You are about to approve next year's AI budget, and a new question has landed on your desk: should money already committed to one AI vendor be allowed to pay for software from other vendors? Since 29 September 2026, that is no longer hypothetical. OpenAI now lets eligible enterprises redirect part of their committed spend to 32 partner products, and the decision it creates is a procurement decision, not a technology one.
This guide explains what the OpenAI Marketplace is, how it works, why Hong Kong enterprises face a different version of the question, and the five tests a COO or CFO should apply before treating committed AI spend as a flexible budget.
What is the OpenAI Marketplace?
The OpenAI Marketplace is a beta procurement programme, launched at OpenAI DevDay on 29 September 2026, that lets eligible enterprise customers apply part of an existing OpenAI spending commitment toward approved third-party products built on OpenAI models. It is a funding route, not an app store, and it launched with 32 partners.
A commitment here means a contracted amount of OpenAI spend that an enterprise agrees to over a fixed term, usually in exchange for better rates. The Marketplace lets some of that committed money flow to partners instead of sitting unused.
According to Channel Insider's 30 September 2026 report, the launch roster spans customer experience, cybersecurity, legal, creative and developer software. Named partners include:
--- Customer experience: Salesforce, ServiceNow, HubSpot, Zendesk, Sierra and Decagon
--- Cybersecurity: CrowdStrike and Palo Alto Networks
--- Legal: Harvey and Legora
--- Creative and knowledge work: Adobe, Figma, Notion and Glean
--- Infrastructure: Baseten, Datadog and Vercel
An analysis by eesel AI counted nine of the 32 partners in customer experience, the largest single category. That concentration tells you where OpenAI expects the most committed spend to move first.
How does buying through the OpenAI Marketplace work?
Buying through the OpenAI Marketplace follows four steps: OpenAI and the partner confirm eligibility, the customer signs directly with the partner, the partner issues the invoice, and OpenAI records the eligible amount against the customer's commitment. There is no self-service checkout during the beta, and only some products qualify.
The detail that matters most for finance teams is step two. You still negotiate and sign a separate contract with the partner, which means the vendor's legal review, security questionnaire and order form all still happen.
What disappears is the internal fight for a new budget line. The money was already approved when the OpenAI commitment was signed, so the purchase can bypass a fresh budget cycle.
OpenAI's published terms leave several questions open. As summarised by eesel AI from OpenAI's help centre and FAQ:
--- How much can be applied? Only "part of" a commitment, depending on programme terms and the purchase
--- Is every product eligible? No, eligibility is decided per product, not per partner
--- Is there a minimum commitment? Not stated
--- What happens at renewal? Not stated
--- Is it available to every account? No, it is rolling out gradually
For a procurement team, those blanks are the real content of the programme. Every unanswered question becomes a negotiation point.
Why does the OpenAI Marketplace matter for Hong Kong enterprises?
The OpenAI Marketplace matters for Hong Kong enterprises in an unusual way: Hong Kong is not on OpenAI's supported-regions list, so most local companies cannot hold a direct OpenAI commitment. The programme mainly reaches them through multinational group contracts, overseas subsidiaries, and as a template that other AI vendors are already copying.
According to PTS Consulting's 2026 guide, OpenAI withdrew service from Hong Kong, Macau and mainland China in July 2024, and that position had not changed by August 2026. Hong Kong organisations typically reach OpenAI models through Microsoft's Azure OpenAI Service, deployed in a nearby region such as Singapore or Japan.
That creates three distinct situations for a Hong Kong leadership team:
--- Regional headquarters of a multinational: your group may already hold an OpenAI commitment negotiated in the US, UK or Singapore, and partner purchases may be routed through it
--- Local enterprise using Azure: your committed spend sits with Microsoft, whose consumption-commitment rules apply instead
--- Local enterprise with no commitment: the programme does not apply today, but the procurement model it represents will reach you through other vendors
The third point is the strategic one. Committed-spend marketplaces are becoming the standard way large AI vendors sell, and Hong Kong buyers will meet the same mechanics in contracts from Microsoft, Google and Anthropic.
How is the OpenAI Marketplace different from AWS, Azure and Google Cloud marketplaces?
The OpenAI Marketplace differs from cloud marketplaces in two ways: the partner, not OpenAI, sends the invoice, and OpenAI has not published caps or renewal rules. Azure counts 100% of eligible purchases toward a Microsoft commitment and bills through Microsoft, while Google Cloud publishes a 25% cap on how much of a commitment can go to marketplace purchases.
Committed-spend marketplaces are not new. Cloud providers have used them for years to help enterprises consolidate software buying against existing commitments.
The published rules compare as follows, based on the vendor policies collated by eesel AI:
--- OpenAI Marketplace: "part of" a commitment counts, cap not published, partner invoices you, no self-service checkout
--- Microsoft Azure (MACC): 100% of the pre-tax amount of eligible purchases counts, Microsoft invoices you, self-service in the Azure portal
--- Google Cloud: eligible purchases count up to 25% of the minimum commitment, Google or a reseller invoices you
--- Anthropic Claude Marketplace: an existing commitment "can apply", cap not published, purchases by request
The invoicing difference has real operational cost. Under the OpenAI model your finance team manages two commercial relationships per purchase, one with the partner and one with OpenAI for reconciliation.
For a Hong Kong enterprise already on Azure, the more relevant comparison is usually Microsoft's own marketplace, which offers clearer rules today.
What are the risks of spending AI commitments on partner software?
The main risks of spending AI commitments on partner software are deeper vendor lock-in, unclear renewal costs, and the temptation to buy before proving value. Each partner funded through one model vendor's commitment makes switching model vendors more expensive later, because unwinding it touches several contracts at once.
Risk 1: Lock-in compounds. Committing more to OpenAI becomes easier to justify when unused spend can flow to apps. Once several tools depend on that commitment, moving away from OpenAI means renegotiating with every partner simultaneously.
Risk 2: Renewal shock. Because you contract directly with the partner, that contract has its own term. If your OpenAI commitment shrinks or ends, the partner subscription may renew at full price from a budget line that no longer exists.
Risk 3: Buying before proving. Committed money feels free. A six-figure platform decision can look smaller than it is when it does not require new budget approval.
Risk 4: Model concentration. Qualifying products must be built on OpenAI models. Funding them through the Marketplace quietly concentrates more of your AI estate on a single model provider, which is a governance question in its own right. We examine this in detail in our guide to AI model concentration risk.
How should a CFO or COO evaluate a committed-spend purchase?
A CFO or COO should evaluate a committed-spend purchase exactly as they would a new purchase, then add five tests: would we buy this anyway, what does renewal cost without the commitment, how much switching cost does it add, who owns the reconciliation, and has it been proven on our own data. Funding source should never select the vendor.
Test 1: The "anyway" test. Would this product win your evaluation if it were paid from fresh budget? If not, the commitment is subsidising a weaker choice.
Test 2: The renewal test. Model the second and third year at full list price with no commitment offset. If the business case fails that model, it is not a business case.
Test 3: The exit test. Estimate what it would cost to leave OpenAI in two years once this tool depends on it. Include data migration, retraining and contract unwinding.
Test 4: The ownership test. Name the person who reconciles partner invoices against the OpenAI commitment each quarter. Unowned reconciliation is how committed spend leaks.
Test 5: The proof test. Require a limited pilot on your own workflows, with a success metric agreed before the contract is signed.
These five questions fit on one slide. They give a board a clear reason to approve or reject without debating the technology.
What does this look like in practice for a Hong Kong enterprise?
In practice, a Hong Kong enterprise usually meets the OpenAI Marketplace through a group-level contract or a competing vendor's equivalent programme. The right response is the same in both cases: treat the offer as a pricing mechanism, run a normal evaluation, and model renewal without the subsidy before signing anything.
Scenario A: Regional office of a global insurer. Group procurement in London holds an OpenAI commitment and offers the Hong Kong customer service team a CX agent platform "at no extra budget". The Hong Kong COO's questions are local: does the platform handle Cantonese and Traditional Chinese reliably, does data processing meet the PDPO, and who pays when the group commitment is renegotiated?
Scenario B: Local logistics group on Azure. The company cannot contract with OpenAI directly, but its Microsoft account team proposes routing partner software through its Azure consumption commitment. The same five tests apply, with clearer published rules.
Scenario C: Professional services firm with no commitment. A vendor pitches a large annual commitment with "flexible spend across partner apps" as the sweetener. Here the risk is reversed: the flexibility is being used to justify a larger commitment than current usage supports.
In all three scenarios, the decision quality depends on whether someone in the room separates how a purchase is funded from whether it should be made.
What mistakes do enterprises make with committed AI spend?
The most common mistakes with committed AI spend are signing commitments larger than proven usage, letting funding convenience pick vendors, ignoring renewal terms, and failing to track consumption monthly. Each turns a discount mechanism into a cost overrun, and each is avoidable with basic procurement discipline applied early.
--- Over-committing to unlock perks: a bigger commitment for partner flexibility only pays if the organisation will genuinely consume it
--- Letting the funding route choose the tool: a weaker product funded by "free" money still costs adoption time and switching effort
--- Skipping renewal modelling: partner contracts outlive the commitment that made them easy to approve
--- No monthly burn tracking: unused commitment discovered in month eleven forces rushed, low-value purchases
--- Treating the beta as stable: OpenAI describes the programme as beta, with terms still being defined
According to McKinsey's State of AI research, most organisations using AI still struggle to show enterprise-level financial impact. Procurement discipline is one of the few levers fully within a leadership team's control.
What is the strategic takeaway for enterprise leaders?
The strategic takeaway is that AI buying is shifting from individual tool purchases to ecosystem commitments. Leaders who understand committed-spend mechanics will negotiate better terms and avoid lock-in, while those who treat committed money as free budget will find their AI architecture decided by their procurement contracts.
The OpenAI Marketplace is small today: 32 partners, beta terms and limited direct relevance for most Hong Kong companies. Its significance is the direction it signals. Every major AI vendor now wants to be the budget that other software is paid from.
The organisations that benefit will be the ones that decide their AI architecture first and let contracts follow, not the reverse. That requires a clear view of where AI genuinely creates value in your operations before any commitment is signed.
New procurement models will keep arriving, and each will promise to make AI easier to buy. Making it worth buying is still your team's job. We understand AI. We understand you. With UD by your side, AI never feels cold.
Before your next AI commitment is signed, know where AI will actually pay back in your organisation. We'll walk you through every step, from an AI readiness assessment to vendor evaluation, contract structuring and performance tracking, backed by 28 years of serving Hong Kong enterprises.
Reviewed by the UD enterprise AI team.
