AI Infrastructure Vendor Scorecard - Lock 2026 Rates Now | Adattero

AI Infrastructure Vendor Evaluation Scorecard

Lock in Q1 2026 Rates Before Costs Spike 40-60%

📅 Last Updated: October 3, 2025

⏱️ Time: 30 minutes

💰 Potential Savings: $150K-$440K annually

⚠️ WHY THIS MATTERS RIGHT NOW

Gas turbines for new data centers are sold out through 2028. OpenAI locked up capacity equivalent to 17 nuclear power plants. AI compute costs will spike 40-60% by Q1 2026 unless you lock pricing this week.

Critical deadline: Most enterprise agreements renew November 1-December 15, 2025. Miss this window and you're stuck with spot pricing through 2026.

PART 1: 12-Point Vendor Evaluation Checklist

Score each vendor 0-2 points per question (0 = no/poor, 1 = partial/unclear, 2 = yes/strong). Maximum possible: 42 points

Infrastructure & Capacity (24 points possible)

1. Energy Strategy (0-2 pts) — CRITICAL WEIGHT: 3x

Has the vendor disclosed their power sourcing strategy for 2026-2028?

Do they have confirmed energy commitments (not just "in talks")?

Red flag: Vague answers about being "bullish on nuclear" without actual contracts (see: Sam Altman's Abilene facility tour comments)

2. Chip Supply Commitments (0-2 pts) — CRITICAL WEIGHT: 3x

Does the vendor have multi-year GPU/TPU purchase agreements with Nvidia/AMD/others?

Can they guarantee capacity allocation during shortages?

Ask directly: "What happens to my workloads if you face chip shortages in Q1 2026? Do I get throttled or cut off?"

3. Geographic Redundancy (0-2 pts)

Does the vendor operate data centers in 3+ regions?

Can they failover your workloads if one region hits capacity?

Test this: Ask which specific data centers would host your workloads. If they can't name them, that's a warning sign.

4. Capacity Guarantees (0-2 pts) — CRITICAL WEIGHT: 3x

Will they guarantee capacity in writing as part of committed pricing?

What's the SLA penalty if they can't deliver promised capacity?

Get specifics: "If I commit to 12 months at $X/month, what capacity do I get in writing? What's your penalty if you can't deliver?"

5. Infrastructure Partnerships (0-2 pts)

Who are their strategic compute partners? (OpenAI has Nvidia, Oracle, CoreWeave, Google)

Are they diversifying or single-vendor dependent?

Why it matters: Single-vendor dependency = single point of failure when capacity gets tight

6. Build vs. Rent Strategy (0-2 pts)

Does the vendor own their infrastructure or resell others'?

What percentage of capacity is owned vs. leased?

Owned infrastructure = better cost predictability (they're not exposed to upstream price shocks)

Pricing & Economics (12 points possible)

7. Pricing Transparency (0-2 pts)

Can they provide per-token AND per-compute-hour pricing?

Are overage charges clearly documented?

Red flag: "Contact sales for pricing" = negotiation leverage for you but also opacity. Demand written rate cards.

8. Committed-Use Discounts (0-2 pts)

What discount do they offer for 12-month commitments?

Industry standard: 15-25% below spot for annual commits

If less than 15%, push back hard. Reference competitive rates from AWS Reserved Instances, Azure Reserved Capacity, GCP Committed Use Discounts.

9. Flexibility Terms (0-2 pts)

What happens if you exceed your committed usage?

What's the overage rate? (Should be ≤10% above committed rate)

Avoid: Vendors who charge 2x spot rates for overages. This is predatory pricing during scarcity.

10. Early Termination (0-2 pts)

Can you exit early if capacity isn't delivered?

What are the penalties for under-usage?

Negotiate: Performance guarantees that void contracts if unmet. Don't accept one-sided lock-in.

Technical & Operational (6 points possible)

11. Model Diversity (0-2 pts)

Do they support multiple AI models (OpenAI, Anthropic, open source)?

Can you switch models without rewriting code?

Critical: Single-model lock-in (e.g., Azure only offering OpenAI) = dangerous concentration risk. Microsoft's hedge with Anthropic proves even they know this.

12. Monitoring & Controls (0-2 pts)

Do they provide real-time usage dashboards?

Can you set spending caps and budget alerts?

Must-have: Alert at 80% budget, hard cap at 110%. Sam Altman admitted ChatGPT Pro at $200/month is unprofitable due to runaway usage—you need controls.

Scoring Your Vendors (Weighted)

Critical items (Q1, Q2, Q4) count triple. Maximum score: 60 points (42 base + 18 bonus)

48-60 points: Strong vendor—lock pricing immediately

30-47 points: Acceptable with negotiated improvements on critical items

Below 30 points: High risk—find alternatives or demand major concessions

PART 2: Vendor Negotiation Email Templates

Template A: Initial Pricing Request (Azure/AWS/GCP)

Subject: Q1 2026 AI/ML Capacity Planning - Committed Pricing Request Hi [Account Manager Name], Given widely reported Q1 2026 capacity constraints (OpenAI's $500B commitments, gas turbine shortages through 2028, Nvidia's sell-out status), I'm initiating pricing discussions for our AI/ML infrastructure. Current Context: - Monthly spend: $[amount] - Primary services: [Azure OpenAI / AWS Bedrock / GCP Vertex AI] - Token consumption: ~[X million] tokens/month - Projected 2026 growth: [10-30%] Request: Please provide your best committed-use pricing for a 12-month agreement (Jan 1 - Dec 31, 2026) by Friday, October 11, 2025. I'm evaluating competitive options and need to finalize before our Q4 planning cycle closes November 15. Specific questions: 1. What capacity guarantees come with committed pricing? (SLA terms, penalties) 2. What's our flexibility if we scale above the commit? (overage pricing) 3. Can we lock token pricing specifically, or only compute hours? 4. What early-commitment incentives are available if we sign before October 31? I'm available for a call this week to discuss our roadmap in detail. Best, [Your Name] [Title] [Company] [Phone]

Template B: Competitive Leverage

Use this when you have multiple quotes. Specific competitor names by provider:

Azure: Cite AWS Bedrock, GCP Vertex AI, Oracle Cloud Infrastructure
AWS: Cite Azure OpenAI Service, GCP Vertex AI, CoreWeave
GCP: Cite Azure OpenAI Service, AWS Bedrock, Anthropic Direct

Subject: Re: Q1 2026 Pricing - Competitive Review in Progress Hi [Account Manager Name], Thank you for the initial pricing proposal. I'm currently reviewing quotes from [Competitor 1] and [Competitor 2] for our $[X]/month AI infrastructure spend. Your proposal strengths: [e.g., "Geographic redundancy across 4 regions with sub-50ms latency"] Areas where competitors are stronger: - [Competitor A] offers 22% committed-use discount vs. your 15% - [Competitor B] guarantees capacity in writing with SLA penalties of 20% monthly fee credit - [Competitor C] provides multi-model support (OpenAI + Anthropic + Cohere) Can we schedule a call this week to discuss closing these gaps? I need to make a decision by [Date] and would prefer to consolidate with [Current Vendor] if terms are competitive. Best, [Your Name]

Template C: Renewal Optimization

Subject: Early Renewal Discussion - AI/ML Contract Expiring [Date] Hi [Account Manager Name], Our current AI/ML agreement expires [Date]. Given the projected Q1 2026 capacity crunch (widely reported: Fortune, CNBC, IEEE Spectrum), I'd like to discuss early renewal at improved terms. Current agreement: - Annual spend: $[amount] - Rate: $[X] per million tokens - No capacity guarantees Proposed new terms: - 12-month extension (Jan 1 - Dec 31, 2026) - 20% committed-use discount from current rates - Written capacity guarantees with SLA penalties - 10% overage allowance at ≤5% markup over committed rate Our budget approval cycle closes November 15. Can we finalize terms by November 1 to secure capacity allocation before year-end rush? I'm open to a longer 24-month commitment if that unlocks better pricing or stronger guarantees. Best, [Your Name]

PART 3: ROI Calculator - Committed vs. Spot Pricing

How to Use This Calculator

These projections assume flat usage (no growth). If you expect 10-30% usage growth in 2026, increase your baseline by that percentage first.

Example 1: $50K/month baseline (flat usage)

Scenario	Q4 2025	Q1 2026	Q2 2026	Annual Cost
Do nothing (spot)	$50K	$70K	$80K	$800K
Lock 12-mo commit (Oct 2025)	$50K	$57.5K	$57.5K	$690K
Your savings	—	$12.5K	$22.5K	$110K

ROI on negotiation time: $110K savings ÷ 2 hours effort = $55K/hour return

Example 2: $200K/month baseline (flat usage)

Scenario	Q4 2025	Q1 2026	Q2 2026	Annual Cost
Do nothing (spot)	$200K	$280K	$320K	$3.2M
Lock 12-mo commit (Oct 2025)	$200K	$230K	$230K	$2.76M
Your savings	—	$50K	$90K	$440K

ROI on negotiation time: $440K savings ÷ 4 hours effort = $110K/hour return

Your Custom Calculation Worksheet

Step 1: Current monthly spend × 12 = $________ annual baseline

Step 2: If expecting growth, multiply Step 1 by 1.20 (for 20% growth) = $________ adjusted baseline

Step 3: Adjusted baseline × 1.50 (50% spot increase midpoint) = $________ spot pricing annual cost

Step 4: Adjusted baseline × 1.20 (20% committed increase midpoint) = $________ committed pricing annual cost

Step 5: Savings = (Step 3 - Step 4) = $________ saved in 2026

⚠️ Risk Adjustment: What If You Lock and Demand Drops?

Downside scenario: If AI adoption slows or OpenAI's bubble bursts, you're locked into committed pricing that may be above spot rates by Q2 2026.

Probability assessment: Given 700M weekly ChatGPT users, $13B annual revenue, and enterprise deals >$100M, a complete demand collapse is low probability (<20%). More likely: growth slows but doesn't reverse.

Hedge strategy: Lock 70% of projected usage with committed pricing, leave 30% flexible for spot. This gives you downside protection while capturing most savings.

PART 4: Decision Matrix - Should You Lock Pricing This Week?

Your Situation	Action	Priority
Spending >$50K/month on AI	Lock pricing NOW	🔴 Critical
Spending $10K-$50K/month	Lock pricing this month	🟡 High
Spending <$10K/month	Monitor monthly, reassess Q1	🟢 Low
AI is business-critical	Lock pricing NOW	🔴 Critical
AI is experimental/pilot	Wait and see	🟢 Low
Contract renewal Nov-Dec	Lock pricing NOW	🔴 Critical
Contract renewed recently	Low urgency	🟢 Low
Single-vendor dependency	Lock + diversify simultaneously	🔴 Critical

PART 5: Negotiation Tactics Cheat Sheet

What to Say to Get Better Terms

Tactic 1: Competitive Pressure

"I'm evaluating three vendors—[Azure/AWS/GCP]. The first to guarantee capacity in writing with competitive pricing gets the business. I need your best offer by Friday."

Tactic 2: Volume Commitment

"If you can match [Competitor]'s 22% committed-use discount, I'll commit to a 20% increase in usage vs. current baseline—guaranteed."

Tactic 3: Multi-Year Leverage

"What additional discount is available for a 24-month commit? I want to avoid renegotiating during the 2026-2027 shortage, so I'm willing to lock longer for better terms."

Tactic 4: Early Signing Incentive

"What's the best rate you can offer if I sign before October 31 vs. waiting until December? I can accelerate approval if terms justify it."

Tactic 5: Capacity Guarantee Demand

"I need written SLA guarantees that workloads won't be throttled during capacity constraints. What penalties apply if you can't deliver? Industry standard is 20% monthly fee credit."

Red Flags That Mean Walk Away

❌ "We can't predict 2026 pricing" — Translation: They'll jack rates when scarcity hits

❌ "Capacity guarantees aren't standard" — Translation: They'll prioritize larger customers over you

❌ Overage rates >2x committed pricing — Predatory pricing during scarcity

❌ No transparency on infrastructure partnerships — They don't control their supply chain

❌ "Just trust us, we'll have capacity" — Not acceptable. Demand written SLAs.

PART 6: Quick Reference - Key Deadlines

Date	Action	Why It Matters
Oct 3-11	Request pricing quotes from 3+ vendors	Start negotiations before vendor calendars fill
Oct 11-25	Review competitive options, run scorecard	Need time to compare vendors objectively
Oct 25-Nov 1	Finalize vendor selection, negotiate final terms	Budget approvals take 2-4 weeks
Nov 1-15	Secure internal approvals (CFO, CTO, procurement)	Q4 planning cycles close mid-November
Nov 15-Dec 15	Sign contracts, lock rates	Most enterprise renewals happen in this window
Jan 1, 2026	New rates take effect	Miss this and you're on spot pricing all year

Note on Non-US Enterprises

This timeline assumes US-based enterprise contract cycles. If you're outside the US, adjust based on your fiscal year and typical renewal periods. The capacity crunch is global, so urgency applies regardless of geography.