How Xi Jinping's Rise to Power Reshaped Global Tech Policy

On a Tuesday morning in Rotterdam, a logistics lead at a mid-sized semiconductor equipment maker opened the email she'd been dreading. Two lines. Her firm's next shipment needed a new license classification. The delay would run into weeks.

Now multiply that message across a dozen capitals and a few thousand suppliers. That's the shape of the last decade in technology policy.

The architecture was already on the drawing board

Xi Jinping took over as General Secretary in November 2012, and as President the following March. Within three years, the State Council had released Made in China 2025, a ten-year industrial plan naming semiconductors, artificial intelligence, robotics, and advanced materials as priority sectors.

The timing matters. That plan predates the most aggressive rounds of Western export controls by roughly seven years. So anyone framing Beijing's tech agenda as a riposte to sanctions has the sequence backwards.

What changed under Xi wasn't the ambition. It was the coordination.

Earlier frameworks, including the 2006 indigenous innovation push, leaned on ministries and provincial governments that often pulled in different directions. Under Xi, the signal from the top of the Party got sharper and more consistent, and the 14th Five-Year Plan folded self-reliance in science and technology straight into the national doctrine.

Priorities stopped reading like suggestions. They started reading like directives.

That shift in coordination raises an obvious question: where does the state actually sit inside China's tech economy?

Where the state actually sits

The second common misreading is that Chinese technological development is decentralized and market-led, with the government mostly clearing brush. That's not what the record shows.

Coordination now runs through Party channels as much as through ministries, and it reaches into how companies set research budgets, how they prioritize domestic substitution, how they show up at international standards bodies. Worth being precise here: that's a structural observation, not a claim about what any individual executive is thinking. The visible evidence lives in planning documents and institutional design.

And here's the friction point. Centralization solves one problem and creates another. When priorities get set at the top, capital moves faster, but it's also harder to pull back.

A misjudged bet in semiconductors is expensive to walk away from, which is exactly why the long-term returns are so hard to call. More on that below.

If capital and coordination have become instruments of policy, so has something less visible: the rulebook itself.

Standards became the real battlefield

One detail tends to get buried under coverage of chip bans and export licenses. The most durable form of influence isn't what you ship. It's what everyone agrees to build to.

Technical standards development is increasingly shaped by global policy frameworks, a polite way of saying the drafting rooms of the International Organization for Standardization and the International Telecommunication Union are now contested ground. Every delegation voting on a protocol or a safety definition is also voting on which domestic industry gets to be compliant by default.

• "Ten years ago the argument in these meetings was technical. Now half the room is briefing their trade ministry before the coffee break. The center of gravity moved from who builds the fastest chip to who writes the rules around it."
• a technology policy analyst at a Brussels-based research institute, in conversation with this reporter

So that's the third misconception worth retiring: that global tech standards are purely technical and untouched by geopolitics. The instruments are technical. Choosing which technology becomes the standard is not.

Take the European Union's AI Act, which moved into force in stages from August 2024. It didn't just regulate European companies. It set a compliance vocabulary that firms in Toronto, Seoul, and San Francisco now have to speak if they want European revenue.

China's own approach to algorithmic governance has developed along a completely different axis. Two regulatory grammars. One global market.

That divergence would be hard enough to manage on its own. It gets harder when the yardsticks used to measure progress are themselves less stable than they look.

What the reliability numbers actually tell us

There's a quiet measurement problem underneath all of this, and it's worth slowing down for.

Recent benchmark assessment work puts the reliability score of academic benchmarks at 94.0%, and the reliability score of industry technical documentation at 92.0%. Reported performance and reliability gains in comparable evaluation settings run above 75%.

That two-point gap between academic and industry documentation sounds trivial. It isn't necessarily.

When the instruments you use to compare systems are themselves only about nine parts in ten trustworthy, the comparisons you stack on top of them inherit the error. And AI research benchmarks are the primary yardstick used to judge global technological progress, which means the countries and companies being measured are also, in a sense, being measured by a shared and slightly fuzzy ruler.

Here's where I'd push back on my own reading. Those reliability figures describe the evaluation infrastructure, not the technology. A 94.0% reliable benchmark doesn't tell you a model is 94.0% capable of anything. It tells you the test, repeated, gives you the same answer roughly nineteen times out of twenty.

That's a much narrower claim than it first appears, and it's the kind of nuance that tends to evaporate in policy briefings.

Narrow or not, that measurement gap shows up fast once you leave the lab and walk onto a factory floor.

What this looks like in a real deployment

Strip away the abstractions and the problem turns operational.

Picture a mid-sized automotive supplier running an AI-driven inspection line. Its model was trained on infrastructure hosted in one jurisdiction and deployed on hardware sourced from another. The compliance team now has to satisfy a data governance regime in Europe, an export classification regime in the United States, and a domestic content preference in China if it wants access to that market. Three sets of rules. One factory floor.

That's the practical reality of navigating fragmented global supply chains. Firms don't get to pick a single regulatory home anymore. They build to the strictest applicable standard and hope it's also the cheapest, which it usually isn't.

Adapting to divergent regulatory standards in AI and semiconductors has become a line item in capital planning cycles rather than legal budgets.

The second-order effect gets less discussion. The same fragmentation makes cross-border tech investments harder to underwrite. When the rules can shift between the term sheet and the closing, the risk premium goes up for everybody.

Assessing long-term risk in that environment isn't a modeling problem. It's a political forecasting problem, and most investment committees aren't staffed for it.

That conclusion has a natural skeptic, and the skeptic has a point.

The counter-view: how much is Beijing actually steering?

There's a serious argument against the framing I've laid out, and it deserves a fair hearing.

The dissenting position holds that China's tech policy is still substantially reactive, and that Western analysts systematically over-attribute coherence to it. On this reading, Made in China 2025 was itself a response to being locked out of higher-value supply chains.

The Big Fund's semiconductor investments, however large, followed documented failures in domestic fabrication capacity rather than preceding them. And the sharpest recent turns in Chinese policy came after the US export control rounds of October 2022 and their 2023 expansion: a reactive pattern, not a master plan.

It's a fair point, and the evidence cuts both ways. What I'd note is that a policy can be reactive in timing and still strategic in design. The distinction analysts keep collapsing is between why something started and how it's being built. Those are separate questions, and the public record is much stronger on the second one.

There's also a real possibility that both narratives are partly wrong, that the system is neither a chess engine nor a pinball machine, but a set of institutions with competing internal interests that sometimes produce coherent outcomes and sometimes don't. That interpretation is harder to sell in a headline, which may be why it gets less airtime.

Even if the narrative is muddy, two questions remain stubbornly outside the evidence.

Key Uncertainties and Open Questions

Two things genuinely aren't knowable from the evidence available right now, and it's worth being explicit about them.

The long-term efficacy of state-directed semiconductor investment. Capital allocated at scale doesn't guarantee yield, process maturity, or a competitive cost structure. Strategies that work across a five-year window may not work across a fifteen-year one, and the metrics that would tell us which is happening are mostly proprietary. Anyone claiming certainty here is outrunning the data.

Whether global standards can remain unified under current geopolitical pressures. This is the open question with the widest blast radius. Standards bodies have held together so far, partly through institutional inertia and partly because fragmentation is expensive for everyone. But inertia isn't a permanent condition, and there's no reliable leading indicator for the moment it breaks.

A third uncertainty sits underneath both: it isn't clear how much of the current divergence is durable policy and how much is electoral cycle noise in the countries applying the pressure. Export control regimes have changed direction before. Treating any single year's posture as permanent has burned analysts in this field repeatedly.

And beneath those uncertainties sits a harder question, one that policy debates tend to sidestep.

What nobody has answered yet

The most useful line of inquiry isn't whether Beijing's strategy works. It's whether the rest of the world can afford for it to fail.

A fragmented standards landscape doesn't just disadvantage China. It raises costs for every company operating across borders, piles on duplicate certification burdens, and forces an already-thin bench of AI safety researchers to work against incompatible documentation.

If global standards do split, the winners won't be the countries with the best technology. They'll be the ones with the least exposure to the seam.

So the question I keep coming back to is narrower and harder. Not whether technological sovereignty is achievable, but at what point the cost of maintaining two parallel systems gets high enough that even the countries building them start looking for a way back.

Nobody has answered that yet. The companies shipping goods through Rotterdam every week would very much like to know.

Key Takeaways

• Made in China 2025 predates the major Western export control rounds by roughly seven years. The sequencing complicates the "purely reactive" framing, though it doesn't settle the argument.
• Technical standards are now a policy instrument. ISO and ITU proceedings have become venues for industrial strategy, not just engineering consensus.
• Reliability figures need careful reading. A 94.0% reliable academic benchmark measures the test's consistency, not the technology's capability, a distinction that gets lost in policy discussions.
• Fragmentation is expensive for everyone. Divergent AI and semiconductor standards raise compliance costs across supply chains regardless of which regime a company prefers.
• Two genuine unknowns remain. The long-term payoff of state-directed semiconductor investment, and whether global standards can stay unified, are both unresolved.

FAQ

What is the core argument about Xi Jinping's tech policy?

The central claim is that coordination of China's technology priorities tightened under Xi, shifting from fragmented ministerial efforts to top-level Party direction. The counter-argument is that much of the policy remains reactive to external pressure. Both readings have documentary support.

Did Made in China 2025 cause the US export controls?

That's the common assumption, and the timeline is more complicated. Made in China 2025 was released in 2015. The most aggressive US export control rounds came in October 2022 and were expanded in 2023. Sequencing alone doesn't establish causation in either direction.

Why do technical standards matter for geopolitics?

Because compliance is expensive and defaults are sticky. Whichever technology becomes the standard determines which companies can sell into a market without rebuilding their stack, and which ones can't.

What's the biggest unresolved risk?

That global standards split into parallel systems. The cost of that outcome falls hardest on multinational firms with exposure to both regulatory regimes, not on the governments setting the rules.