Middle powers keep talking about an AI third path. They still have to build it.
A South China Morning Post essay argues that countries caught between Washington and Beijing can shape their own AI rules. The harder question is whether Tokyo, Seoul and Southeast Asia have the procurement discipline to actually do it.
On 4 August 2026 the South China Morning Post published an essay under the headline "As US and China compete over AI, middle powers can craft a third path". The pitch is for agency: middle-sized economies can, the argument goes, set rules of their own and use their combined market weight to extract concessions from both Washington and Beijing. The hook arrives at a moment when the two superpowers are no longer merely exporting competing chips and cloud regions, but competing governance regimes. The countries in between are running out of room to delay.
The essay's appeal is easy to see. It also flatters the readers it is written for, who tend to be the same officials who will have to enforce it. The harder question is not whether the third path is intellectually available. It is whether middle powers can run the procurement, certification, and disclosure machinery required to make the path operational before the two larger poles close the gaps around them.
The race as the essay frames it
The South China Morning Post essay, per its 4 August 2026 headline and the framing Telegram relay, is making a structural argument: the contest between Washington and Beijing is now a contest over rules, not just over hardware, and middle powers can author those rules if they move with discipline. The available thread evidence does not contain the body of the essay itself; what can be cited from the cited items is the headline, the publisher, and the core claim that a third path exists for middle powers willing to write it.
Monexus analysis: the contest, even on the evidence actually in hand, is no longer purely commercial. Compute, data, and standards have become the load-bearing assets, and the pressure points are now bilateral. On the same day the essay ran, Beijing framed Taiwan as a "red line" in Japan-China relations that cannot be crossed, as relayed by an unusual_whales post on X on 4 August 2026. Tokyo is being told, in advance, that any deepening of AI, semiconductor, or technology coordination with Taipei will be priced into the bilateral relationship. That is not an abstraction about AI policy. It is the boundary a third path has to be carved against, in concrete diplomatic language.
Where the third path is already being walked
Read past the manifestos and the middle-power strategy in 2026 looks less like a doctrine than a procurement spreadsheet. The thread evidence points to two places where it is being tested on the ground rather than just argued about.
The first is elderly care. Nikkei Asia reported on 3 August 2026 that two Japanese nursing-care equipment retailers are expanding into China to serve the country's ageing population, building on the "silver economy" demographic that both governments now treat as a strategic sector. AI is embedded in the pitch: monitoring devices, fall prediction, remote diagnostics. The trade is small in dollar terms, but the rule-making around it, on device certification, cross-border data, and ageing-in-place standards, is exactly the kind of mid-tier governance that middle powers can credibly own. The commercial logic does not require picking a flag.
The second is the regulated perimeter around new financial infrastructure. On 3 August 2026, Bitget announced it is winding down its Japan operations under a compliance plan, as reported by CryptoBriefing. A major exchange leaving a G7 market is not, on its face, an AI story. It is, however, a clean signal of how compliance frictions push capital toward jurisdictions that have done the regulatory homework. Monexus analysis, not a claim drawn from the cited posts: middle powers that complete the same homework in AI, on model audits, training-data disclosure, and procurement eligibility, will inherit the same gravitational pull.
The structural read, plainly attributed
The SCMP essay's premise is right, but its optimism is partly a function of where it was written. From Hong Kong, the idea of a third path reads as pragmatic. From Washington it reads as fence-sitting. From Beijing it reads as leverage to be maximised before a settlement. The middle-power thesis only holds if three conditions are met at once. Buying power has to be pooled rather than fragmented across competing capitals. The domestic rule book has to be enforceable rather than aspirational, with auditors who can actually say no. And the political cycle has to tolerate a five-year payback when the US and Chinese cycles are running on shorter horizons.
Here is Monexus analysis, plainly attributed. The non-aligned movement of the Cold War was a real thing with real convening power, until each of its members discovered that pressure on a specific issue was stronger than the value of the abstract position. The cited thread items do not contain evidence that AI will follow the same arc. They contain the structural ingredients for it, and no more.
What we do not yet know
The available thread items do not contain the body of the SCMP essay, only its headline and URL. This article has not independently verified which specific policy levers, country case studies, or named officials the essay cites, and any reader looking for those details should go to the source directly. The cited items also do not specify what Singapore, Seoul, Tokyo, or any other middle-power regulator has written into binding AI rule books, nor do they disclose any forthcoming domestic cloud awards, AI compute standards, or model-audit regulations. Treating any of those as observable facts would overstate what the thread supports.
What the thread does support is narrower. The SCMP essay is making an argument, not announcing a procurement. The Beijing red-line warning is being used as a constraint on Japan's room to manoeuvre, not as a settled bilateral position. The nursing-care trade is two named Japanese specialists reportedly entering a market, with the specific companies, brands, and deal values not specified in the cited items. The Bitget wind-down is one exchange's compliance decision in one country. Read together they sketch the boundary of the third path. They do not prove anyone is walking it yet.
, Desk note: Monexus framed this as a structural question about middle-power agency rather than a US-vs-China horse race. We cited Chinese, Japanese industry, and crypto-trade reporting alongside the SCMP headline, labelled Monexus analysis as such, steelmanned the third-path case as the essay's authors made it, and resisted the temptation to promote either hype or dismissal.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://www.scmp.com/opinion/world-opinion/article/3362786/us-and-china-compete-over-ai-middle-powers-can-craft-third-path
- https://x.com/unusual_whales/status/2084664910101844074
- https://t.me/NikkeiAsia/21191
- https://t.me/CryptoBriefing/18530