AI compute procurement intelligence

Managing large-scale AI compute spend involves the same signal layers that DPX already monitors for cross-border settlement: commodity markets, supply chain disruption, sovereign and geopolitical risk, energy costs, and macro conditions. The compute.* MCP tools expose these signals framed for compute procurement decisions.

The compute desk problem

A compute desk buying $100M+ of GPUs, building data centers, or negotiating long-term chip supply agreements needs to answer:

Is now the right time to commit to a large order, or will prices fall in 90 days?
What happens to my supply chain if TSMC fab capacity drops 30%?
Which jurisdictions are safe for data center investment given current geopolitical conditions?
Is energy cheap and available in the markets where I’m building?
Are shipping lanes to key supplier regions clear?

These are intelligence questions, not payment questions — but DPX’s oracle infrastructure already monitors all of them.

MCP tools

Add to your Claude Desktop or Cursor config:

{
  "mcpServers": {
    "dpx": {
      "command": "npx",
      "args": ["-y", "@untitledfinancial/dpx-mcp"],
      "env": {
        "INTELLIGENCE_API_KEY": "your_key_here"
      }
    }
  }
}

`compute.supply_chain`

Semiconductor and AI compute supply chain risk — 30–90 day lead signals across chip fabrication, advanced packaging, memory, rare earth inputs, and logistics chokepoints.

compute.supply_chain

Returns: risk score (0–100), active alerts by segment (fab / packaging / memory / rare earth / logistics), recommended procurement actions, AI narrative.

When to use: Before placing large GPU or ASIC orders. When evaluating TSMC/Samsung/ASML concentration risk. When monitoring CoWoS/HBM packaging constraints.

`compute.energy`

Energy cost and availability for AI data center procurement — power grid stress, renewable capacity, natural gas pricing, carbon costs, and energy transition risk by jurisdiction.

compute.energy

Returns: energy stress score, jurisdiction rankings (US/EU/APAC), cost trajectory (30/60/90d), recommended hedging windows.

When to use: When evaluating data center locations. When negotiating long-term power purchase agreements. When assessing energy cost exposure for new compute capacity.

`compute.sovereign_risk`

Sovereign and geopolitical risk for compute infrastructure — jurisdiction-level assessment across Taiwan, South Korea, Japan, Netherlands, Singapore, US, and China.

compute.sovereign_risk

Returns: stability score per jurisdiction, active geopolitical alerts, cross-strait risk level, export control change probability (90d), recommended diversification actions.

When to use: When stress-testing supply chain dependencies. Before committing to single-source procurement from Taiwan-based fabs. When evaluating export control exposure.

`compute.cascade`

Simulate what happens downstream when a specific chokepoint fails. Returns cascade sequence, time-to-impact per supply chain tier, cost impact estimates, and recommended hedges.

compute.cascade shock="TSMC N3 fab offline 60 days" magnitude=80
compute.cascade shock="ASML EUV embargo expanded" magnitude=70
compute.cascade shock="Taiwan Strait closure" magnitude=90
compute.cascade shock="CoWoS packaging capacity -40%" magnitude=60
compute.cascade shock="rare earth export quota cut 30%" magnitude=65

Returns: ordered cascade sequence (which suppliers/sectors affected and when), estimated price impact, procurement hedge recommendations, alternative sourcing options.

When to use: Supply chain stress testing. Procurement contingency planning. Board-level risk briefings.

`compute.shipping`

Shipping and logistics stress for AI hardware — container availability, port congestion (Rotterdam, Singapore, LA/Long Beach, Kaohsiung), air freight rates, customs delays.

compute.shipping

Returns: shipping stress score (0–100), per-route alerts, estimated delay by corridor, recommended logistics windows.

When to use: Planning GPU cluster delivery timelines. Assessing whether spot hardware orders will arrive on schedule.

`compute.macro_timing`

Macro conditions for large capital commitments — credit conditions, FX risk for cross-border procurement (USD/TWD, USD/JPY, USD/EUR), inflation trajectory for construction and energy costs.

compute.macro_timing

Returns: macro stress score, credit spread conditions, FX regime per pair, commitment window recommendation (now / delay 30d / delay 90d).

When to use: Before signing multi-hundred-million-dollar supply agreements. Before announcing data center build programs. When assessing whether to lock in pricing now or wait.

`compute.commodity`

Commodity intelligence for compute inputs — rare earths (neodymium, dysprosium, gallium, germanium), industrial metals (copper, cobalt), and energy commodities (natural gas, silicon).

compute.commodity

Returns: spot pricing trends, supply concentration risk by country, 30/60/90d outlook, active alerts.

When to use: Assessing cost risk in long-term supply agreements. Evaluating supplier ESG exposure in rare earth sourcing.

Example agent session

In Claude Desktop with DPX MCP configured:

You: I need to assess whether now is a good time to commit to a $500M GPU
     order from TSMC and sign a 3-year supply agreement.

Claude: I'll check supply chain conditions, macro timing, and sovereign risk.

→ compute.supply_chain
→ compute.macro_timing
→ compute.sovereign_risk
→ compute.cascade shock="TSMC fab capacity -20%" magnitude=75

Based on current signals:
- Supply chain: CoWoS packaging is constrained (alert: 6-8 week lead time
  extension active). Locking in now captures current allocation before
  Q3 demand surge.
- Macro: Credit conditions are favorable for large capital commitments.
  USD/TWD is stable — no FX hedging urgency.
- Sovereign: Cross-strait risk elevated vs 6-month baseline but below
  threshold for supply chain restructuring recommendation.
- Cascade simulation: TSMC -20% capacity causes ~40% GPU price increase
  within 45 days. Securing allocation now is the hedge.

Recommendation: Proceed with commitment. The supply allocation risk
outweighs the macro case for delay.

Connecting payments to procurement

Once your agent makes a procurement decision, DPX handles the settlement — same tools, same session. Cross-border payments to TSMC (TWD), ASML (EUR), or Samsung (KRW) route through the DPX settlement rail with full compliance documentation.

compute.macro_timing          → check conditions
settlement.quote              → price the cross-border payment
compliance.screen             → screen the counterparty
settlement.execute            → move funds

This closes the loop from intelligence to execution in a single agent session.

Setup

MCP (Claude Desktop / Cursor)

{
  "mcpServers": {
    "dpx": {
      "command": "npx",
      "args": ["-y", "@untitledfinancial/dpx-mcp"],
      "env": {
        "INTELLIGENCE_API_KEY": "your_key_here"
      }
    }
  }
}

Get an intelligence API key: docs.untitledfinancial.com/beta

REST

All compute.* tools call intelligence.untitledfinancial.com directly:

curl https://intelligence.untitledfinancial.com/v1/intelligence/supply-chain \
  -H "X-API-Key: your_key"

curl https://intelligence.untitledfinancial.com/v1/intelligence/cascade \
  -X POST \
  -H "X-API-Key: your_key" \
  -H "Content-Type: application/json" \
  -d '{"shock": "TSMC N3 fab offline 60 days", "magnitude": 80}'

For AI builders — payment tools alongside these intelligence tools
Intelligence API — full endpoint reference
Commodity Forecast — 11-symbol climate-driven outlook

AI compute procurement intelligence

The compute desk problem

MCP tools

compute.supply_chain

compute.energy

compute.sovereign_risk

compute.cascade

compute.shipping

compute.macro_timing

compute.commodity