A sourced record of potential catalysts considered for coverage. A status describes the editorial screening result, not a recommendation, price view or judgment about the company.
How to read the record
Passed means the potential catalyst cleared the screen and remains under observation. It has not been selected for a thesis.
Proposed means a sourced candidate is awaiting screening.
Under review means the decision is open through the displayed review date.
Deferred means consideration is paused; the rationale preserves evidence, replacement or coverage restrictions.
Rejected means the candidate did not clear the stated screen.
POTENTIAL CATALYSTS
Potential coverage record
8 records
5 companies
Selected radar record · NVDA
NVIDIA
3 screened developments
Under review
Development Sep 21, 2026
African AI-factory infrastructure buildout
The primary source identifies a concrete regional AI-infrastructure buildout with named NVIDIA platform exposure. Keep under review as a potential demand-diversification catalyst until deployments convert into measurable NVIDIA-attributable shipments and economics.
Four AI factories were announced or brought online across Africa within a year, with another 656 MW of capacity in the pipeline.
Cassava's NVIDIA-backed expansion spans South Africa plus planned deployments in Egypt, Kenya, Nigeria and Morocco.
Stratos Lab plans an AI cloud using more than 50 NVIDIA HGX B300 systems, providing a concrete NVIDIA system-count datapoint.
The dated primary source supplies measurable partner validation and a production grid-response demonstration. Keep under review until paid adoption and attributable economics are established.
Lambda measured 24% more cluster token throughput and 23% better performance per watt on a five-rack, 19-node HGX B200 deployment.
In a concurrent 20-node mixed workload under an 80% power policy, Lambda measured 17% higher training-cluster throughput and 20% higher inference-cluster throughput.
NVIDIA's Eos AI factory responded successfully to more than 200 Silicon Valley Power demand signals; one disclosed response reduced power from 4 MW to 3 MW without interrupting high-priority work.
DSX pricing, named paid customers, repeated external deployments and NVIDIA-attributable economics remain unestablished.
NVIDIA, Emerald AI and Google relaunched AEMA in 2026 as the next evolution of the Advanced Energy Management Alliance, which AEMA says was founded in 2014.
The alliance proposes response-speed, duration, predictability, ride-through, curtailment and operational-data standards, but disclosed no paid deployment or NVIDIA economics.
AEMA cites research estimating that moderately flexible data centers could unlock 100 GW of capacity from the existing U.S. power system.
AEMA cites 5–7+ year grid-connection timelines in key U.S. markets and an estimate of $733 million in avoided power-system costs for each GW of new flexible AI data-center capacity.
These coalition-cited estimates describe the potential grid and interconnection mechanism; they do not establish DSX adoption, NVIDIA pricing, revenue, margin or realized economics.
AEMA says it builds on existing relationships with FERC, DOE, state utility commissions and policymakers while targeting a 5–10 year interconnection backlog; this is organizational and policy capacity, not DSX adoption or NVIDIA economics.
The partnership creates a measurable potential ecosystem catalyst, but the current evidence is roadmap-stage and is not tied to an entitled published NVIDIA thesis.
Named d-Matrix and NVIDIA NVLink Fusion integration path.