What we cover. AI, robotics, chips and compute, data centres, digital infrastructure, advanced manufacturing, medical technology and autonomous systems, across China, South and Southeast Asia, the Middle East, Africa, Central Asia and Latin America. Adjacent sectors — energy, EVs, supply chains, digital trade, industrial policy — when there is a technology or industry angle.
Announcements are not deployments. This is the standard that shapes most of what we publish. A memorandum of understanding is not a contract. A contract is not a delivery. A delivery is not an operating system serving real users. Every project we track in Deployment Watch carries an explicit stage label, and we do not describe a facility as operating until there is evidence that it operates.
Evidence tiers are stated, not implied. Each tracker entry names the class of evidence behind it — regulatory filings, procurement records, ministerial statements on the record, or a company press release replicated across outlets. Where a figure cannot be confirmed, we mark it Unclear rather than omitting the doubt. Thin evidence is published as thin evidence; it is not padded to match the length of better-sourced entries.
Documents before statements. Where a claim can be checked against an exchange filing, a procurement notice, a regulator’s register, a licence file in a public repository or a permit, we check it there and cite that. Company-stated figures are attributed as company-stated. A number repeated by ten outlets from one press release is one source, not ten.
Numbers. We state the basis of every figure — installed versus pipeline capacity, IT load versus total power, announced versus delivered. Where sources conflict, we say so and show the gap rather than picking the more impressive number. Currency conversions carry the date.
Sources. Named by default. We grant anonymity only where a source faces a concrete professional or personal risk, and we say in the piece why it was granted. We do not print anonymous criticism of a named person or company without giving them a chance to respond.
Translation. Chinese-language and other non-English source material is read in the original, not through machine translation of secondary coverage. Where an official English rendering differs materially from the original — as it sometimes does with figures and scope — we note the discrepancy rather than silently choosing one.
Proportionate criticism. We judge projects against what they claim at their own stage. An early pilot is not held to the standard of mature infrastructure, and a completed system is not excused for failing to be one. We are neither promotional nor reflexively sceptical; both are ways of not doing the reporting.
Use of AI tools. We use AI tools for research assistance, translation support and drafting. Every published fact is verified by a person against a primary source, every piece is edited by a person before publication, and no article is published as generated text. Errors are ours.
Corrections. Factual errors are corrected as soon as they are confirmed. Substantive corrections carry a dated note at the foot of the piece explaining what changed. Tracker entries carry an update history: earlier assessments are not silently overwritten when a project’s status changes.
Independence. SouthwardTech accepts no payment, travel, equipment, hosting or compute from the companies or governments it covers. Where a reporting trip has been funded by a third party, that is disclosed in the piece.