To invest in China AI, just follow the money

What happened

Bloomberg reported that Tencent and Alibaba are in talks to invest in DeepSeek's first external funding round, with a valuation benchmark around $40bn. Tencent proposed acquiring up to a 20% stake, though DeepSeek is pushing back on ceding that much control.

This tells something important about where the value chain is heading: the mega-caps can't build everything themselves, so they're buying into the companies that can. That's the pattern that will repeat across the ecosystem, especially in verticals.

Impact on our Investment Case

Mega-caps: spending big, monetizing slow

The 15th Five-Year Plan commits roughly $2tn in cumulative tech spending through 2030. That policy mandate, combined with the competitive arms race between Alibaba, Tencent, and ByteDance, is forcing massive capital deployment.

Following in the footsteps of their western peers, Alibaba has pledged RMB380bn (~$52bn) in AI and cloud infrastructure through FY2028, while Tencent's capex hit RMB79bn (~$11.5bn) last year and is set to rise. Neither can afford to fall behind. 

But the spending is running ahead of the revenue. Alibaba's net income dropped 66% in 2025 as capex surged and API prices were slashed to win adoption. Costs arrive now, revenue follows in two to three years. In that window, the value sits not with the platform builders, but with the companies applying AI in specific industries.

Mid-caps: where the money actually lands

As the Chinese government mandated 90% AI adoption across 150,000 State-Owned Enterprises (SOEs) by 2030, who gets the contracts?

Not Alibaba Cloud directly. SOEs buy verticalized solutions from companies that understand their specific operations. iFlytek didn't win AI deployment across State Grid's 1,100+ cities because it had better models; it won because it spent years building the operational relationships. That kind of moat doesn't get replicated by throwing cloud credits at the issue.

The regulatory framework reinforces this. Big Tech must direct 30% of R&D toward national priorities, open their APIs to competitors, and source 70% domestically by 2027. Beijing is deliberately pushing value from platforms down to ecosystem players. The mega-caps become the infrastructure layer; the mid-caps become the monetization layer.

Could the mega-caps choose to build vertically rather than buy? In some domains, yes. But the data moats in healthcare records, industrial IoT, and government fintech take years to accumulate. What's more likely, and what the DeepSeek talks illustrate, is the flourish of strategic investment and partnership: the mega-caps provide capital and infrastructure, the mid-caps provide domain expertise and customer access. Tencent sits on roughly $60bn in net cash. That money needs to go somewhere, and Beijing's framework is designed to channel it into the ecosystem, not back into platform expansion.

The valuation setup

The gap between Chinese and US tech has narrowed since our earlier reports, as US tech has de-rated from its peaks, but it remains significant on a growth-adjusted basis.Hang Seng Tech trades at roughly 21x trailing earnings vs the Nasdaq-100 at around 35x. The Magnificent Seven's PEG ratio sits near 1.7. MSCI China overall trades at approximately a 40% discount to developed markets. Our ChinaTech index trades at 21.6x P/E with revenues showing a 14% 3Y CAGR, and PEG sitting at 1.00, still well below the US equivalent.

Why hasn't this re-rated? Because most allocators still price "China tech" as the 2020 trade: mega-cap platforms with regulatory and geopolitical risk. The mid-cap cohort with domestic revenue and policy-driven demand is a different animal, but foreign ownership near historical lows tells the market hasn't made that distinction yet. A renewed sanctions cycle around the late 2026 midterms or a sharp yuan move could delay the re-rating, but the earnings trajectory is independent of either.

Looking at the most widely held China ETFs, like KWEB China Internet or Invesco China Tech illustrates the structural mismatch. KWEB's top four holdings are Tencent (10%), Alibaba (9.4%), PDD (8.2%), Meituan (7.3%), in other words exactly the mega-caps whose margins are under pressure. Its YTD performance (as of 22 april) is -14%. Our strategy, built from a 1,083-company proprietary universe with a mid-cap tilt and pure A/H-share exposure that includes many SOEs, has returned +10.3% YTD and +50% since inception in February 2025. The edge comes from being positioned where policy capital flows, not where index funds passively allocate.

Liquidity as accelerant

Monetary policy in China is accommodative, with money-market rates near multi-year lows and subdued inflation leaving ample room for further easing. This matters more for mid-caps than for mega-caps: mid-caps tend to carry more debt and their valuations are more sensitive to discount rate changes. When deposit rates hit historic lows, investors move down the cap spectrum looking for growth.

The 15th Five-Year Plan's growth target of 4.5–5% is the lowest in nearly 30 years, reflecting Beijing's deliberate shift toward quality over headline speed. But the RMB1.3tn ultra-long bond program announced in March funds infrastructure and tech spending off-budget, creating fiscal room without headline deficit expansion. The money flows into exactly the sectors our portfolio captures.

Our Takeaway

The logic is straightforward. Beijing mandated AI self-reliance and backed it with $2tn. The mega-caps are deploying that money into infrastructure, compressing their own margins in the process. They can't verticalize fast enough, so they'll partner with and eventually acquire the mid-caps that hold the domain expertise. Policy channels the spending through those same mid-caps. Cheap liquidity amplifies the bid. And the market hasn't re-priced any of this yet.

We launched at $100 on February 4, 2025. We're at ~$160 today. The structural drivers suggest the party has only begun.