Baillie Gifford says Chinese AI buyers may redirect software margins at home

The prevailing view suggests Chinese tech firms will continue adapting Western technologies, relying on established infrastructure and incumbent leads. However, this consensus, while valid at the engineering level, overlooks a crucial aspect: procurement. Buying committees increasingly prioritize approval risk, supply continuity, and board optics over pure technical benchmarks, reshaping the competitive landscape for AI software in China.

The investor note is really about who gets protected demand The obvious read is that Baillie Gifford is making a stock-selection argument: its China equities team is pivoting toward companies it believes can grow by inventing rather than adapting Western models. The more useful business read is narrower and less comfortable for foreign-linked AI vendors: if the policy emphasis described in the research becomes a procurement preference, the margin pool moves before product superiority is settled.

The work question is not whether Chinese AI tools are better across the board; it is whether buyers inside China face enough regulatory, political, or trade pressure to choose domestic-first software even when a foreign-influenced option appears technically convenient.

That makes this a regulation story before it is a technology story. The Baillie Gifford source summary points to “regulatory and trade headwinds” and a policy focus on domestic innovation; it does not show the enforcement machinery that would convert preference into revenue.

Still, enterprise procurement rarely waits for a formal ban when the direction of state policy is visible. In AI software, where model choice, middleware, data hosting, and integration partners can all be framed as strategic dependencies, the first-order effect is vendor selection; the second-order effect is that domestic integrators can preserve services margin that might otherwise leak to foreign-influenced stacks.

The source names a pivot but not the machinery behind it The weakness in the Baillie Gifford packet is exactly where an operator would need detail. It says the China equities team is pivoting toward domestic firms that prioritize original innovation, and it names Sophie Earnshaw as the investment manager making the case, but the supplied material does not specify subsidies, procurement rules, security reviews, licensing burdens, or state-backed research directives.

Without those mechanisms, the margin-shift thesis remains plausible rather than proven. A chief AI officer should read the note as an indicator of investor expectations, not as evidence that a particular software category has already been reassigned to domestic suppliers.

That distinction matters because “inventing, not copying” is a broad investment phrase, not an enterprise buying rule. A hospital system, automaker, bank, or manufacturer operating in China would still need to know which AI workloads are covered by localization expectations, which vendors qualify as domestic-first, and whether foreign intellectual property inside a product creates compliance risk.

Baillie Gifford’s source, at least in the reporting packet provided here, does not answer those questions. The omitted machinery is load-bearing: without visible policy enforcement, the story is portfolio rotation; with it, it becomes a protected demand channel for domestic AI software and services.

Why the Western-copying consensus misses the buying committee

The consensus take this rejects is that Chinese technology companies will keep adapting Western technologies because the incumbents have a lead and the infrastructure is already there. That can be true at the engineering level and still fail at the procurement level.

A buying committee does not select only for benchmark performance or developer familiarity; it also selects for approval risk, supply continuity, board optics, and the odds that a vendor will remain usable under future trade pressure. Baillie Gifford’s own framing, as summarized in the packet, puts regulatory and trade headwinds beside the shift toward original domestic innovation, which is the mechanism by which a technical comparison turns into a policy-filtered purchase.

The counter-read is that the asset manager may be overfitting a policy narrative to an investment thesis. “Inventing, not copying” is a clean story for a China growth portfolio, but it does not prove that Chinese buyers will abandon foreign-influenced AI frameworks, middleware, or applications where those tools are cheaper, more mature, or already embedded in workflows. The packet also provides no customer names, procurement examples, spending data, or independent analyst corroboration.

If the domestic-first claim cannot be observed in actual purchasing behavior, it remains a market story told from the seller side of capital, not a work story visible inside enterprises.

The under-noticed middle is the local integrator, not the model lab

If the thesis holds, the most exposed actors are not only foreign AI vendors trying to sell into China. The under-noticed middle is the layer of systems integrators, consulting teams, implementation partners, and internal IT groups that translate policy preference into working software.

When procurement tilts toward domestic firms, those teams gain leverage because buyers still need migration, customization, compliance documentation, workflow redesign, and maintenance. The margin does not simply move from one model provider to another; it can move from packaged foreign-influenced software toward domestic-first development and implementation work that sits closer to the buyer.

For executives, the labor consequence is subtle. A domestic-first AI procurement environment increases the value of employees who can evaluate local vendors, manage regulatory ambiguity, and rebuild workflows around tools that may not match the assumptions of Western software stacks.

It can also reduce the bargaining power of teams whose expertise is tied mainly to foreign vendor ecosystems. Baillie Gifford’s source does not say this directly, but its described pivot toward domestic inventors implies that the commercial premium may accrue to firms and workforces able to make indigenous products usable inside real organizations.

The next evidence should show up in procurement, not slogans The thesis is falsifiable. It would weaken if Chinese government procurement data for 2025 showed over 50% spend on non-Chinese AI software solutions, if major Chinese tech companies reported increased reliance on US or EU developed AI frameworks and middleware in their 2025 annual reports, or if there were a significant relaxation of Chinese regulatory penalties for using foreign AI intellectual property by mid-2025.

It would strengthen if the opposite appears: more domestic-first language in purchasing documents, more enterprise case studies where local AI vendors displace foreign-influenced tools, and more investor commentary linking China software margins to policy protection rather than only product quality.

The practical forecast, clearly labeled as analysis, is that within 24 months the important China AI software question will be less “who has the best model?” and more “whose stack can a regulated buyer defend?” That is not a claim that domestic tools automatically outperform foreign-influenced alternatives. It is a claim that Baillie Gifford’s reported pivot, if matched by procurement behavior, points to margin moving toward domestic-first development because policy can change the buyer’s risk calculation before engineering consensus changes.