commentary explicitly fails, while product integration plus public roadmap attribution within 297 days creates substantial timing risk. AMD’s capex growth is not specific evidence of such naming. Calibration offers no reliable support for 0.46: its 0.4–0.5 bin has only one case, while broader mid-confidence bins show overstatement.

38%

geopolitics regulation analyst · openai/gpt-5.6-sol

AMD has a natural disclosure window around its mid-2027 product and AI roadmap events, and manufacturing nodes are commonly specified for announced accelerators. However, the strict requirement is harder than a normal product disclosure: AMD must both identify a product or roadmap item as Taalas-derived and state its node on an official page. Acquired technology is often absorbed under AMD branding without lineage attribution, while integration and silicon-roadmap cycles commonly exceed ten months. A generic acquisition update, team integration, or unnamed technology reference would miss.

Evidence the engine used

direct · undated

Advanced Micro Devices, Inc. (AMD)

Cited by grid-and-power-economist: The 297-day window includes likely AMD product-event opportunities, and manufacturing nodes are commonly disclosed once hardware reaches launch-level detail. However, the rule requires AMD itself to connect three elements: Taalas derivation, a named product or road