Why us / The AI feed

A plain-text feed of your deck, built for LLMs.

The one-pager is for humans. The AI feed is the same narrative, structured as plain text, designed specifically for ChatGPT, Claude, Gemini, Copilot or an internal RAG. Two artifacts, two audiences. Both ship with every deck.

Dell Technologies · Infrastructure Solutions GroupTwo artifacts ship with every deck

Launch deck32 / 14

One-pager · for humanspoweredge-xe9680.pdf · 1 page · 412 KB

Dell Technologies

PowerEdge XE9680

Ships H1 2024

AI factory in a 6U rack.

9.4×

vs V100

× H100

$0.46

/1M tok

GPU 8 × NVIDIA H100 SXM5

Mem Up to 2 TB DDR5

Net NVLink 900 GB/s

slidedeckstudio.com/dell/poweredge© Dell Technologies 2023

Brand-matched. Printable. Forwardable. The leave-behind after a CTO meeting.

AI feed · for LLMsai-feed.txt · 1.8 KB · UTF-8

# Dell PowerEdge XE9680, AI factory in a rack

## Product

Dell PowerEdge XE9680. 6U rack server.

Purpose-built for generative AI training

and large-scale inference.

## Specifications

- GPUs: 8 × NVIDIA H100 SXM5 (80GB HBM3)

- Interconnect: NVLink 900 GB/s

- Memory: up to 2TB DDR5

- Power: 10.2 kW peak, 4 × 2800W PSU

- Form factor: 6U rack

## Performance vs prior generation

- Llama-2 70B fine-tune: 9.4× V100 baseline

- Inference TCO: $0.46 / 1M tokens

- 2.9× throughput vs A100 reference

## Availability

Ships H1 2024. Channel partners enabled.

## Sources

[1] Dell PowerEdge XE9680 spec sheet, Nov 2023

[2] NVIDIA H100 Tensor Core GPU datasheet

[3] Internal Dell ISG TCO model, v3.2

Plain text. Semantic. Cited. Drops cleanly into ChatGPT, Claude, Gemini, Copilot, Dell's internal RAG.

From the engagementSame source of truth. Two outputs. The one-pager went to 1,247 Dell channel partners as a leave-behind; the AI feed populated Dell's partner-portal RAG so every rep could ask "what's the H100 NVLink bandwidth?" and get a sourced answer.

How it works

Clean structure. Any LLM can read it.

The AI feed mirrors the deck, same narrative, same numbers, same hierarchy, but flat, semantic and machine-readable. Real headings. Real lists. Real footnoted sources. No exported-to-PDF mess of overlapping text boxes. No design chrome to confuse the model.

The recipient feeds it to ChatGPT, Claude, Gemini, Copilot or their company's internal LLM, and asks for whatever format they actually want. Bullets. Exec summary. Q&A. Translation. Slack thread. Email draft.

Your deck stops being a static file. It becomes a knowledge artifact.

← See the one-pager

The flow

From your deck to any LLM, any output.

Dell Technologies · Infrastructure Solutions GroupSlide 28: Where the AI feed travels

Launch deck28 / 14

Source · 1 file

ai-feed.txt

Shipped with every PowerEdge XE9680 deck. Cited, structured, machine-readable. 1.8 KB.

Sections

## Product
## Specifications
## Performance
## Sources [1][2][3]

Ingested by · 5 LLMs

CChatGPT
CClaude
GGemini
CCopilot
DDell RAG

Outputs · 6 channels

→ Draft partner-enablement email · 247 words
→ CFO summary · 5 bullets · TCO focus
→ Spec sheet · translated to JP, KR, DE
→ Slack thread · 3 posts · launch-day cadence
→ Q&A bot · 1,247 partners answered in 24h
→ Analyst briefing note · Gartner / IDC ready

From the engagement24 hours after the PowerEdge XE9680 announcement, Dell's partner RAG had answered 1,247 questions across 14 countries - each citing the same source file we shipped with the deck.

01 / Use

Summarise on demand

Paste the URL into ChatGPT or Claude. Ask for an exec summary, a five-bullet recap, a thread.

02 / Use

Translate without re-design

Drop it into Gemini for a Spanish or Japanese version. The structure survives the translation.

03 / Use

Q&A for internal LLMs

Copilot, Claude for Work, custom RAG can answer staff questions on the deck without anyone reading it cover to cover.

04 / Use

Forwardable, not a 40 MB attachment

A real text file. No fonts to embed, no aspect ratio to fight.

Start a project