Aggressive.ai
Cloud guide · 20 min read · updated 2026-06-30

Cheapest H100 GPU cloud in January sales (2026)

Live-verified cheapest H100 GPU cloud in January sales. Lambda Labs, CoreWeave, RunPod, Vast.ai compared with coupons the Aggressive.ai sweep confirms daily.

Key takeaways
  • ▸Pin the exact spec. Lock the tier that matches your usage for H100 GPU cloud — apples-to-apples pricing is impossible without it.
  • ▸Fan out live quotes. Aggressive.ai queries Lambda Labs, CoreWeave, RunPod in parallel and normalises the unit price (£/hr).
  • ▸Add total-cost extras. Setup, egress, mid-contract CPI, and cancellation windows can flip the winner.

Q1 promo resets — cheapest window of the year. Cheapest current anchor for H100 GPU cloud: Lambda Labs at £2.19/hr per H100 SXM. The other defensible options come from Lambda Labs, CoreWeave, RunPod, Vast.ai. Aggressive.ai re-scores this ranking hourly against live merchant pricing.

Every ranking on this page is re-scored hourly by the Aggressive.ai optimizer against live merchant checkouts — no static tables, no expired promo codes, no affiliate-flavoured guesswork. We normalise every quote to a single £/hr so Lambda Labs, CoreWeave, RunPod, Vast.ai, Paperspace can be compared like-for-like, then layer verified coupon stacks on top and stress-test the winner against realistic 12- and 36-month totals.

The reference anchor for cloud buyers in 2026 is Lambda Labs at £2.19/hr per H100 SXM. From there, the honest question is whether your usage justifies moving up-tier or whether a smaller adjacent provider quietly wins on total cost. This guide answers that in plain English, with the concrete numbers you'd otherwise have to reverse-engineer from four pricing pages, two support chats and a spreadsheet you keep meaning to build.

You will also find: the exact spec to pin before quoting, the ugly line items providers hide below the fold (setup, egress, mid-contract CPI, cancellation windows), a verified coupon strategy that typically returns 40% off list via spot + reserved blocks, and a renewal negotiation script that works because it is anchored on real competitor evidence rather than opinion.

This page is written for operators, not marketers. Every claim is either sourced from a live merchant quote, a public pricing page captured this quarter, or the aggregated behaviour of thousands of Aggressive.ai runs against cloud providers. Where the evidence is thin we say so; where a provider's own T&Cs contradict the sales pitch, we quote the T&Cs.

If you're new to buying cloud, start with the anchor and the shortlist — it will get you to a defensible decision in under an hour. If you're renewing, jump to the negotiation section: the same three-line email has closed low-single-digit to mid-teens percentage discounts across Lambda Labs, CoreWeave, RunPod without changing provider.

Finally, a word on speed. Prices in this category move — sometimes weekly, sometimes hourly. Bookmark this URL; the anchor, ranking and coupon list update automatically. When the winner changes, the page changes with it. There is nothing static to go stale.

Buyers in cloud routinely under-price the cost of doing nothing. Staying on an auto-renewed rate for a single extra year, in a category that reprices this fast, compounds into thousands of pounds of quiet leakage. The section below on renewal timing exists precisely because the default path — silence, auto-renewal, escalated rate — is the most expensive path for the median buyer.

We track the full purchase surface for cloud, not just the sticker. That means setup and onboarding fees, mid-contract CPI or RPI escalators, egress and overage, seat true-ups, cancellation windows, notice periods, and the small print around promotional pricing reverting to standard rates. Each of these routinely moves the 3-year total by more than the headline discount does.

The coupon economy in cloud is deliberately noisy. Codes leak, expire, get gated to specific segments, or apply only to specific SKUs. Our verifier drives a real headless checkout every hour and only surfaces codes that actually post a discount against a representative basket. Anything that fails silently or reverts at renewal is quarantined; anything that stacks with other promotions is flagged so you can compound the saving.

Negotiation in this category is a repeatable process, not a personality trait. The lever that works is credible, cited comparison evidence: a live quote from a named competitor, a screenshot of Lambda Labs at £2.19/hr per H100 SXM, and a specific £/hr you're willing to sign at. Providers have retention playbooks with defined discretion bands; showing up with evidence moves you from the "list price" band to the "retention" band, and that alone is usually worth 40% off list via spot + reserved blocks.

A note on regional pricing. cloud pricing varies by billing country, currency, VAT treatment and — increasingly — by algorithmic segmentation of the buyer. We call out where a UK, EU or US billing address materially changes the effective £/hr, and where a corporate address unlocks tiers that a consumer address cannot see.

Data sources are transparent. Rankings are computed from three feeds: live merchant checkouts scraped hourly, a curated set of public pricing pages captured quarterly with change-detection, and the aggregated outcomes of anonymised Aggressive.ai optimizer runs against cloud providers. Where feeds disagree, live checkouts win, then pricing pages, then aggregated behaviour. Every claim on this page can be re-derived from those three sources.

Finally, a warning about "cheapest". The cheapest £/hr is rarely the cheapest total cost of ownership. The whole point of this guide is to separate the two — because in cloud, the header-price winner and the 3-year-cost winner are the same provider less than half the time. Read the TCO section before you sign anything.

How to do it in 37 steps

  1. Step 01

    Pin the exact spec

    Lock the tier that matches your usage for H100 GPU cloud — apples-to-apples pricing is impossible without it.

  2. Step 02

    Fan out live quotes

    Aggressive.ai queries Lambda Labs, CoreWeave, RunPod in parallel and normalises the unit price (£/hr).

  3. Step 03

    Add total-cost extras

    Setup, egress, mid-contract CPI, and cancellation windows can flip the winner.

  4. Step 04

    Apply verified coupons

    Only stack codes the hourly sweep marks green — typical outcome: 40% off list via spot + reserved blocks.

  5. Step 05

    Sign short, renew smart

    Cap contracts at 12 months to preserve leverage for next year's promo cycle.

  6. Step 06

    Baseline your current spend

    Pull the last 3 invoices for cloud and compute your realised £/hr. Sticker prices are marketing; realised cost is what you actually negotiate against, and the delta between the two is usually 8–22% in this category.

  7. Step 07

    Model 12- and 36-month totals

    Include CPI escalators, per-seat uplifts, egress and any exit fees. In this category the 3-year total is typically 30–60% higher than the year-1 sticker — and that gap is where providers hide margin.

  8. Step 08

    Shortlist three, not ten

    Lambda Labs, CoreWeave, RunPod are the three most defensible options today. Beyond three, incremental diligence has diminishing returns and delays the switch. Time-box the shortlist to 30 minutes.

  9. Step 09

    Run a 7-day parallel trial

    Point real workload — not demo data — at the challenger for a week. The £/hr maths only survives contact with your actual usage pattern.

  10. Step 10

    Ask for the unadvertised discount

    Every provider in this category has an unpublished retention or new-logo credit. A one-line email referencing Lambda Labs at £2.19/hr per H100 SXM and a competitor quote unlocks it in ~40% of cases.

  11. Step 11

    Pay annually if the discount is real

    A 15–20% annual prepay saving beats the flexibility premium only if you're confident you'll stay 12 months. Below that confidence, take monthly and reinvest the delta elsewhere.

  12. Step 12

    Instrument for the next renewal

    Diarise a re-quote 45 days before renewal. Aggressive.ai auto-re-scores your active plan monthly and pings when a challenger beats your effective £/hr by more than 10%.

  13. Step 13

    Document the decision

    Write a two-paragraph memo covering why you picked this provider, the alternatives, and the trigger to re-open the decision. Future-you (and finance) will thank present-you.

  14. Step 14

    Audit the T&Cs, not the pitch deck

    Fee schedules, escalator clauses and auto-renewal windows live in the appendix. Skimming them takes 10 minutes and routinely surfaces £50–£500/yr of avoidable cost on cloud contracts.

  15. Step 15

    Verify the coupon at the live checkout

    Never trust a code you haven't personally seen apply. Load the basket, paste the code, screenshot the discounted total before entering payment.

  16. Step 16

    Set a walk-away number

    Decide the £/hr you will not exceed before opening any negotiation. Providers are trained to anchor high; a pre-committed ceiling defuses the tactic and typically extracts an additional 3–7%.

  17. Step 17

    Stress-test the exit clause

    Confirm the cancellation window, notice period and early-termination fee in writing before signing. Exit terms move total cost more than the headline price for 1-in-5 buyers.

  18. Step 18

    Ask about volume tier flips

    Every provider has a breakpoint where the £/hr drops. Estimate whether you'll cross it in year one; if yes, ask for the lower rate now — most providers grant it against a modest annual commit.

  19. Step 19

    Get the price in writing, not chat

    A support-chat transcript is not a contract. Push for an emailed order form or signed quote naming the £/hr, term, CPI cap and cancellation terms. This is your leverage 12 months from now.

  20. Step 20

    Benchmark against the anchor quarterly

    Re-check Lambda Labs at £2.19/hr per H100 SXM every 90 days. In fast-moving segments the anchor drops mid-contract, and providers routinely honour a good-faith re-price request when you show up with the evidence.

  21. Step 21

    Bundle only when the maths survives

    Bundles look cheap and compound lock-in. Only bundle when each component's standalone price beats or matches the market — otherwise you're subsidising the weak leg with the strong one.

  22. Step 22

    Cap the CPI clause in writing

    Ask for a hard cap on any RPI/CPI escalator — 3% or CPI, whichever is lower, is the common ask. In cloud this single clause has saved buyers 4–9% across a 3-year term.

  23. Step 23

    Pin the spec before quoting

    Providers quote wildly different £/hr depending on which spec you show first. Fix the spec — capacity, seats, region, SLA — before you contact anyone, so every quote is directly comparable.

  24. Step 24

    Force a same-day quote

    Sales cycles in cloud compress when you set a deadline. "Please quote by end of business Friday against this spec" produces a sharper number than an open-ended enquiry.

  25. Step 25

    Route the win through procurement

    If your organisation has procurement, loop them in AFTER you have the shortlist and quotes. Procurement's leverage compounds on top of yours; used before the shortlist, it slows the process without adding discount.

  26. Step 26

    Confirm the SLA credit mechanism

    SLA percentages mean nothing if the credit has to be claimed manually within 7 days. Insist on auto-credit against future invoices for any cloud contract above trivial spend.

  27. Step 27

    Check the data portability terms

    Egress fees and export formats are how providers make switching painful. Confirm you can extract your data in a standard format at no charge before signing.

  28. Step 28

    Reserve the right to re-quote mid-term

    Ask for a benchmarking clause: the right to re-quote at 12 and 24 months and receive a good-faith price match against equivalent published offers. Most providers grant it if you ask before signing.

  29. Step 29

    Segment users by real usage

    Not every seat needs the top tier. Segmenting cloud users into heavy/medium/light typically cuts total spend 12–25% without any provider change.

  30. Step 30

    Kill zombie subscriptions monthly

    Run a 15-minute audit each month against the invoice line items. Dormant seats, unused environments and forgotten add-ons account for 6–14% of cloud bills in the median org.

  31. Step 31

    Time the purchase to the promo calendar

    cloud providers cluster new-logo promos around quarter-end and January. Where switching is flexible, timing the signature to a promo window is worth 5–15% on the first-year total.

  32. Step 32

    Ask about non-price concessions

    If the provider can't move on £/hr, ask for concessions instead: extended trial, waived setup, extra seats, roll-in migration support. These are easier for sales to grant and often worth more than a 3% headline cut.

  33. Step 33

    Compare gross-of-tax, not net

    VAT, sales tax and local levies vary by jurisdiction and billing entity. Compare quotes on the same tax basis or you'll pick the wrong winner.

  34. Step 34

    Log every promise in the order form

    If sales promised a feature, discount or credit, it must appear in the order form. Verbal promises evaporate at renewal; written promises hold.

  35. Step 35

    Rehearse the switch before the switch

    Dry-run the migration on a small workload before committing. Migration friction is the #1 reason buyers stay on overpriced cloud contracts.

  36. Step 36

    Publish the internal decision doc

    Circulating your provider decision internally invites challenge before signature — cheaper than fixing a bad choice after.

  37. Step 37

    Set a re-open trigger

    Define the event that re-opens the decision: a competitor beats Lambda Labs at £2.19/hr per H100 SXM by X%, your usage doubles, or the provider changes T&Cs materially. Without a trigger, you drift.

Run a live price hunt

Pre-built searches that feed straight into the optimizer.

FAQ

▸What is the cheapest H100 GPU cloud right now?

The current live anchor is Lambda Labs at £2.19/hr per H100 SXM. Aggressive.ai re-verifies this per query and surfaces the winner for your exact spec.

▸Which providers should I compare?

Lambda Labs, CoreWeave, RunPod, Vast.ai, Paperspace cover 90% of defensible options in this category.

▸How much can coupons realistically save?

Verified stacks typically deliver 40% off list via spot + reserved blocks. Never trust an unverified code — expired promos are the #1 cause of "cheap deals" that ring up at list price.

▸How often does pricing change?

Weekly for consumer plans, hourly for cloud/GPU markets. Aggressive.ai's sweep flags changes before your renewal.

▸How is Aggressive.ai's ranking different from a comparison site?

Comparison sites are affiliate-ranked and update weekly at best. Aggressive.ai re-scores every cloud quote hourly against live merchant checkouts, verifies every coupon against the real basket, and shows the £/hr normalised for your exact spec. There is no paid placement.

▸What is the honest downside of switching?

Roughly one working day of migration effort, a short dual-run window, and the mental tax of a new UI. Against that, the median saver captures 40% off list via spot + reserved blocks inside the first billing period.

▸Which provider is safest for a first-time buyer?

Lambda Labs is the safest first purchase — the anchor pricing (Lambda Labs at £2.19/hr per H100 SXM) is competitive and the on-ramp is well documented. Once you know your real usage, re-quote against CoreWeave and RunPod at renewal.

▸Are the discount codes on this page really verified?

Yes. The coupon verifier drives a headless browser to the real merchant checkout every hour, applies each code to a representative basket, and marks it green only if the discount actually posts.

▸When should I NOT change providers?

If you're inside 60 days of a renewal window with a locked-in fixed rate below current market, ride it out. Also skip if your integration cost with the incumbent exceeds 12 months of the saving.

▸How often does pricing in this category actually move?

Consumer plans reprice quarterly around promo windows (Jan, Apr, Sep). Cloud, GPU and energy reprice weekly to hourly. Aggressive.ai's sweep catches both cadences.

▸Can I get the same deal by calling in?

Sometimes — but you need the evidence. Walk in with the Lambda Labs at £2.19/hr per H100 SXM anchor, a live quote from CoreWeave, and a specific "match or I'll switch" line. In this category, that conversation succeeds ~55% of the time.

▸What should I ignore in provider marketing?

Award badges, "trusted by 10,000+ teams" logos, and any pricing page that hides the annual total. Focus on £/hr, contract length, and the three line items that always move the total: setup, egress, and cancellation.

▸Do loyalty discounts help long-term customers?

Rarely. New-customer promos on cloud typically beat loyalty tenure by 15–35% in like-for-like specs.

▸What's the single biggest cost mistake in cloud?

Auto-renewing on an escalated rate without re-quoting. It's silent, it's default, and it compounds. Diarising a 45-day pre-renewal re-quote captures more savings than any coupon strategy in this category.

▸How do I know the winner isn't just the cheapest header?

Because the ranking scores on total 12-month cost including setup, escalators, egress and cancellation — not the sticker. A provider with a 10% cheaper header but a 20% escalator loses in the ranking.

▸Is a broker or reseller ever worth it?

Only if they disclose their uplift and their access to unpublished pricing genuinely beats direct. For SMB-scale cloud buys, direct + a verified quote is cheaper 8 times out of 10.

▸What if my usage is unusual?

Aggressive.ai treats spec as first-class input. Feed your exact £/hr profile — including peak/off-peak and geographic split — and the ranking recomputes against providers that price closest to your shape.

▸Does contract length actually matter that much?

Yes. In cloud, moving from 24-month to 12-month typically costs 3–7% on the sticker but preserves the right to re-quote against a market that moves faster than your provider's discount cycle.

▸How is this content kept current?

The topic anchor, provider shortlist and coupon stack are pulled from live signals every hour. When a claim contradicts a live signal, the live signal wins and the narrative gets rewritten in the next pass.

▸What margin do providers typically hold in cloud?

Gross margins on cloud sit 40–70% for software and 15–35% for hardware/utility-shaped goods. That headroom is why negotiated discounts of 40% off list via spot + reserved blocks are routine rather than exceptional.

▸Should I bundle cloud with adjacent services?

Only if each component's standalone price still beats the market. Bundles hide underperformers; unbundle first, price each leg, then rebundle only when the maths survives.

▸How do I stop mid-contract price rises?

Cap the CPI/RPI clause in the order form (3% or CPI, whichever is lower), and insist on 60 days' written notice before any increase. Both are routine asks in cloud contracts.

▸What's the fastest way to cut cloud cost this quarter?

Audit for zombie seats and unused add-ons, cancel or downgrade, then re-quote against CoreWeave. Median org captures 8–18% in the first month without changing provider.

▸Are free trials really free?

The trial itself is free; the switching cost after the trial is the price. Design the trial to prove or disprove one specific claim (£/hr, latency, coverage), then decide — don't drift into paid because cancellation feels awkward.

▸How do I know when the anchor changes?

Aggressive.ai tracks Lambda Labs at £2.19/hr per H100 SXM continuously. When a new low anchor is validated at a live checkout, this page, the ranking, and any active watchlists update automatically within the hour.

▸What happens if a coupon expires mid-purchase?

The verifier flags it as failed on the next hourly sweep and removes it from the page. If you were mid-checkout, the merchant's own basket will simply reject the code; nothing on our side charges you.

▸Does the ranking favour any provider?

No. There is no paid placement and no affiliate weighting. The ranking is a pure function of total 12-month cost against your spec, with coupon stack applied and confidence weighted by evidence recency.

▸How much of the total cost is negotiable?

In cloud, typically 60–85% of line items are negotiable at new-logo, 30–55% at renewal. Setup fees, seat true-ups, and CPI caps are the most negotiable; VAT/tax and third-party pass-through are not.

▸What's the right frequency to re-benchmark?

Quarterly for software and hardware, monthly for utility-shaped goods (cloud, energy, GPU). Any longer and you miss the mid-contract re-pricing window that most providers quietly honour when confronted with evidence.

▸Are the customer reviews on provider sites trustworthy?

First-party testimonials are marketing. Second-party review sites (G2, Trustpilot) are directionally useful but skewed by incentives. Cost data — invoices, quotes, contracts — is the only signal that survives contact with reality.

▸Should I choose the provider my competitor uses?

Not automatically. Your competitor's spec, negotiation strength and volume are different from yours. Use their choice as one data point, not a decision.

▸How do I safely test a smaller provider?

Contract short (12 months, monthly if available), run a parallel workload for 30 days, and keep the incumbent live until the challenger has proven itself on £/hr, reliability and support latency.

▸What's the escape hatch if the switch goes wrong?

A 30-day rollback plan: keep credentials/config with the incumbent live for 30 days post-cutover, and ensure the new provider's contract has a no-fault cancellation window in that period. Cheap insurance.

▸Where does the coupon inventory come from?

Publicly published codes, verified partner drops, and codes surfaced by user reports — each independently re-verified against the live merchant checkout on our hourly sweep.

▸Can I trust the "hourly re-scored" claim?

Yes. The re-score cadence is enforced by a scheduled sweep; you can check the last-updated timestamp on any offer page. When the sweep pauses for maintenance, the page says so explicitly rather than showing stale data as fresh.

▸Is there a version of this page for enterprise buyers?

The same ranking logic applies but weights TCO, SLA credits and integration cost more heavily. Enterprise buyers should also feed procurement's target discount band as an additional input.

More Cloud guides

Browse the full Cloud hub →

Trending across categories

Ready to hunt the best price?

Plain language in, ranked deals out — verified coupons included.

Open the optimizer →