A $61.57m Yunnan AI tourism tender shows where bidders get stuck

AI procurement has changed flavour. Buyers once purchased servers and storage; now they purchase platforms, agents and digital scenarios — and even advice on how to price such things. A cluster of notices issued at the end of September brought the shift into the open: an AI tourism project in Chengjiang, Yunnan, with total investment of 61.5735m yuan, is tendering for whole-process cost consultancy, while a securities firm in Dongguan is procuring AI editor services. For would-be entrants, the obstacle is not the quote. It is whether they saw the notice in time, and whether they meet the hard conditions.

Treat this as a category of its own

Such projects are scattered across tourism, securities, healthcare and computing power — industries with little in common. Checking provincial platforms one by one by hand guarantees misses.

The evidence is right here. A notice published by Guangdong Zhouji Tendering Agency shows that on 30 September 2026, Dongguan Securities issued an open tender for AI editor services. A day earlier, on 29 September, the Chengjiang AI tourism project issued its tender notice for whole-process cost consultancy. One sits in finance in the Pearl River Delta, the other in tourism in central Yunnan — one day and more than 1,000km apart.

On the same day, public resource trading platforms in various regions were themselves being partly taken over by AI. The homepage of the Guangdong tendering and bidding supervision website already carries an entry point for a "tendering information AI assistant". Buyers are using AI to process information while bidders still flip through pages by hand. That gap translates directly into missed tenders.

The same move is not confined to China. According to Xinhua, the Trump administration launched America.gov, an AI-driven federal government information website, on 29 September. AI build-out on the public-sector side is advancing in parallel.

Read what is actually being bought, and where the packages split

The Chengjiang notice deserves a close reading.

Total project investment is 61.5735m yuan, of which engineering costs are 51.803m yuan. Main construction content includes an AI platform, software systems, agents, VR digital scenarios, software and hardware integration, and tourism scenario development. This particular tender covers only whole-process cost consultancy services, with a control price of 650,000 yuan, running from review of the preliminary design estimate through to review of the final settlement.

On receiving a notice, first break down the "main construction content" column and check each item against your own capability list.

One line in the notice stands alone — pricing consultancy for new-type content such as data assets, AI large models, VR large spaces and game IP experiences, with software costing using the function point method, the workload method and the market analogy method. Even how to price AI content has become a business in itself. That demand barely existed two years ago.

Hard thresholds go beyond certificates, and two are easy to trip on

Chengjiang's bidder qualifications are set out clearly:

• Consortium bids are not accepted;
• At least one project in informatisation, smart tourism, smart cities or digital platforms within the past five years (from 1 January 2021 to date);
• At least one person on site, who must be a formal employee with social security contributions for the past three months and hold at least a second-class cost engineer qualification or an intermediate professional title or above;
• On credit, those listed as dishonest judgment debtors, major tax violation subjects or serious illegal and dishonest acts in government procurement records are directly excluded, based on checks of Credit China and the National Enterprise Credit Information Publicity System.

Read the timing on that last point carefully: the check period runs from the date the tender notice is issued to the deadline for submitting bid documents. Credit checks should therefore be a fixed routine — run one again after the notice appears and keep it on file. Do not submit a screenshot from three months ago.

The window is narrower than for conventional engineering tenders

Chengjiang's tender documents are available from 29 September 2026 to 12 October, excluding statutory public holidays and rest days, and only in two daily slots: 9:00–11:30am and 2:00–5:00pm.

The real deadline is not the bid submission date but the few days in which the documents can be bought. On seeing a notice, first calculate the document acquisition deadline, then work backwards to schedule credit checks and preparation of performance materials. With holidays in between, the working days actually available are fewer than the calendar suggests.

Budgets are shifting towards computing power and foundations

According to the tendering section of Smart Healthcare Network, Guangzhou Pharmaceutical Baiyunshan's AI computing power construction project is worth 120m yuan; the Greater Bay Area Bioinformatics Centre's information infrastructure budget is 440m yuan; and the Xiamen health and medical big data centre's AI medical large model foundation and application project has a budget of 10.8m yuan.

Add "computing power", "large model foundation", "data platform", "agent" and "editor" to your keyword list, and loosen the industry terms instead. Looking only at your own sector means missing more than half of what is out there.

How to act

Missed tenders are hard to eliminate by human monitoring alone. SHEYU LEIDA (招标雷达) aggregates government procurement and tendering information nationwide, with real-time subscription alerts by keyword, region and industry; alongside its own iOS app, it offers a white-label mini-program channel version. In the context of this article, that means handing the keyword list above to it to scan the platforms for you, so you only deal with the few notices pushed to your desk.

The window will not stay open indefinitely. The first to arrive is not the lowest bidder, but the one who saw the notice first.