1. You missed the 30 January deadline
Biggest group. You want to know when you can get on, and what to do with the wait.
G-Cloud 15 (RM1557.15) was awarded on 6 August 2026 and goes live for buyers in the week commencing 7 September. G-Cloud 14 runs until 28 October, so the two frameworks overlap through the autumn. Applications closed on 30 January 2026, and the framework reopens at 18 and 36 months, which puts the next chance to join around early 2028.
G-Cloud 14 keeps running until 28 October 2026. G-Cloud 15 was awarded on 6 August and buyers get access in the week commencing 7 September, per the Government Commercial Agency. If you are on 14 but not 15, your existing call-offs continue on their own terms, but your shop window closes on 28 October. The first reopening lands around early 2028.
Three kinds of supplier land on this page. Skip to yours.
Biggest group. You want to know when you can get on, and what to do with the wait.
You need to know what happens to your existing work after 28 October 2026.
There are deadlines attached to your place, and missing them can cost it.
| Date | What happened, or happens |
|---|---|
| 17 April 2025 | Pipeline notice published. First official signal G-Cloud 15 was coming. |
| June and July 2025 | Supplier and customer engagement sessions. |
| 15 August 2025 | Preliminary market engagement notice. |
| 23 October 2025 | Tender notice and ITT published. Applications open. |
| 12 December 2025, 15:00 | Clarification questions close. |
| 16 January 2026 | Answers to clarification questions published. |
| 30 January 2026, 15:00 | Applications closed. Hard stop, no late entry. |
| 24 July 2026, 19:20 | Award notice published. Award decision dated the same day. |
| 5 August 2026 | Eight working day standstill period ends. |
| 6 August 2026 | Earliest date the framework could be signed. Brought forward from 17 September. |
| mid-August 2026 | Buyers get access. Trade press expectation, stated as an aim. |
| 28 October 2026 | G-Cloud 14 and G-Cloud 14 Lot 4 both end. |
| 27 November 2026 | Cloud Compute 2 (RM6292) ends, with a one year extension option. |
| expected early 2028 | First reopening to new suppliers, at 18 months. |
| expected mid 2029 | Second reopening, at 36 months. |
| 2030 | Framework ends. The exact date is disputed, see below. |
Worth knowing before you quote a date in a board paper. Four official artefacts give four different answers.
The award notice and GCA's agreement page both show 6 August 2026 as estimated, not confirmed execution. If you are a successful supplier reading your own award form and wondering why it says October, you are not misreading it. A contract details notice is due within 30 days of entering the framework, not within 30 days of the award notice, and should settle the actual start. Until then, treat the end date as approximately August 2030 and do not build a renewal plan on the precise day.
Crown Commercial Service became the Government Commercial Agency on 1 April 2026, so anything published before then says CCS. Same agreements, same RM numbers, new name.
The platform naming is a genuine mess, and it is worth ninety seconds of your time because people lose whole afternoons to it.
The open-framework documents provide for reopenings at 18 months and 36 months. If the estimated 6 August 2026 start is confirmed, that points to around early 2028 and mid 2029. No calendar date has been published for either.
Here is why that is much better news than it sounds.
G-Cloud 15 is an open framework under section 49 of the Procurement Act 2023. That is genuinely new, and it kills the old rule that missing a round costs you the whole framework. Rather than one supplier list locked at award for four years, an open framework is a scheme of three successive frameworks: 18 months, another 18 months, then a final 12. Between them it has to reopen so new suppliers can join.
Two things make this a real opportunity rather than a technicality.
You join on the same terms. The tender documents commit to awarding the later frameworks "on substantially the same terms" as the first, under section 49(1). Late joiners are not second class citizens on worse terms.
You will not need a G-Cloud 16 date to plan your next entry point. Under the old model, missing a round meant waiting years for the next numbered version. The published open-framework documents instead provide for G-Cloud 15 to reopen at around 18 months.
The reopening is advertised as a new tender notice on Find a Tender, and the tender documents put the responsibility squarely on you. Their words: you "must ensure that you monitor Find a Tender Service for the publication of the associated tender notice and tender documents".
There is no published notice period. The window will be measured in weeks. If you are not watching Find a Tender in early 2028, you will find out when a competitor mentions it.
Waiting is the expensive option. Three things that are not.
The official instruction is that monitoring is your job. Automate it. BidSkim does this, and so do other tools. Pick one.
G-Cloud is not the only way to sell cloud and IT to the public sector. GCA runs other technology agreements, and there are ESPO, YPO and NHS routes, plus a large amount of below-threshold spend that never touches a framework.
Every call-off award notice names the service, the buyer and the price paid. Eighteen months of reading those beats the application you would have written cold last January.
When the reopening comes, the suppliers who win work quickly are the ones who already know which buyers buy what they sell. A framework place is a licence to sell. It is not a customer.
At each reopening, existing suppliers pick one of three options on the Digital Platform.
Under all three, GCA checks the debarment list and exclusion grounds again. And a useful thing to know: you do not have to wait for a reopening to change your prices. Baseline prices can be increased or decreased by the supplier at any time during the framework period.
This is the group the transition catches out, usually because two true things sound like they contradict each other.
You cannot win new work through G-Cloud 14 after 28 October 2026. But a call-off you sign before that date runs for its own term, up to 36 months plus a 12 month extension, long after the framework itself is gone.
A G-Cloud 14 call-off signed in October 2026 can legitimately run into 2030. For a buyer who already knows you, already has your pricing and already understands the terms, that is a real reason to sign now rather than start again on a new framework.
One thing to plan around: central government departments need Cabinet Office spend controls approval for call-offs beyond 24 months. That takes time you do not want to discover you needed in late October.
So the next twelve weeks are the last window to turn G-Cloud 14 conversations into signed call-offs. After 28 October, reaching that buyer means G-Cloud 15 if you are on it, a different framework, a direct procurement, or the reopening.
Note that G-Cloud 14's Lot 4 ends on the same day, and Cloud Compute 2 follows on 27 November 2026. All three are being replaced by the single G-Cloud 15 framework.
Being named on the award notice is not the finish line. There are deadlines attached, and missing them can cost you the place.
After that, the harder part. Your listing goes live into a catalogue with thousands of entries and nobody is coming to find you.
People search this catalogue for the outcome they need in plain English, not for your product name.
On three of the five lots, price is 80% of the score. More on that below.
Award history on G-Cloud 14 is the only honest guide to what this market pays.
Buyers can now run a competition in any lot, but a framework place creates eligibility rather than demand. See the strict recorded-award analysis below.
G-Cloud 15 absorbs three agreements at once: G-Cloud 14, G-Cloud 14's separate Lot 4, and Cloud Compute 2. The old three lots plus Lot 4 come out the other side as five.
| Lot | Official name | Applied | Awarded | Awarded SMEs | Max call-off |
|---|---|---|---|---|---|
| 1a | IaaS and PaaS | 96 | 93 | 46 | 5 years + up to 3 |
| 1b | IaaS and PaaS above OFFICIAL | withheld | 16 | withheld | 5 years + up to 3 |
| 2a | Infrastructure Software (I-SaaS) | 312 | 311 | 217 | 4 years + up to 2 |
| 2b | Software as a Service (SaaS) | 2,744 | 2,744 | 2,361 | 4 years + up to 2 |
| 3 | Cloud support services | 2,282 | 2,282 | 1,953 | 4 years + up to 2 |
Cyber Essentials Plus is required for Lots 1a and 1b, Cyber Essentials for the rest. Max call-off durations are in years.
On Lots 2b and 3, every single applicant was awarded a place. On Lot 2a, 311 of 312. Only Lot 1a saw meaningful attrition, and only three suppliers. If you expected a framework place to distinguish you from rivals, the data says otherwise.
Awarded SMEs are 86% of Lot 2b and 86% of Lot 3, but about 50% of Lot 1a. Lot 1b's SME count is withheld, so no percentage is supportable there. The further up the infrastructure stack, the more the published data shifts toward large suppliers.
The award notice's structured data shows a single supplier, because all 16 awarded suppliers are withheld under section 94 and appear behind one placeholder entry. The notice text states the real figure: 16. Anyone reading the data rather than the words will publish "1". The published roster therefore contains 4,231 named supplier records plus that placeholder, not a knowable count of unique legal entities. The named records map to 4,210 organisation records.
One more from the notice: the structured release contains 4,233 tenderer party records. That is not a reliable count of unique legal entities because the section 94 redactions replace multiple parties with a placeholder. The award data publishes 4,231 named supplier records plus the Lot 1b placeholder, but it cannot reveal the exact number of unique legal entities awarded.
You can bid for as many lots as you like, and unlimited places were available on every lot. But each framework contract is lot-scoped, so you sign a separate technical acceptance certificate per lot and you cannot deliver in a lot you were not awarded.
G-Cloud 14 had one software lot. G-Cloud 15 has two, and this is where suppliers will misfile themselves. The good news is that the boundary is decided by a taxonomy, not by your interpretation. G-Cloud 15 adopts IDC's technology taxonomy, and the framework specification carries a table saying exactly which parts of it belong to which lot.
Lot 2a is everything in the Systems Infrastructure Software category: system and service management, cloud financial management, network, security, storage, endpoint management and physical or virtual computing. Plus exactly two sub-sections of the Application Development and Deployment category: Integration and orchestration, and (Deployment centric) Application platforms.
Lot 2b is everything else in Application Development and Deployment, plus the whole Applications category: collaboration, content workflow, ERP, supply chain, production and operations, engineering, and CRM.
The trap is real and specific. The taxonomy code for Application platforms appears twice, once under Lot 2a and once under Lot 2b, split on whether it is "deployment centric". If your product sits near that line, resolve it in the full taxonomy annex rather than guessing. Getting it wrong puts you in a lot with nine times more competitors, or gets your service rejected as out of scope.
One nuance that is widely missed: hosting suppliers on Lots 1a and 1b may also offer software and support services as ancillary items, but every Lot 1a or 1b call-off must include at least one core hosting service. Similarly, Lot 2a and 2b suppliers can offer cloud support as an ancillary.
The official test is sole control, not ownership. You qualify if you are "in sole control of the infrastructure that underlies its Services", or you can evidence that you are an accredited reseller of such infrastructure. Resellers are explicitly permitted, and there is a dedicated reseller branch in the tender questions: a reseller relying on its cloud provider does not have to hold ISO 14001, 27017 and 27018 itself.
Beyond that, the hosting lots carry the heaviest gates on the framework: a Gold financial viability assessment, mandatory validated ISO 9001, 20000-1 and 27001, ISO 27018 if you offer public cloud, Cyber Essentials Plus, a carbon reduction plan, and pay as you go availability with no minimum commitment.
Lot 1b adds security clearance for staff, an accredited secure facility within 6 months, security aspects letters, and compliance with UK embargoed countries policy. Its pricing is not published publicly, and its service information is treated as classified.
The specification excludes four things outright: hardware bought independently of an as-a-service purchase in the same call-off, co-location services, contingent labour and any services inside IR35, and anything beyond enabling access to the services themselves.
On G-Cloud 14, Lot 4 existed for one reason: it was the only lot where buyers could run a real competition. G-Cloud 15 allows call-off with or without competition in every lot, so a separate lot for it became pointless. Lot 4's function is absorbed into Lot 3.
Lot 3 has two useful properties. It can be bought entirely separately from the hosting and software lots, and even from a different supplier, including to support cloud services procured outside G-Cloud altogether. And non-cloud services get removed from the platform, so the lot stays what it says it is.
This is the part most guides skip, and it is the part that determines whether a listing earns anything.
Lots 1a and 1b: Quality 80%, Price 10%, Social Value 10%.
Lots 2a, 2b and 3: Quality 10%, Price 80%, Social Value 10%.
If you are selling software or support, this is close to a price auction with a quality threshold. Some trade coverage reports different weightings. The published tender documents say 80/10/10 and 10/80/10.
A quirk of Lot 3 worth knowing, straight from the tender notice: if a buyer awards without competition, the price defaults to your framework maximum day rates. The notice says so plainly, and adds that better value for money is likely from competing. Read that from both sides. A buyer who cannot be bothered to compete pays your ceiling, and a buyer who does compete will expect you to come down from it.
The specification bans three habits that are normal everywhere else. No price on application. No "from £x per day". No minimum-only prices or unexplained ranges. Public cloud pricing must be publicly visible on your own website and linked from your listing.
Framework prices are not indexed, so you cannot raise them with inflation. But baseline prices can be moved up or down by you at any time during the framework.
The management charge is 0.75% of everything you invoice to buyers, excluding VAT, payable within 30 days of an undisputed invoice. Monthly management information returns are mandatory, and there is a penalty charge if you fail to file them.
Call-off liability caps are the greater of a floor or 125% of estimated yearly charges.
| Lots | Call-off liability floor | Insurance required |
|---|---|---|
| Lots 1a, 2a, 2b and 3 | £500,000 | £1m PI, £1m public, £5m employers |
| Lot 1b | £20,000,000 | £50m PI, £20m public, £5m employers |
| Framework level, all lots | £150,000 or 125% of the management charge | n/a |
Data protection liability is £10m for most lots and £20m at framework level for Lot 1b. Termination for convenience notice was cut from 90 days to 30.
Lot 1a sits with the software lots at £500,000, not with Lot 1b at £20m. They are both hosting lots, so the assumption is easy to make and expensive to get wrong.
The general terms document still carries the standard public sector baseline of £5,000,000 or 150%. The framework award form overrides it. If you price your risk off the general terms alone, you will have both the floor and the percentage wrong.
The clarification log was extensive, and the requirements shifted more than once during the process. Lot 1b insurance came down to the figures above. ISO evidence was relaxed to accepting evidence that accreditation had been started. The one penny pricing floor was removed so sub-penny and zero tiers are allowed. And Cyber Essentials was made mandatory for Lots 2a, 2b and 3, having previously been optional.
If you are reading a guide written before spring 2026, it probably states the pre-climbdown version of at least one of these.
G-Cloud is not a side door. GCA's own tender documents put G-Cloud spend at £3.1bn in FY23/24, inside a public sector cloud market it estimates at around £6bn a year. The published ceiling for G-Cloud 15 is £14bn excluding VAT, or £16.8bn including it.
That ceiling grew as the framework was planned. The first pipeline notice in April 2025 estimated £4.8bn. By August 2025 it was £8bn. By the time the tender published in October it was £14bn, nearly three times the original figure. Note also that no lot has its own value: all five carry the same £14bn total, with the caveat that lot values may be shared between lots. Anyone quoting a per-lot value has invented it.
Now the uncomfortable half: a listing is not revenue.
Getting on is the easy bit. It makes you eligible, and that is all it does. The suppliers who make money here treat the listing as the start of the sales job rather than the end of it.
In our fixed-date analysis, the 4,231 named supplier records map to 4,210 organisation records. Across those organisations we found 61,861 recorded award participations and £79.35bn of exact, supplier-allocated GBP value. Framework-parent placements and framework ceilings are excluded; shared or otherwise unallocated values stay unknown rather than being assigned to a supplier. See the definitions, limits and full supplier analysis.
G-Cloud 15 was visible in public data from April 2025, nine months before applications closed. Most suppliers still reacted late, or not at all.
The contracts underneath it work the same way. Every G-Cloud 14 call-off signed since 2024 has an end date sitting in public data right now. Each one is a conversation somebody is going to have with that buyer, months before anything is re-tendered. Across the UK notices BidSkim tracks, the pattern holds: by the time a re-tender is published, you are near the end of the opportunity, not the start.
The best time to notice a contract is not the day it comes back to market. It is months earlier, while there is still time to talk to the buyer.
Compiled 7 August 2026 from the RM1557.15 tender pack (Framework Schedule 1 v2.1, Framework Award Form v1.2, Attachments 1, 2 and 9, and the clarification response log) and all six Find a Tender notices published under OCID ocds-h6vhtk-0504a3, cross-checked against the GCA agreement pages. Where the tender pack and the notices disagree, this page says so rather than picking one silently.
A fixed-date analysis of 4,210 matched organisations: 61,861 award participations and £79.35bn of exact supplier-allocated GBP, excluding framework placements and ceilings.
Missing a join window can lock out years of work.
The best time to notice a contract is months before it returns to market.
What the official Britain-wide source carries, and what it does not.
Set up alerts for framework and DPS routes in your market, then get on with the work.