Summary of "20% BURN is COMING: ICP Cloud Engines Demo!"

Overview / Purpose of the Demo

The video is an ICP Cloud Engines demo led by Dominic Williams (DFINITY).
It focuses on the value proposition of “cloud engines” as a productized path into the decentralized cloud market.
The speaker(s) argue ICP is positioned to capture significant market share in a multi-trillion-dollar cloud revenue narrative, targeting a $1T/year cloud market.
A separate mention is made of an “intelligence gateway” for a future/other demo.

Designed for enterprises and governments.
Hosted apps/services are presented as immune to infrastructure hacks because they run on infrastructure governed by ICP’s network technology (described as “advanced network math”).

Claim: “AI will build nearly all our online functionality.”
Cloud engines can include AI nodes to run LLM inference on open-weight models on hardware the customer provisions.
If AI nodes are not added, inference can route through an “Internet Intelligence Gateway” (described as coming soon), which routes/funnels inference elsewhere.

A cloud engine is created by combining nodes.
All nodes must share the same node class (e.g., standard vs. high performance).
The wizard warns about fault tolerance if nodes are selected improperly.
- Example mentioned: choosing two nodes from the same provider reduced resilience.

Demo shows:
- Minimum nodes: 4
- Suggested “good” counts: 7, 10, 13
- Scaling example: up to 100 nodes
Claimed effects of more nodes:
- Higher security
- Higher resilience
- Higher query workload throughput
Claimed scaling approach:
- Increase node class to increase throughput.
- At the highest node class, split the engine into two to double update-workload throughput.

Pricing is framed as mass-market, quoted in dollars.
Backend flow described:
- Payment processor takes money → converts to ICP tokens → converts to cycles → cycles are used to charge/top up the cloud engine.
“Frozen engine” behavior is configurable:
- Demo mentions setting emergency reserve to none to minimize cost, with a warning about the risk of having no emergency reserve.

Engines spin up are shown on testnet, with the claim of same behavior on the live network.
A pre-created engine is used to prevent the demo from pausing.

Commentary argues about ICP burn:
- Institutions/governments can pay in fiat (dollars) rather than dealing with crypto directly.
- Flow described as: pay in dollars → ICP conversion → ICP is burned (presented as a desirable “burn” mechanism).
Incentives for node providers:
- Node providers are expected to have revenue incentives to run nodes.
- The video claims 20% of revenue is burned in a fully deflationary model.

Apps can be installed into the cloud engine.
Apps are packaged as .Ipn / inp-like bundles (subtitles suggest “inI files” / “inp files”).
The demo uploads a “private instant messaging” bundle after a trust/source check.

Highlighted app: “Sovereign Slack” (sovereign team chat).
Positioned as a sovereign messaging system where messages are stored on customer-controlled nodes.
Contrasted with AWS/centralized providers where customer control is limited.

Commentary argues centralized providers and AI tooling create risks:
- Potential extraction of sensitive information (e.g., secrets, formulas, classified-type information).
- Using AI with sensitive data can mean complete loss of control/data.
ICP is framed as enabling full data control and tailored deployments, including for military/classified scenarios.

The demo/onboarding highlights Pakistan because:
- Target governments in emerging markets can leapfrog older Western infrastructure.
- Western/EU governments are described as less likely to switch due to existing Web2 infrastructure investments.
- Pakistan is described as having a young population and widespread crypto awareness, with claims like 20M+ hold crypto and 9M+ Bitcoin.
- Geopolitical/economic pressures are presented as increasing urgency to adopt new infrastructure.
Mentions that other regions/countries may come later, including the UAE.
The demo reportedly references the “first Pakistani cloud engine” onboarding:
- 4 nodes across 4 independent Pakistani cloud providers, with a tailored infrastructural design.

How to create a cloud engine
- Choose a node class
- Select nodes (minimum 4)
- Ensure correct provider diversity for fault tolerance
- Configure top-up duration and reserve behavior
How to scale
- Increase node count
- Increase node class
- Or split the engine at the maximum node class
How to deploy software
- Use the App Center to install private bundles into the cloud engine
How to enable AI inference
- Add AI nodes for controlled inference on provisioned hardware
- Or use the upcoming intelligence gateway route

Dominic Williams (DFINITY / Internet Computer) — presenter of the cloud engines demo and value proposition.
Video commentator/host (unnamed in subtitles) — provides additional commentary and analysis, especially on burn/tokenomics, payments, and Pakistan targeting.