CITIC Securities calls Tencent's AI spending "downside protected." I call it a reentrancy attack on shareholder value.
Let me walk you through the numbers with the same forensic lens I use on DeFi protocols. The logic held until the liquidity dried up.
Context: The Hype Cycle Repeats
Tencent just reported Q2 2026. Domestic gaming and advertising beat expectations. Operating profit, excluding new AI product investments, grew 19% YoY. Core business profit quality is improving.
Then the report drops the bomb: Tencent has clarified four major AI strategies. Management is supporting "more aggressive capital expenditures." CITIC Securities raised capex forecasts for 2026 and 2027 to HKD 215.7 billion and HKD 260 billion respectively. That's roughly $27.6 billion and $33.3 billion USD.
But here's the kicker: they lowered core net profit estimates for 2026-2028 by 5% to 9%. Expected core net profit growth? 2% in 2026, 3% in 2027.
Core: The Structural Deconstruction
Let me break this down like I would a yield aggregator with a suspicious rebalancing function.
First, the capex-to-revenue ratio. Tencent's total revenue for 2025 was approximately HKD 780 billion. If capex hits HKD 260 billion in 2027, that's 33% of revenue going into capital spending. In the crypto world, that's like a protocol allocating 33% of its TVL to a single liquidity pool with unknown counterparty risk.
Second, the depreciation trap. They acknowledge rising depreciation costs. New AI servers, GPUs, data centers—all depreciate fast. In my audits of AI-agent protocols last year, I found that teams often ignore the depreciation of their own infrastructure. The math works until the hardware ages.
Third, the "downside protection" claim. CITIC says the increase in capex will bring "some financial pressure" but can be alleviated by core business profit improvement. This is reminiscent of Terra's argument that Anchor's 20% yield was sustainable because of adoption. Code does not lie, but incentives do. The incentive here is to keep spending because the alternative—admitting the AI moat is weak—would tank the stock.
Let me quantify this. If core profit grows at 2% in 2026, that's roughly HKD 10 billion in additional profit. Meanwhile, additional capex year-over-year is about HKD 40 billion (from 2025 to 2026). That's a 4x gap. The so-called "alleviation" is a prayer, not a plan.
Trace the gas, find the truth. Here's the raw on-chain equivalent: Tencent is spending 4 units of gas to earn 1 unit of profit. That's a negative EV trade every time.
I've seen this pattern before. In 2022, I reverse-engineered the Terra/Luna oracle feed. The team kept minting LUNA to defend the peg, believing the loop would sustain itself. The exploit was in the trust, not the contract. Tencent is trusting that AI spending will create a moat that justifies the debt. But unlike a blockchain, a corporate balance sheet doesn't have a built-in slashing mechanism. Shareholders bear the full cost.
Contrarian: What the Bulls Got Right
I have to give credit where it's due. Tencent's core business is genuinely strong. WeChat ecosystem, gaming, advertising—these are cash cows with network effects that rival any DeFi protocol's TVL. The 19% operating profit growth (excluding AI) is real.
Also, unlike many crypto projects that raise capital and then fail to deploy, Tencent has a track record of building. Their AI products like Hunyuan have shown decent performance. The ecosystem moat is not just hype—it's weaved into the daily lives of a billion users.
But here's the blind spot: Entropy always wins if you stop watching. The AI race is not a winner-take-most game like social media. It's a capital-intensive arms race where every player is buying the same GPUs, hiring the same researchers, and fine-tuning the same open-source models. The moat is not the technology; it's the distribution. Tencent has distribution, but so do ByteDance, Alibaba, and Meta.
In my 2026 audit of three AI-agent platforms, I found a critical reentrancy vulnerability in their payment routing. The fix was simple: add a mutex. But the root cause was the rush to integrate AI without understanding the failure modes. Tencent is rushing. The capex is the mutex they're skipping.
Takeaway: The Accountability Call
Silence is just uncompiled potential energy. Tencent's AI spending is a bet that the future will be exponentially more valuable than the present. That may be true. But the timeline matters.
If this were a DeFi protocol, I'd flag the leverage ratio and recommend a circuit breaker. For Tencent, the circuit breaker is the stock price. If the ROI doesn't materialize within 18 months, the market will execute a mass liquidation.
I'm not saying sell the stock. I'm saying read the revert strings in the earnings transcript. The depreciation is real. The profit growth is slowing. The capex is accelerating.
Math does not have feelings. Neither do I.